AI in Marketing

Prof. Zillur Rahman

Department of Management Studies
Indian Institute of Technology, Roorkee
Week-8
Lec 36-Personalization and hyper-personalization Using AI-IV

Welcome to this NPTEL online certification course on artificial intelligence and

marketing. Now we will talk about module 36 and as you can see from this slide, we are
still talking about and discussing personalization and hyper-personalization with the help
of artificial intelligence. So this is part 4 of this discussion, personalization and hyper-
personalization using AI. And these are the things that we will cover in this module. We
will start with the need for hyper-personalization. Then we will talk about hyper-
personalization customer journey map for different sectors.

Thereafter, we will understand the steps to be followed to start off with hyper-
personalization and the risk involved in hyper-personalization and how can we mitigate
those risks. So why does hyper-personalization matters? The first is, the first reason for
that is organizations are now facing competition from non-traditional digital first brands.
Traditional marketing and business models are becoming increasingly outdated as digital
data first and direct-to-consumer brands penetrate the marketplace. This is making
digital marketing increasingly crowded leading to increased spending in hopes of
reaching consumers.

And as a consequence of the increased intensity of outreach, 3 out of 4 customers have

indicated they receive too many email promotions from brands and 69 percent have
unfollowed brands they once followed on social media. These trends point to a lack of
consumer interest in the content or offers they are being served. Even traditional
businesses like McDonald's are adopting through the acquisition of companies like
Dynamic Yield to personalize their drive-thru experiences. Digital-first companies like
Amazon, Facebook and Google are leading the charge into new markets through the use
of their comprehensive customer databases. By understanding their customers, they are
able to target very specific customer segments using the right media and content.

This technology uses external data such as weather, trending topics on social media,
products, time of the day, seasons combined with customer order patterns to recommend
additional products for that particular time and place. Given feasuous competition, digital
advertising is becoming increasingly expensive. Studies have indicated that over the last
2 years, digital ad spend had increased 12 percent with no discernible increase in the
results. So, now as you can see the rate of return on the digital spend is coming down
because of increased competition. To effectively compete organizations need to make a
meaningful impression on their customers taking full advantage of the limited exposure
available.

So, what have what is happening is that with increased competition, the exposure time
available to a company is decreasing. Exposure time per ad is decreasing. Therefore, they
have to take increase the effectiveness of this. Hyper personalization is a means to cut
through the node noise and provide customers exactly what they are looking for. So, this
is done by way of hyper-personalization.

What they are look it assists customers with decision-making and fosters deeper
relationships that will keep them from seeking out competitors. The second is that the
technology is enabling advanced marketing solutions and consumer interaction. With
more connected devices and robust data models, organizations are finding unique means
of collecting data and connecting with customers. Historically consumer data could only
be gathered at point of sale and was mostly expressed by customers themselves. So, that
was the problem with historical data.

Earlier the data was collected only at point of sale and then what happens nobody knew.
However, the usage of online tracking through cookies and other means allow brands to
gain a deeper understanding of customer preferences and behavior. They are there are
various tools which enable this level of customer data to be effectively utilized including
customer data platforms and loyalty programs. This data can then be combined with
third-party data from social media and censuses and ethically shared among peers to
establish a detailed description of customer and understand them far beyond traditional
customer segments. Data collection has also been augmented through the advancements
of analytics and AI technology.

The combination of data and technology is disrupting the traditional means through
which organizations interacted with the customers. Data and technology are now enabling
organizations to reach customers through targeted media with content that is relevant and
uniquely tailored to them. So, now that data which is uniquely content that is uniquely
tailored to them and also relevant to them at the same time. Well-executed hyper-
personalization can deliver 8 times the return on investments of marketing spend and lift
sales by 10% or more. 22% of customers are happy to share some data in return for a
more personalized service or product.

Now let us look at how this is happening in financial sector. So, we are talking about the
customer journey map. Personalization is a key enabler to driving customer centricity
across the value chain for financial institutions. Because different customers have
different needs at different point in time and also different ways to reach and buy from
the market. As traditional financial institutions face increasing competition from
technology-first organizations they are looking for ways to retain current customers 1 and
attract new customers 2.

Their strategy includes personalized product design, tailored advice and

recommendations, pricing optimization and tactical communications and servicing
strategies. All of these contribute to future-proofing the institution while integrating
customer-centricity. So, these are the two things that organization is future-proof also and
customer-centric also at the same time. Or you can say that taking customer centricity
path to become future-proof. Gathering contextual data generates information about
customers profile, consumer behavior and transactions which can help not only with
profiling and segmenting customers but also drive more targeted offers.

Financial organizations have deployed tools like external data, voice analytics and natural
language processing to capture their clients profiles and sentiments through phone, in
person and digital interactions. This has enabled them to become increasingly more data
driven to better understand their customers and predict varying customer demand. So, this
is the financial sector customer journey map. So, it starts with researching financial
products. AI allows companies to ensure content is tailored to the customer making it
easier for them to find the information they are looking for.

So, easy for customers to find what they are looking for. Then it comes there comes
meeting a financial advisor, real time analytics and customer insights will ensure advisors
know exactly what customer needs. So now, this these advisors are also well educated
about the various products about the company's products and also each customers need.
So, they now they are well equipped to tell which product is more suitable for which
customers. The next step is to develop a personalized offer.

Granular customer data allows for the tailoring of financial offerings for specific
customers, reducing risk and improving their experience. The next story is that of retail
sector. So, now we are talking about the customer journey map in the retail sector. The
availability of customer data allows marketing professionals to understand their
customers desires before they even enter the marketplace. These insights are being used
to promote their brands in an relevant manner attracting customers with a higher
probability of conversions.

So, now they are attracting customers with high probability of conversion. Reliable offers
use multiple different levels such as pricing, advertising and bundling. Once a customer
enters the marketplace or a retail location, personalization can be used to ensure their
shopping experience is unique to them. Each customer have a unique experience. This
can be seen through the use of virtual shopping associates or customizable products.

Allowing for these experience changes retail from being a shopping destination to
becoming an online activity for customers to experience the brand. So, it is not about
shopping but it is about experience. This is only accelerated with the rise of omnichannel
brands. Customers are now equipped with the tools to digitize their experience even
inside a physical retail location bringing their online customer experience into the store
through the use of technology. This starts with using behavioral analytics to sort products
in a way that maximizes conversation.

Once a customer walks into the store, he can use his or her mobile technology to get
product recommendations and be directed towards various products and departments.
Again, now let us look at the customer journey map in retail sector. So, starting with
relevancy through loyalty programs mapping customers and understanding their needs
and predicting segment changes early allows organization to provide targeted offers. So,
with this customer understanding, now companies can come up with targeted offers. The
next step is tailored web content.

Companies are able to offer tailored product and recommendations and insights by
changing website content in line with the customer's profile. So, depending on the
customer's profile, they are changing the website content. So, it may be so that each
customer is looking at a different content on the same website. And then the next step in
this journey is personalized pricing. Knowing a customer's willingness to spend and
external motivators allow organizations to use dynamic pricing and increase the
conversion rate.

The next customer journey map is about the healthcare sector. As healthcare technology
advances, so too does the depth of interaction that healthcare professionals have with
their patients. Health being a very personal matter naturally lends itself to hyper-
personalization to ensure each individual is getting the specific treatment he or she
requires. So, as with the case of financial products, similar is the situation here in
healthcare that the needs are very specific.

What fits A may not fit B. So, because of this, the scope for hyper-personalization
increases tremendously. This industry can be broken down into health consumers and
healthcare patients. Personalization in this sector helps customers find what they are
looking for based on their unique health and behavioral needs. In a similar way to retail,
an example of personalization is the use of recommendation engines to suggest healthier
alternative products based on past behavior and gathered data. This can be further
extended into general wellness through nutrition fitness, meditation and support tools.
The next step in personalization in the healthcare world is to create a 360-degree view of
health that encompasses both consumers and patients. Equipping healthcare professionals
with a person's lifestyle data and empowering individuals with their clinical data
contribute to seamless service and personalized healthcare experience. So, in healthcare,
this is personalization is all the more important. In this world, customers would receive
recommendations based on their medical history and healthcare professionals would be
able to prescribe care plans based on a patient's lifestyle preference. So, this is the
customer journey map in healthcare sector and it starts the first step is data consolidation.

Bringing together scattered pieces of healthcare data from visits to the doctor, the
pharmacist, wearable technology and demographics. So, all this whole lot of data is
integrated. Whole lot of data from these disparate places and that leads to care
recommendations. Utilizing customer health data to provide personalized diets and
exercise routine to fit not only these consumers lifestyle but also their constraints. So,
each patient will obviously need a different plan.

Depending upon their lifestyle and healthcare conditions and that leads to personalized
care. Although correctly tightly regulated, AI shows the promise of being able to develop
personalized medicine. But still this this space is tightly regulated. Now, when we are
talking of hyper-personalization, where to start with? So, the first thing is to identify
customer needs. The first step to improving customer experience with hyper-
personalization is understanding the customer himself or herself.

This no longer means simply high level demographics such as age, gender and education
level but rather calls for a deeper layer of customer data to be gathered and analyzed. So,
these are some superficial data. We are looking for some deeper layer of customer data.
This includes behavioral, psychographics, geographic and demographic data and allow
for the targeting of each segment and person differently and for more personalized and
relevant communication. The various layers of information through which we can better
understand the customer's needs can be sourced from a variety of first and third-party
data.

Each source provide a deeper insight into a customer's preference and thus boost
personalization efforts. Each source is helping in personalization effort. First party
information is that which an organization already collects and owns. This type of
information includes data on previous interaction with the company, shopping behaviors,
type of product customers are made favorable along with any demographies gathered as a
result of doing business. So, for example, a customer is using a credit card or home
delivery.
So, lot of this data is gathered about his address and other data from this credit card
details. This layer of information provides more limited insight but can be used to clarify
and distinguish customers who drive the most value. So, now we are not looking at at
huge quantity of customers. But we are looking at to distinguish customers who drive the
most value. As well as internal information, organizations can also obtain and harness
external data.

While internal data is limited to interactions with the organization. So, this is the
limitation of this internal data. External data allows organization to fill in the gap of
knowledge and identify behavioral traits that may not have been obvious from an
organization's own interaction with their customers. Simple example includes social
media posts, credit scores and corporate partnerships. This enriched data sets can then be
used to tailor messaging, design products or develop content that is more relevant to the
customer.

The combination of these two sources of data leads to the personalization even allowing
brands to respond to dynamic life events. So, now the brands they are themselves
becoming dynamic. Dynamic real-time data ensures customer information is up to date
and offers the ability to target customers with relevant messaging at the right place and
time for optimal conversion. The second is understanding the current state of data and
technology. Once an organization understands the role personalization play in this
strategy and how it aligns to its customers needs, implementations will require the right
data and also the technical capabilities.

Whether it be tracking customers on e-commerce platforms or using facial recognition as

a means of providing additional security, the art of hyper-personalization starts with how
information is gathered, analyzed and used. So, this art of hyper-personalization is
dependent on how we are gathering, analyzing and using the data. A personalization
strategy alone is not enough and will need to be enabled by the appropriate data
technology and architecture. So, strategy needs to be implemented by way of data
technology and the third is architecture. Once the organization knows its data
requirements, it will need to collect that data which can be done through the appropriate
combination of customer tracking, purchasing of third-party data or capturing data
throughout the customer journey.

This also requires organization use data platform such as customer relationship
management systems and customer data platforms. When the relevant data is gathered,
implementing a method of targeting customers with the correct content can be achieved
through technology. Knowing where to meet customers and the channel through which it
can be done is a crucial part of this process. This can be something as simple as
personalized email marketing with names and relevant products included based on
previous purchases or something much more complex such as custom landing pages on
the website. Some examples of these technical solutions include personalized
recommendation engines, chatbots and serving specialized ads on social media to a
narrower audience.

The third is creating a roadmap for success. Hyper personalization is not a single piece
of technology that can be installed and immediately work for every organization. It
involves a strategy that affects multiple elements in the organization's business and
requires a consolidated technical solution. It is crucial for an organization to develop a
roadmap to decide which personalization element should be implemented, when and
how. Although there are various personalization tactics that can be deployed throughout
the customer journey, it is important that organization recognize the importance of
prioritization, which elements should come first. Obviously, those elements which are
more important for the customer and also which the company is better at doing.

This is done by balancing the impact of each effort with the cost of implementation.
Impact can be financial that is the return on investments, the impact on customer
experience or how it changes the competitive landscape. Some personalization efforts
carry greater benefits than others and it is important that an organization know what it
wants to achieve. If a brand is looking to improve customer loyalty, it is more likely to
focus on improving the customer journey by making the shopping experience easier or
more connected through personalized communication. Alternatively, if a brand is
looking for quicker conversion, it can use personalized offers and pricing to entice
customers to make an impulse purchase without going through the decision-making
process.

As the demands of the organization and customer vary, so do the types of personalization
solutions that need to be used. Hyper personalization is all about timing and calls for the
constant gathering of right data at real time basis to make informed decision. So, it is a
constant gathering of data it is not one time because the customers they keep on evolving.
This means that the cost of implementation will vary depending on an organization's
maturity and willingness to invest. To make the best use of resources, organizations
should target low cost and high impact solutions that will drive benefits in a manner that
is least disruptive to the customers and their needs.

It is important that an organization choose the solutions that are best suited for its brand
image, customer relationship and the industry itself. Now, let us look at the hyper-
personalization risks. The first is reinforcing biases. If the generative AI models are
trained on biased data, then obviously hyper personalization could reinforce existing
biases and stereotypes leading to discriminatory or exclusionary recommendations. The
next is limited exposure to diverse viewpoints that could create a filter bubble effect
where users are only exposed to content and recommendations that reinforce their
existing beliefs and values, limiting their exposure to diverse viewpoints.

The next is over reliance on technology that could lead to an over reliance on technology
where users become less capable of making decisions without the help of AI generated
recommendations. The next are the privacy concerns. Hyper-personalization requires
collecting and analyzing large amounts of personal data which could raise concern about
privacy and data protection. Then comes security risk. With more data being collected,
there is an increased risk of security breaches, hacks and data thefts.

The next risk is susceptibility to manipulation. Could make users more susceptible to
manipulations, particularly if the generative AI models are misused for political or
commercial purposes. Reduced serendipity could limit serendipity as discoveries or
surprise recommendations that could broaden a user's horizon. Loss of human touch
could lead to a loss of human touch where recommendations become purely algorithm-
driven and lack the personal touch that comes with human interactions. The next risk is
bias amplification. Hyper personalization could amplify biases that exist in the data
leading to recommendations that are even more biased than the original data sets.

The next is undermining trust. If hyper-personalization is not transparent, it could

undermine trust in the technology and the companies that use it, leading to a backlash
against AI-generated recommendations. The next is transparency. Be transparent about
the data being collected, how it will be used and who it will be shared with. Explain to
the customer the reasoning behind personalized recommendations.

Content generations or decisions and also solicit active feedbacks. Consent. Obtain
explicit consent from customers before collecting and using their personal data for hyper-
personalization. Allow customers to opt out of personalized experience if they choose to
do so. Data security. Explain strong data security measures to prevent data breaches or
thefts.

Ensure that data is encrypted, access is controlled and they regularly undergo security
audits. Explainability. Ensure that hyper-personalization algorithms are transparent and
explainable. It should be possible to explain how the algorithm arrived at a particular
decision or recommendation. The next step that a company can take to mitigate hyper-
personalization risk is diversity and inclusion.
Take steps to eliminate biases in data and algorithms and ensure that personalized
experiences are designed to be inclusive and diverse. Provide personalized experience for
a broader range of customers and take steps to avoid stereotypes. The next thing that can
be done to mitigate hyper-personalization risk is human oversight. Provide human
oversight of hyper-personalization algorithms. Ensure that humans are involved in the
decision-making process and that they can step in when necessary to prevent errors or
biases.

Ethics and accountability. Develop ethical guidelines for hyper-personalization and

ensure that they are followed. Hold individuals and organizations accountable for any
breaches of trust or privacy. Model monitoring. Ensure that models are monitored and
that preferences are not favoring toxic or bias views.

Ingress or aggress filters. Filter content coming into the system and going out of hyper-
personalization system to ensure that bias is minimized. Reasoning models. The usage of
models that can reason about truth and discern facts, opinions, beliefs and prejudice from
one another can more easily eliminate biases and reduce hallucinations as they are able to
discern facts from personal opinions, beliefs and prejudices. So, to conclude in this
module, we have first studied the need for hyper-personalization. Then we have
understood a customer journey map of hyper personalization from three different sectors.

Thereafter, we have also understood the steps to start off with hyper-personalization.
And finally, we have studied about the risk involved in hyper-personalization and ways to
mitigate the same. And these are the seven sources from which the material for this
module was taken. Thank you.
 AI in Marketing
Prof. Zillur Rahman
Department of Management Studies
Indian Institute of Technology, Roorkee
Week-8
Lec 37-Personalization and hyper-personalization Using AI-V

Welcome to this NPTEL online certification course on artificial intelligence and

marketing. And now we will talk about module 37. So, as you can see that this is the last
module in which we will continue with the discussion and understanding of
personalization and hyper personalization using AI. So, this is part 5 and we are talking
about B-to-B perspective in this module. So, in earlier modules, in earlier 4 modules we
have talked about B to C perspective, now we are talking of B to B perspective. And
these are the things that we will cover in this module.

In this module, we will start with discussing about personalization using AI in B-to-B
perspective. What steps need to be implemented? What steps are to be used to implement
AI for personalized B to B customer experience? The rise of AI in B-to-B e-commerce
and steps to integrate it in B-to-B e-commerce. How various AI tools are helping in
personalization of B-to-B e-commerce buying experience? How AI aided CPQ is
benefiting B to B e-commerce? The best practices to be followed and challenges that are
associated with implementing AI in B to B selling and the future of AI in B to B sector.
So, now let us look at personalization in B to B that is business to business.

Modern business to business buyers seeks the same self-service options they enjoy in
their personal shopping experiences, even when buying complex products. So, now these
buyers they also want to map the same kind of experience that they have in B to B to B to
C that they have in B to C to B to B. Yet many B-to-B companies still treat their websites
more like static catalogs than dynamic sales engines. More and more B-to-B enterprises
are leveraging artificial intelligence to bridge the gap between outdated ways of selling
and current customer expectations. Spurred by technological breakthroughs like the
introduction of chat GPT, AI is no longer just about automating simple tasks or content
creation.

It is now a linchpin for delivering personalized experience that empower sales team to
operate more effectively. Empower sales team, personalized experience that empower
sales team. So, now these sales people are also treated as customers. So, they can they are
also delivered this personalized experience so that they also feel empowered and become
more effective in making a sale. AI is definitely becoming an indispensable asset in B to
B e-commerce with whether it is enhancing sales strategies through guided selling or
streamlining conversations on price.

AI powered configure price code tools like visual configurators. Now these are the steps
to implement AI for personalized B to B customer experience. Step 1 is to identify your
goals. Before implementing AI, it is essential to identify your goals for personalized B to
B customer experiences. Are we looking to increase customer satisfaction? That is one
goal.

Boost sales to or improve efficiency. What are we trying to do? Understanding our goals
will help us identify the right AI tools and strategies for the businesses. So here also as
we were discussing in the last module, there is a need to understand the importance of of
priorities. For example, customer satisfaction can be given 50 percent, sales can be 30
percent and efficiency can be 20 percent. So, it is not either or or.

It is the different priorities that are given to the different goals that the company has.
The second step is to collect and analyze customer data. To provide personalized
experience, we need to collect and analyze customer data. This can include data on past
purchases, website interactions and the feedback. Analyzing this data can help us identify
patterns and preferences that can be used to provide tailored recommendations and
solutions.

So, keep in mind that we are always on a lookout for tailored recommendations and the
solutions. Step 3 is choose the right AI tools. There are many AI tools and platforms
available. So, it is essential to choose the right one for your business. We should
consider factors again such as cost, ease of use and customization option.

Again, here it is important to prioritize. Option cost can be 40 percent, ease of use can
be 30 percent and customization option can again be 30 percent. So, this makes it 100
percent. Step 4 is to train your AI algorithm. Once you we have chosen the AI tool, we
will need to train the algorithm on our customer data.

This process can take some time, but it is essential to ensure that our AI is providing
accurate and relevant recommendations and solutions. So, this time is not important here.
More important is providing accurate and relevant recommendations and solutions. The
fifth step is to test and refine your AI. After implementing the AI, it is essential to test
and refine its performance continuously.

This can involve analyzing customer feedback and adjusting our algorithms as necessary
to provide the best possible personalized experience. Now, let us look at the rise of AI in
B2B e-commerce. While AI has achieved buzzword status in 2023, it has been
transformative force in e-commerce for over a decade. With milestones ranging from
chatbots to recent innovations like AI-driven content and supply chain optimization, AI
has evolved from nice to have to a necessity in today's B2B e-commerce landscape. For
example, in the area of e-commerce data management, AI has significantly reduced
manual labor by automating the ingestion, cleaning, structuring and enrichment of
product information.

These streamlines operations and enhances the customer journey through personalized
experiences. AI also enables businesses to gather and analyze customer data more
effectively, including the use of zero-party data to better understand and meet customer
needs. This in turn boosts engagement, satisfaction and sales. The next is to look at how
to integrate AI into B2B e-commerce. Inaccurate, outdated bad data cost e-commerce
businesses money.

For instance, the Nels and Norman Group, founded 20% of unsuccessful e-commerce
purchases, are a result of a lack of product information. But having comprehensive
product descriptions, content and images is not enough. We also need optimized and
enriched data to unlock their full potential. Optimized and enriched data. This
underscores the importance of data cleansing, structuring and enhancement processes to
maximize the e-commerce effectiveness.

For example, data cleansing removes errors and standardized units for weight, size,
temperature and speed. When combined with data structuring and enrichment, this
process enhances product visibility, improves click-through rates and boosts
conversations. We will discuss some steps that need to be followed to be accurate,
complete, consistent, relevant and up-to-date with our data to make the most out of it.
So, let us start with content ingestion. AI can also assist you with integrating your content
and data, whatever the format or source.

This replaces the need for dedicated marketing or productives to digitize resources
manually. Instead, platforms like ZOO, Automate Design, this process to help mitigate
human error. So, you increase efficiency, save time and provide more accurate product
data to build brand loyalty. Then comes cleansing. Having quality product data is
essential for a streamlined, efficient e-commerce business, especially for multi-channel
companies.

Data cleansing helps you remove duplicate content that may negatively impact your
ranking, identify and monitor inaccurate product information and update incomplete
information. Data cleansing is not a one-time activity, but it is an ongoing process. It will
ensure your data stays up-to-date and so customers have access to accurate information to
facilitate their buying journey. Structuring your data should be meticulously organized
and standardized to enable tools to efficiently search and understand the information.
Take customer searches which often hinge on categories like type, brand and price.

A simple error like inconsistently labeling the same product type as both web-hosted and
cloud-based can result in inaccurate product recommendations. Similarly, mixed price
formats such as $200 or $200 can compromise the data and ultimately impact the
customer experience. Enrichment. Product data enrichment enhances your raw product
data. Tools can also help translate technical specifications into understandable language
customers can use in the context of their needs.

They can also use feedback from customers to identify how to add value to your existing
product data. Personalized buying experience. Some AI tools that can help adopt and
personalize customer e-commerce buying journey. The first is guided selling assistant.
AI-powered guided selling assistant can enhance customer experience and augment the
abilities of best salespeople on the e-commerce side.

These tools helps customers find the right product using need-based questions. These AI
guided selling assistants are fast, seamless and automate one-to-one engagement and that
too at scale. Customers can communicate what they need and get instant real-time
personalized recommendations, making it easier to find what they want and in thereby
increasing their satisfaction. Intelligent searches. Customers do not want to search for
product using item numbers and exact product names.

So, these are the two things that customers do not know or remember at the time of
searching. Intelligent searches make it easier for customers to find what they need faster
using the way they naturally search. Intelligent search tools use natural language
processing technologies like contextual understanding and semantic search to improve
the search process and get information and answers from various sources without
necessarily needing to exactly match the information. It uses natural language
technology to identify means and make correlations providing a better customer
experience. It uses natural language technology to identify meaning and make
correlations providing a better customer experience.

The next is visual product configuration. Visual product configurators replace the need
of searching through outdated sales spec sheets to determine if the product is compatible.
Instead visual configurators ensure the product bundles, designs and configurations are
all possible through its understanding of the connections and dependencies between your
product catalog. These visual product configurators are designed to help make the
configured price code process simple for new sales people and ensure that they do not
make any mistake while ordering. They are constrained by preset rules so only viable
configurations can be made based on elements like function, technical aspects and budget
limitations.

The next is buyer intelligence. Now buyer intelligence provides insights into customers
including why they buy, how they make decisions when purchasing and the type of
product and content messaging that work. So, you can better understand their priorities
and challenges now and in the future. As a result, companies can create marketing
messages and make business decisions that are better informed. Buyer intelligence is
assisted by the collection of zero party data. So, what is zero party data? Zero party data
is information customers proactively and intentionally share through discovery
experiences such as through the guided selling assistant.

This type of data is considered more trustworthy and definitive compared to other forms
of data since we are getting it directly from the customer. We can use zero party data to
help communicate with customers and improve personalization such as guiding them to
products or information based on their answers. Now let us look at AI powered configure
price code for e-commerce. Previously configured price code that is CPQ data has often
been stored in different programs like spreadsheets, email and even internal knowledge
within different teams making it harder to fully analyze and use that information. Now
advances in AI technology allows CPQ systems and solutions that can help streamline
sales operations, simplify and optimize the process of configuring products and
identifying pricing and more.

These AI tools help to automate task, increase efficiency to provide a centralized

database, visualize trends and identify recommendations. So, these are the five things
that these AI tools help in doing. Some of the benefits of AI powered CPQ for e-
commerce are discussed in the upcoming slides. So the first is precision and accuracy.
AI powered CPQ tools help to reduce human error in configuration and pricing.

So that is one thing. Additionally, they can reduce the time needed to understand
complex catalogs and product suits to help guide customers to the right product. AI
algorithms can also create accurate codes efficiently even when using customized
parameters making it easier to tailor codes and information to specific customer needs
when necessary. Dynamic pricing. Many e-commerce stores no longer rely solely on
supply and demand rule when establishing pricing. Instead, more companies embrace
dynamic or variable pricing and selling strategies like offering a discount for a specific
time.

AI powered CPQ tools help to gather and analyze data to help companies create and
manage their dynamic pricing. Increasing sales efficiency and streamlining omnichannel
selling. Using AI will help businesses optimize their dynamic pricing algorithms. One
identify patterns and enhance efficiency.

Enhance customer experience. AI powered CPQ tools can help e-commerce companies
elevate their customer experience through personalized configurations and
recommendations based on customer data and predictive analysis. For instance, shoppers
can receive faster responses to inquiries, especially regarding customized codes and
pricing. Codes also will be more accurate and can be better personalized to that customer
preferences and needs. So it is all about the personalization.

Streamlining workflows. Using AI powered CPQ tools help to reduce the time from lead
to conversion by using automated processes. You can also generate codes faster, reduce
time to approval, automate product configuration and enhance accuracy and consistency.
Streamlining workflows also help improve your sales process, customer experience and
at the same time, save the time. Scalability. AI powered CPQ software helps the e-
commerce businesses to scale without compromising on the quality, cost and time.

These tools help by handling increased amount of users, data and tasks. For instance, it
can help business manage a growing number of products, configurations and pricing
models without added complexities. So, products there and the pricing models. So these
combinations, these number of combinations can be handled without any kind of
complexity. What are the best practices for utilizing AI in B2B customer experience?
The first is to keep your data clean and accurate.

AI is only as good as the data it is trained on. So, it is essential to keep the customer data
clean and accurate. This can involve regularly updating your data and removing any
irrelevant or duplicate information. The second is to provide transparency. It is crucial to
provide transparency around how your AI is making recommendations and decisions.
This can involve explaining how your algorithm works and provide customers with the
option to opt out of personalized recommendations.

The third is to continuously measure and improve performance. To ensure that your AI is
providing the best possible personalized experiences, it is essential to continuously
measure and improve its performance. This can involve analyzing metrics such as
customer satisfaction, sales efficiency and then adjusting your algorithms as necessary.
So whether all our requirements so far as customer satisfaction, sales and efficiency are
being achieved and then if not then adjusting your algorithm as necessary. Challenges
when implementing AI for B2B The first challenge is lack of data.
 One of the most important challenges when implementing AI in B2B context is a lack of
data. If you do not have enough customer data, your algorithms will not be able to
provide accurate recommendations or solutions. To overcome this challenge, consider
collecting additional data through surveys or website interactions so that some minimum
amount of data is available to train the AI engine. The second is lack of understanding.
Another common challenge is the lack of understanding of AI and its capabilities.

To overcome this challenge, we should consider investing in training or hiring an AI

expert to help implement and optimize the AI. The third is and the most significant is the
resistance to change. Finally, resistance to change can be significant challenge when
implementing AI. To overcome this challenge, it is essential to communicate the benefits
of AI to your employees and customers and provide transparency about how it is being
used. Fourth is matrix to measure the success of AI in B2B experience.

So for all these the customer satisfaction, sales and efficiency, the matrix are to be
worked out. So to measure the success of your AI for personalized B2B customer
experience, a few metrics to keep in mind are first customer satisfaction. It is a crucial
metric to measure when implementing AI. By analyzing customer feedback and
sentiment, you can determine whether your AI is providing the personalized experience
your customers are looking for.

The second is sales and revenues. Sales and revenues are also essential metrics to
measure when implementing AI. By analyzing sales data, you can determine whether
your AI is providing effective recommendations and offers that are driving sales. Now
let us look at the future of AI in B2B selling. As AI technology advances, it is reshaping
both operational landscape and customer expectations within the B2B E-covers context.
One significant area of impact is the rising demand for high quality automated self-
service options.

So this is, the demand for this is increasing. Modern online B2B buyers crave immediate
access to information and the ability to complete tasks autonomously without any kind of
human interaction. The effectiveness of AI-powered self-service tools hinges on the
quality of underlying data, better data, more data, more effective AI. Clean, structured,
enriched data are crucial for optimizing these automated customer interactions. As AI
technologies become more advanced, we can expect to see even more personalized
experiences for B2B customers. This could involve technologies such as emotion
detection or even personalized virtual assistants.

Moreover, as natural language processing technologies continues to improve, we can

expect to use more natural interactions between customers and AI-powered chatbots and
virtual assistants. This could involve technologies such as voice recognition or even
customer avatars even in the B2B space as well. So to conclude in this module, we have
discussed personalization in a B2B perspective where we have first studied about the
emergence of personalization in business to business with the help of AI. Then we have
studied the steps to implement AI for personalized business to business customer
experience and then we have studied the rise of AI and how to integrate AI in the B2B e-
commerce. Thereafter, we have studied how various AI tools are helping in personalizing
B2B e-commerce buying experience.

After that, we have studied how AI aided CPQ is benefiting B2B e-commerce. We have
also gone through the best practices to be followed and challenges that are associated
while implementing AI in B2B selling. Finally, we have also discussed briefly about the
future of AI in B2B sector. And these are the 8 sources from which the material for this
module was taken. Thank you.
 AI in Marketing
Prof. Zillur Rahman
Department of Management Studies
Indian Institute of Technology, Roorkee
Week 8
Lec 38-Implementation of AI by Product Managers

Welcome to this NPTEL online certification course on artificial intelligence and

marketing. And now we will talk about module 38. So as you can see from this slide, we
are talking about the implementation of AI by product managers. So this is module 38
and we are in chapter 6. Now these are the things that we will cover in this module. So
we will start with the guidelines to be followed for AI infusion based on the key phases
of product development cycle to be followed by the questions to be answered in each key
phase of the product development life cycle along with use case examples.

So let us start with the first one that is implementation of AI by product managers. As
AI gets even more deeply embedded in society and people's everyday life, business
leaders, data scientists and engineers have a responsibility to guide AI product
development towards responsible outcomes. So this is very important word responsible.
Product leaders typically oversee the development of one or more AI products.

Their teams have simultaneous responsibility to advance an organization's values and

business objectives while guiding product development towards the responsible
outcomes. So now they have to do look after the organization values, business objectives
and responsible outcomes. Their place at the center of innovation and product
development positions these leaders as essential actors in operationalizing responsible
AI. The range of AI products that oversee is broad with the underlying algorithm ranging
from novel complex models to pre-trained models built by others. Organization leaders
focused on product development and delivery should remember to take sufficient
measures to scale responsible AI across their organizations.

Ultimately responsible AI is about driving a cultural shift within an organization. Truly

operationalizing responsible AI requires broad changes across leadership, governance,
processes and talent. We will discuss some guidelines focusing on product development
and delivery processes to provide a formalized way to stimulate the challenging
conversations and critical thinking necessary to anticipate and mitigate the risk of AI
systems. These guidelines will help product leaders create deeper synergy between
performance, organization goals and values. Throughout the product development life
cycle, leaders should keep in mind their organization's values and AI's ethical principles
to the extent their organization has formalized them.

Now these are some of the guidelines. Assess and prepare, design, build and document
and validate and support. The guidelines have been organized according to the key
phases of the product development life cycle while recognizing that AI product
development often cycles through these phases iteratively. So, the first is to assess and
prepare. Assess merit of developing the product considering organization values and
business objectives.

Assemble teams reflecting diverse perspectives and with clearly defined roles and
responsibilities. Assess potential product impact by including inputs from domain
experts and potentially impacted groups. So, the first one is to assess merit of
developing the product considering organizational values and business objectives. So,
now let us understand what needs to be done in this first step. So, what are the primary
use cases and benefits of the proposed products which users are explicitly out of scope?
What is the desired business outcome for this product? How will the business impact be
measured? How might the operations of the product and use of its output for business
decisions, enroach on core organizational values? The second is assess merit of
developing the product considering organizational values and business objectives.

So, use case example is automating loan approval decisions. A financial institution
would like to automate its loan application approval process with the goal of increasing
efficiency and reaching new customer basis. The product leader should immediately
recognize that because of the AI's product's potential to impact individual's economic
well-being and quality of life, they must proceed with care. The system should be
designed to predict an applicant's earning potential over the lifetime of the loan under
review and thus the likelihood of repayment. The business impact should be measured
with the key performance indicators.

The first is default rate against a historical baseline. The second is increase in overall
loan volume and the third is speed of approval or rejection decisions. There may be
race, gender or other biases reflected in the historic data. Exploratory data analysis will
need to include steps to bias assess the data set composition so that appropriate
mitigations can be implemented. There is a significant risk that proxy variables are race,
gender and other protected categories could reinforce biases and impact outcomes.

By letting company values around fairness and the law, the product team will have to
proactively engage on these topics. The next is to assess and prepare. Assemble team
reflecting diverse perspective and with clearly defined roles and responsibilities. Do we
have a diverse that is gender, age, ethnicity, multi-disciplinary team with a range of
functional expertise? What prospective or expertise are missing and how can we
introduce them, including sources outside the team or organization? Is the team is
structured so that domain experts can impact relevant design choices? The next is
assemble team reflecting diverse perspective and with clearly defined roles and
responsibilities. The product leader should begin to build the team based on initial
assessment that the team should include data scientist with experience applying fairness
tools to machine learning models, a lawyer with expertise in the Equal Credit
Opportunity Act to make sure the team fully understands the regulatory environment.

Two loan officers with extensive experience helping a variety of customers navigate the
application process. This will enable the team to design the product to augment the
decision making capabilities of loan officers and integrate their feedback in real time. A
product leader should know that since the system will deal with sensitive personally
identifiable information, system security and privacy are critical. We must gain approval
to bring on an external consultant with cyber security and AI expertise to provide
guidance on optimal system architecture, data storage and differential privacy across the
build. A user researcher to ensure customer needs are at the forefront from idea inception
to final product so that customer's understanding and satisfaction can be reconciled with
business goals.

A designer to improve product usability and accessibility across a diverse user base. A
diverse team reflecting a variety of backgrounds and lived experience. The third is to
assess and prepare. Assess potential product impact by including input from domain
experts and potentially impacted groups. What are the foreseeable models of failure for
this product? What edge scenarios could lead to failure and harm? What are the societal
and environmental implications of foreseeable product failure, misuse or malicious
attack? What are the product's potential unplanned uses? What external SMEs or groups
could provide input in forming design choices that would reduce the risk of negative
societal impact and harm to individuals directly or indirectly affected by the AI product?
Assess potential product impact by including input from domain expert and potentially
impacted groups.

A preliminary discussion should take place where the loan officers should share their
experiences engaging with the financial institutions current customer base. As they detail
interactions with various customers over the years, the product leader should realize that
the AI may engage new customer demographics who may have needs and expectations
that differ from those captured in historical data. The team should engage an economist
with expertise in banking relationships within the demographic groups to whom the bank
might expand its activities. The team should also work with a vendor to deploy a survey
for potential borrowers in the new customer communities to better understand how great
access to credit might impact earning potential in ways both consistent and inconsistent
with communities overrepresented in the historical data and thus the likelihood of
repayment. The loan officer should also note that even with the new customer base, one
thing is likely to remain constant.

That is having a loan application declined in an unpleasant and potentially painful

experience. Having a loan application declined is an unpleasant and potentially painful
experience, one that can be delivered more respectfully by a skilled and experienced
professional. Based on this insight, the product team should decide that all rejection
decisions shall be communicated to applicants by a loan officer. Design, build and
document. So, the first step here is to evaluate data and system outcomes to minimize the
risk of fairness harms.

The second step is to design AI products to mitigate the potential negative impact on
society and the environment. The third step is to incorporate features to enable human
control. The fourth is to take measures to safeguard data and AI products. And the fifth
is to document throughout the development life cycles to enable transparency. So, let us
start with the first step that is evaluate data and system outcomes to minimize the risk of
fairness harms.

What fairness metrics? For example, statistical parity, equalized odds, test and chipping
criteria will be used. How will the product team validate that the training data includes
data collected via APIs capture the different groups and types of people likely to be
impacted by the system's output? How will the product team measures whether the AI
product outcomes are consistent with the chosen objective, that is avoid target leakage.
Fairness metrics, tests and chipping criteria across a wide variety of potentially impacted
groups or intersection of groups. How will the product team ensure continued adherence
to fairness metrics, tests and criteria post deployment? Now let us look at a user case
example of demand forecasting for a fashion retailer. A fashion retail chain hopes to
transform its in-store inventory management with AI.

Using historical sales data, the company wants to optimize the amount of inventory held
in stores so that it can maximize sales per square foot. For this, the product leaders
should clarify the business objectives and assemble a team. A team should now shift its
focus to the design, build and document stage of work. The product leader should
convene the product team for a discussion around fairness. A poorly designed product
could lead to service discrepancies across different demographics within the customer
base, which would violate the company's values.

Data scientists on the team should note that customer's feedback for the retail chain
may vary dramatically across neighborhoods in large metropolitan areas, with stockouts
more common in some locations than others. Pivoting from this, the team should align
on the following approach. A team should check for parity of service level based on
stockout reported in its historical data. Cross-referencing the results with census data
will allow team to assess whether stockouts reported are correlated with certain
neighborhoods and perhaps demographic groups. To prevent unacceptably low service
levels in specific situations, the product team should establish a minimum inventory
level for each SKU at each location to ensure minimum service levels that will avoid the
possibility that stockouts would impact certain demographic groups more than others.

The second is to design AI products to mitigate the potential negative impacts on society
and the environment. Its negative impact that is the system's failures, unplanned use,
abuse, attack or simply side effects of normal use is possible. What design processes
that is human centered design and choices can reduce, mitigate or control them? What
design choices will help minimize the adverse environmental impact on the product
output and related decisions? What design choices are critical to ensuring proper use,
legitimate and transparent data collection and respect for user privacy? The team's
sustainability experts should ask the group if system usage might have some second-
order environmental effects. Optimizing inventory levels should maximize profit at the
store level, but smaller and more frequent inventory replenishments would rely on
greater air and ground cargo traffic. At the country level, the environmental consequence
of that cargo traffic could be significant.

Furthermore, since significant amount of excess inventory would need to be returned by

stores, optimized inventory levels could create additional cargo traffic. Using internal
logistic data, the product teams can build a feature that highlights trade-off between
inventory levels and transportation emission at the store and region level. The third is to
incorporate features to enable human control. How is the product team designing the
product to empower humans by augmenting their decision making, streamlining tasks
and otherwise making them more effective? Which decisions or functions requires
human oversight as a critical component of the AI product? The product mechanism
will support end-user comprehension of the system to enable continuous audit,
monitoring and human intervention. What product features allow users to customize AI
performance? What channels will the product utilize to collect live feedback? What
product features will ensure inclusive experiences for people with disabilities? Next we
will look at incorporating features to enable human control.
 A specialist in human-centered design should understand that the retailer often learns
about new trends from its front-line workers, particularly store managers. The historical
data has limited predictive power for trend spotting. By enabling human control over the
inventory system and augmenting store managers' decision-making abilities, stores
should make on-the-fly adjustments to match changing consumer preferences. Based on
further exploration, the team should design a feedback mechanism through which store
managers can indicate emerging trends and newly popular products at a specific location,
thus allowing for the pooling of insights across the country to help spot trends and adjust
inventory levels accordingly. The fourth is to take measures to safeguard data and AI
product.

How will the product team prioritize data privacy across product designs with respect to
cloud infrastructure, encryption, anonymization, and access control? How will the
product team make sure the product does not inadvertently disclose sensitive or private
information during use, for example, indirectly inferring user location or behavior? What
methods will the team use to identify and address security vulnerabilities, including
those such as data poisoning, that are unique to AI products? The fifth is document
throughout the development lifecycle to enable transparency. What standards and
processes are in place to ensure the entire product team consistently document design
and development choices, rationals, and assumptions? How can the product team best
keep track of data sources and their authorized users? What types of models, tools, or
techniques will be used to document product behavior? Next comes validate and support.
Validate product performance and test for unplanned failures as well as foreseeable
misuse unique to AI products. Communicate design choices, performance, limitation,
and safety risk to end users. So, let us look at the first one in more detail.

Validate product performance and test for unplanned failures as well as foreseeable
misuse unique of unit to AI product. How will the product team validate the AI product's
performance against agreed upon business KPIs and metrics, tests, and criteria? How
will the product team validate the AI product's performance against technical standards
and benchmarks? What are the target environment and conditions under which this
product can be expected to function properly and safely? How will the system be tested
and evaluated for safe and effective operations, that is graceful failure in both business as
usual and edge case scenarios? What are the mechanisms for continuously monitoring
business, technical, and fairness performance as well as for modeling drift post-
production? Validate product performance and test for unplanned failures as well as
foreseeable misuse unique to AI product. Use case example of predictive lead times for
a manufacturer. An industrial goods manufacturer is experiencing repeated delays in the
delivery of parts purchased from suppliers. This has caused disruptions to
manufacturing schedule and ultimately, led deliveries to customers and the potential to
damage key relationships.

The manufacturer has decided to develop an AI product to estimate lead time for
supplier procured component based on historical data. Providing purchasing managers
with an early warning of potential delays will enable them to proactively engage with
suppliers and adjust the manufacturing schedule accordingly to avoid missed delivery
deadlines. This AI product might change the way the company engages with suppliers,
customers, and workers. Although the product is expected to have positive impacts
across operations and customer relations, erroneous lead time estimates could also create
more work for purchase, ring managers and further damage, supplier relationships and
negatively impacting the bottom line. This could have significant and far-reaching
consequences for the business.

Furthermore, because assembly is a labor-intensive process requiring specialized skills

and safety criteria at different stages, ad hoc adjustments to the manufacturing schedule
could place factory employees at risk. The product leaders along with the product team
should reach alignment on the approach to validating product performance and
robustness to unplanned failures. The following should be looked into. Testing the
system's output against historical supplier promise dates to determine the model's ability
to flag potentially delayed shipment before it's too late. There are a wide range of
operational scenarios including varied suppliers and component types as well as edge
scenarios, that is parts and suppliers not found in the historical data.

For each scenario tested that requires adjustments to the manufacturing schedule,
developing associated work schedules to be validated by factory floor leadership for
feasibility and safety. For example, during the COVID scenario, the data scientist team
should note that factors related to the ongoing COVID-19 pandemic such as limited
trucking capacity and economic shutdowns in certain states would not be captured in
historical data but could impact lead times. To capture these insights, the product team
should integrate COVID case loads in geographies proximate to suppliers such as the
way to capture the pandemic's potential impact on the manufacturer. In conversation
with factory floor leaderships, the product team learned how scheduling has recently
changed to minimize risk of COVID-19 exposure and that any changes to the
manufacturing schedule would have to be consistent with the new scheduling policies.
The second is to communicate design choices, performance, limitation and safety risk to
end users.

What information and instruction should the product team provide to end users to ensure
safe and reliable use? How will the product team make sure end users understand the
primary use case, underlying assumptions and limitations of the product? How will the
output of the systems be communicated in a way that helps end users understand how the
system works? Now for communicating design choices, performance, limitation and
safety risk to end users, we will use the use case example of predictive lead times for a
manufacturer. The team's UX lead should push the product team to consider how best to
augment the purchasing manager's current decision making process. The team should
align on the following steps. The first is to calculate confidence interval alongside
estimated lead times to enable purchasing managers to responsibly leverage system
outputs. The second is to design the dashboard to sort components for a certain product
by a predicted delay, focusing the end users' attention on prioritizing key products and
engaging problem suppliers early.

Make sure the system never automatically updates customers on delivery dates, a task
that program managers should themselves continue to perform. Build on additional
supplier level as opposed to component level dashboard to equip the company for
strategic engagement with suppliers that consistently struggle to deliver parts on time.
Output a draft work schedule based on proposed changes that will support purchasing
managers' decision making regarding adjustments to the manufacturing schedule. The
factory floor leadership will have to sign off on the revised work schedule to ensure
adherence to workers' safety standards. The team, if earlier identified and engaged with
the products end users, should conduct several meetings and trainings to ensure that
these features will be leveraged effectively.

It should also design a modular training on tool use to be integrated into future
onboarding programs for purchasing managers. So, to conclude, in this module we have
continued explaining the other two key phases of guidelines for product development
cycle. We have also discussed the use case for each phase. And these are the five
references from which the material for this module was taken. Thank you.
 AI in Marketing
Prof. Zillur Rahman
Department of Management Studies
Indian Institute of Technology, Roorkee
Week 8
Lec 39-AI in Service

Welcome to this NPTEL online certification course on Artificial Intelligence and

Marketing. And now we will talk about module 39. In this module, we will talk about
how AI, the role of AI in services. So, these are the things that we will cover in this
module. So, the current trends of AI used in the service sector. One, the AI tools most
effectively used in service sector along with some current industry related cases.

That is the second thing that we will discuss in this module. The various benefits of
using AI in customer service and the limitations of AI in customer service and some key
tasks to understand the potential risks in customer experience. For every brand, helping
customers make purchase decisions by delivering on their needs is a top priority. And
that is not as easy as it sounds.

A surge in digital channels and hybrid online-offline journeys along with heightened
customer expectations have complicated brands' desire to create and deliver a dynamic
experience that equips customers with the right information and incentives when they
need it the most. Deloitte surveyed 11,500 global customers to better understand what
information they found most helpful while making purchase decisions across a variety of
product as a service categories. Two tactics stood out – timely offers and knowledgeable
customer service. In fact, across 8 sectors, respondents cited a timely offer most often
with knowledgeable customer service as the second most helpful option in 6 categories.
These two dimensions routinely outperform customized recommendations.

Free trials and samples, no-hazel return and cancellation policies and augmented
technology such as virtual showroom. So this chart here, it shows timely offer. These
are the various features of the service – timely offer, knowledgeable customer service,
no-hazel return policy, technology-weighted, easier to learn more, customized
recommendations, free trials. And these are the products where the and services where
the survey was carried out. So you see that in automobile, timely offer is 37% important
while in household equipment it is 41%.
 Knowledgeable customer service is 39% important. 39% of the people said that it is
important in auto and the rest of the four options had limited importance with the
customers. So you see there are 20% and 30% and 23% etc. So you see that these two
things – timely offer and knowledgeable customer service, these are the two most
important features, options given to the customers across the product and service
categories. So building on the feedback, brand marketers and customer experience
leaders should be asking how can we get the best offer in front of the customers.

How can we get the best offer in front of the customers when they need it most and
equip service agents with right information to offer those customers a level of service
that helps them make the best purchase decisions. The answer is to optimize artificial
intelligence within the customer experience achieving harmony between human tasks
and machine capabilities. A dynamic experience for customer means delivering the
assistance and information they need whenever, wherever and however they want it.
One prong of this is the utilization of AI that has enabled brand to deliver more
personalized and creative marketing material and help predict customer behavior.
Designing and deploying an AI strategy that assists customers in their moment of needs
may be the most effective way to ensure the right offers meets customers at any point in
their customer journey.

Smarter chatbots One of the first line of AI defense in terms of reaching your customers
is the increasingly popular chatbots. Unfortunately, in the early stages, chatbots have
proven to be a bit frustrating for many customers equipped with too few phases and too
little helpful information. While they may have saved the company's time, in some cases
chatbots just left customers feeling more helpless than before. Companies are now
starting to invest in smarter conversational AI that pulls history and experience into the
conversation while also using emotional recognition tools to better understand the
context of the customer inquiry. Because they are able to engage in deeper, more
complex conversations, they are able to gather much more data about the customer and
answer their problems much more quickly.

A recent report from Capgemini indicated that 54% of customers said they have daily
AI-enabled interactions with businesses, including chatbots, digital assistants, facial
recognition and biometric scanners and 49% of those customers found AI interactions to
be trustworthy up from only 30 in 2018. Chatbots of today won't be confused by a
customer changing the topic of conversation. It can jump from topic to topic and even
channel to channel such as starting a conversation on the brand's website and then
transitioning to WhatsApp if the customer needs to leave their desktop to meet the
customer where they are and provide service that rivals that of their human counterparts.
With the help of conversational AI, consumers can connect with brands right in the same
channel they use most of the time. Ancient virtual consigrees and bots instantly greet
them, answer their questions and carry out transactions and if needed connect them to
agents with all the contextual data it has collected over the course of the conversation.

Service Robots As the global service industry continues to manage rapidly changing
customer trends and a deluge of external factors such as labour costs, trade concerns and
supply chain interruptions, some are seeking new and cutting-edge technologies such as
service robots as a solution. The use of robotic technology for improving productivity
and enhancing efficiency in business is not a new idea. Industrial robots have been used
in manufacturing and warehouses for nearly 50 years and are still used today. However
the notion of using robot is specifically for assisting human workers in completing
service related tasks or helping customers get the best experience possible is relatively
new with the first service robot being implemented in only the last decade. Despite their
novelty, service robots are quickly becoming an essential part of business for service
focused companies in healthcare, hospitality, logistics and retail that are looking for
innovative ways to delight customers while boasting their bottom line.

Now how the service robots are being used in retail? In retail organization service
robots help enhance and personalize the customer experience as well as improve in-store
operations. For example, several global retailers in Americas, Europe are using the
autonomous service robot to automate in-store retail operations. Intel's Computer Vision
Technology, TELI, orders around 10,000 items in 30 minutes to provide near real-time
inventory information and insights to store associates, ultimately optimizing store
execution and improving customer satisfaction. Many retail companies seek out service
robots to simplify and enhance their customers' experiences. SM Supermalls, a chain of
shopping malls in the Philippines, turn to New Era AI Robotech or New Era to
streamline and improve their visitors' experience by developing a fleet of voice
interactive smart service robots.

Guests can ask the AI-enabled robot engaging questions, getting simple directions to the
desired store location or upcoming showtimes for the on-site movie theater. SM
Supermalls also use their service robots to deliver tailored promotions and marketing to
each customer while gaining deeper visitors' intelligence through data collection. Now
how are service robots being used in hotels? In hotels across the globe, service robots are
being used to enhance and personalize the customer experience. Interactive AI-powered
QoSX greet customers when they arrive while service robots take guest luggage straight
to their rooms to make the check-in process seamless and efficient. AMRs bring them
room service and humanoid robots act as their personal conceit.

The food service industry is also taking advantage of service robots. In Singapore, the
Crown Coffee Bar recently onboarded its newest barista, Ella, who happens to be a fully
autonomous 6-axis robot. So this is that. From taking orders via the online app to
making and serving coffee to finalizing the bills and charging credit cards, the robot does
it all. Ghost kitchens, also known as delivery-only kitchens, with no dine-in service are
leveraging autonomous service robots to deliver food orders to customers.

Until recently, ghost kitchens were typically used by pizza restaurants or other delivery-
only businesses. However, the challenges that have emerged from the COVID-19
pandemic, including the need for contactless delivery, increased cost and labor shortages
have helped ghost kitchens proliferate and become more of a standard restaurant
business model. Many modern ghost kitchens located in densely populated or campus-
like areas are now experimenting with self-driving service robots as an innovative and
cost-effective delivery option. How service robots are being used in logistics? As e-
commerce sales continue to surge, logistics businesses are using service robots to help
overcome current labor shortages, assist current workers to avoid workforce burnout and
enable warehouse automation. Robotic arms are often assigned tasks like picking,
placing and sorting objects and because AMRs can navigate warehouses independently,
they are used to deliver material to human workers for accurate and efficient order
fulfillment.

To tackle some of the same-day delivery businesses, FedEx created an autonomous

mobile robot named FedEx Same Day Bot to work with local retailers to deliver
customer orders directly to their houses. The bot navigates sidewalks, unpaved services
and steps while carrying cargo which can include hot or cold food orders. So this is that
machine that is used by FedEx for delivery. Both type of service robot can be
empowered with vision sensors and artificial intelligence technologies to allow them to
see which enables them to more accurately detect, classify, sort, pick and package
objects. Some logistics companies such as FedEx are experimenting with using AMR for
last-mile delivery of goods which is often the most expensive and least productive part
of the entire delivery chain.

Next comes Augmented Reality. Augmented reality is the overlay of digital content on
the real-world environment. Virtual objects can be in various forms. Images, videos or
interactive data. In other words, it supplements the real world with digital objects.

It can be used in two main ways. Portable devices and smart glasses and AR headsets.
What are these portable devices? Augmented reality is the most accessible reality
technology as people can use their smartphones or tablets to run augmented reality
applications. AR apps use a phone camera to capture the real world. Virtual objects are
then overlaid and users can see them on their smartphone screens.
 Smart glasses and AR headsets. Another way to create AR experience is to use special
smart glasses or headsets. Unlike VR headsets, these AR glasses and headsets don't
immerse users into fully virtual environment but just add a digital object to the real
world. With Glass for example, digital data is projected right in front of the user's eye.
One of the key features that set Lenscard apart from other eye wear companies is its
virtual try-on feature.

This feature uses advanced augmented reality technology to allow customers to try on
eyeglasses and sunglasses virtually before making a purchase. So you see that find your
perfect fit with AR, explore and try 10,000 frames and find your AR frame size using
our AR feature. So this is what this picture is from, from the Lenskart and it shows the
person with various eyeglasses. This technology has revolutionized the way people shop
for eyewear online and had made the process much more convenient and enjoyable. To
use the virtual try-on feature, customers simply need to visit the Lenscard website or app
and select the frame that they are interested in.

They can use their phone's camera to see how the frame looks on their face in real-time.
This allows them to get a realistic and immersive experience and see how the frame will
look from different angles. The virtual try-on feature also allows customers to compare
different frames side by side and see which one looks best on them. This eliminates the
need to go to a physical store and try on multiple frames which can be time-consuming
and frustrating. Another benefit of the virtual try-on feature is that it helps customers
make more informed decisions when purchasing eyewear online.

They can see how the frame looks on their face and get a sense of its size and fit before
making a purchase. This can help reduce the likelihood of returns and exchanges which
can be a hassle for both the customers and the company. Virtual Reality Virtual Reality
is an artificial digital environment that completely replaces the real world. With VR,
user experience, artificial sounds and sights and feel as if they are in a digital world. VR
has a lot of applications like in healthcare, education, marketing, real estate and many
more.

Businesses can reach out to their audiences by launching immersive and engaging
marketing campaigns. This is particularly important in the age of online shopping as VR
experiences help people get a feel for products without leaving their homes. IKEA had
introduced a VR application for its customers that works as a virtual reality showroom.
The VR experience works like a creative primer by inviting them to visualize different
combinations. The application provides an immersive and emotional product experience
that revolutionizes the way customers engage in interior decoration.
 So this is how it is working. It empowers the visitor to visualize their own taste, inspire
and expand their mind with new ideas. After slipping into IKEA's virtual showroom,
visitors are able to explore and configure furniture items and their environmental context
in real time. In the blink of an eye, they can try different fabrics, swap the wall color and
even change the time of the day to see their combinations in a different light. Next
comes Mixed Reality. In mixed reality, sometimes called hybrid reality, virtual content is
not only overlaid on the real environment as in AR but is also anchored to and interact
with that environment.

Put simply, a mixed reality can use virtual objects just like you can in augmented reality
but these objects can also interact with the real world. In a sense, mixed reality is more
immersive and interactive type of augmented reality. It is best suited for entertainment
and sports events. From concerts with musicians interacting with digital pops to
watching the action of football game appear to go down in your living room, developers
are working on many mixed reality apps that could create these and similar experiences.
Several years ago, PGA Tour fans were able to use an app with a 3D rendering of golf
courses.

The Tour is also looking to use HoloLens for tournament setup and golf course design
where players can prepare their strategy using mixed reality well before hitting their first
drive on the links. The next comes holographic technology. 83% of brands are either
adopting and investing now or plan to in the future on this technology which shows great
potential in areas such as in-store advertising, interactive gaming and public events
through an expanded three-dimensional viewing experience for an entire audience, not
just a single individual. While AR-VR technology allow a single user to experience a
virtual visualization through a smartphone or headset, holographic technology brings that
ability into the real world, allowing multiple people to engage or experience
simultaneously and allowing brands to bring singular experience to a broader
collaborative audience. Holographic telepresence can enhance remote collaborations,
sales presentation and product demonstrations, enabling organizations to save on-table
expenses and reduce their carbon footprint.

Embedded AI and IoT immersive technologies can only be successful if the ability to
capture, analyze and react to data in real time is present. The application of AI and IoT
sensors is critical to the enabling of voice-based AI assistant, AR-VR and holographic
technologies that are the tools that connect immersive technologies from the consumers
to the brand. AI and IoT have been traditionally thought of as operational resource but
today consumer facing technologies are built from the ground up with sensors and
processing powers. Retailers are leveraging that technology to improve the immersive
experience, capture and analyze consumer behavior and build a more rewarding
customer experience. Unilever has 3 million ice cream freezers around the world.

For retailers, maximizing seasonal sales means working to ensure that they never run
out of stock. Making sure retailers don't run out of stock has seen Unilever incorporate
digital tools and AI into the freezer fleet. Today 50,000 freezers are AI enabled when it
comes to putting the right product in the right freezer. The camera fitted inside the
cabinet takes a photo periodically sent to the cloud and it is analyzed using AI to let
shopkeepers know what to restock and submit orders in a frictionless way. Some stores
have seen an uplift in sales of 15 to 35%.

There is a potential for stock data to be used for market research to shape target
promotions and further help retailers understand what their customers want. Now let us
look at the benefits of AI in customer service. The first is improved customer support
workflows. AI technology can be used to reduce friction at nearly any point of the
customer journey. Chatbot allows us to quickly answer routine questions from new
customers.

We can also use conversational AI to develop personalized emails and knowledge base
articles for existing customers. AI helps to streamline internal workflows and in return
maximize customer service interactions. The second is to better prediction of customer
behavior. AI uses real-time data to make predictions. That means it can be used to
determine how customers are likely to behave based on their purchase history, buying
habits and personal preferences.

Predictive AI can help identify patterns and proactively make improvements to the
customer experience. The next benefit of AI in customer service is sentiment analysis.
Today many bots have sentiment analysis tools like natural language processing that
helps them interpret customer responses. This tells the bot how to respond when a
customer behaves in a certain way so it won't escalate the customer's emotion when it is
trying to help them.

The third is request routing and prioritization. We can also use AI to identify keywords
and analyze the nature of the request before assigning it to one of the reps. AI tools
allow to analyze the customer's problem, the perceived urgency and their associated
emotions. We can also program AI models to tag in the right agent to handle the case.
This is another time-saving process that makes it easier for the support team to
troubleshoot and resolve issues. The next benefit of AI in customer service is the voice
analysis.
 Voice AI has made it easier to automate call center interactions. These tools can be
trained in predictive call routing and interactive voice responses to serve as the first line
of defense for customer inquiries. Yet another benefit of AI in customer service is omni-
channel service. AI can support omni-channel service strategy by directing customers to
the right support channels. So that is the omni-channel, directing customers to the right
support channel.

If all the chat reps are busy taking cases, it can tell the customers to use LiveChat for
quicker responses. Or if a customer is typing a very long question on the email form, it
can suggest that they can call in for a more personalized support. The next benefit of AI
in customer service is data management. Not all AI features are customer facing. In fact,
some of the most useful tools are the ones that are integrated with internal software.

As an example, AI can be paired with the CRM to recall customer data for service
agents. Customer service teams can use this feature to proactively serve customer based
on AI generated information. Now what are the limitations of AI in customer service?
The first is that it can be unpredictable. While AI itself is not exactly new, the sheer level
of availability of this technology to human is. While testing and tinkering is going on to
see what this tech can do.

As such, AI tools can be predictable and in some situations potentially dangerous. The
next limitation is that it can alienate segments of the market. By adopting a full AI
approach to customer service processes, it may risk alienating different parts of the
customer base. That could be those with growing security concerns about how AI uses
their personal data. It may also exclude elderly individuals who don't feel comfortable
using this technology.

Another limitation is limited understanding of the context. AI can automate and

segment tickets based on the nature of the queries. It can even go as far as identifying
customer sentiment based on the tone of voice. It can even use the data to follow up with
basic account actions. But AI may struggle with more challenging issues like high
cancellation intent, high frustration and other nuances.

Simply put, AI needs help understanding the wider context. Thus, its customer
recommendations can lack nuances. Lack of human touch. Integrating AI into customer
service processes can bring incredible advantages. But there is one thing everyone who
shared their insights about AI in customer service mentioned, the human touch is
irreplaceable. Customers can feel when it is missing, leading to frustration and
increasing the chance of churn.
 Fewer interaction with customers using AI in customer service can reduce the need for
human agents to intervene with issues. Though it might save time and allow reps to
focus on more complex tasks, but the team will have fewer interactions with customers,
meaning fewer chances to organically gather critical feedback and make customer feel
valued. Then another limitation is it needs continuous maintenance updates. AI systems
rely on data algorithms and if algorithms are not adequately trained or updated, there is a
risk of providing incorrect or misleading information. Because of that, regular
monitoring and maintenance of AI systems are crucial to minimize the occurrence of
errors.

Now there are three key tasks to better appreciate the customer experience and its
potential deficit. The first is to understand the experience strategy. Identify each
channel and touch point along the customer journey. Make sure they are connected. The
channels and touch points should create a congruent experience for customers.

And the third is to design with human-centered factors at the forefront. Consider the
entire service experience including the customer, talent and third-party service providers.
So, to conclude, in this module, we have first discussed about the current trends in AI use
in the service sector. Then we have studied about the various AI tools that are being most
extensively used in service sector along with some current industry related cases. Then
we have also gone through the various benefits of using AI in customer service.

Finally, we have discussed about the limitations of AI in customer service and some key
tasks to understand the potential risk in customer experience. And these are the six
sources from which the material for this module was taken. Thank you.
 AI in Marketing
Prof. Zillur Rahman
Department of Management Studies
Indian Institute of Technology, Roorkee

Week 8
Lecture-40 Pricing Strategies Using AI -I

Welcome to this NPTEL online certification course on Artificial Intelligence and

Marketing and now we will talk about module 40. So, as you can see here these four
modules are dedicated to understanding pricing strategies using AI. So, this is module 1
and part 1 of this chapter 6 and we will talk about pricing strategies using AI. So, these
are the things that we will cover in this module. We will start with an introduction to
pricing strategies using AI. Then we will talk about the journey to intelligent pricing and
steps to reach the goal and various ways in which AI can assist in pricing.

Now let us start with the first topic, that is pricing strategy using AI. As AI
transformation can vastly improve a company's existing data flows and processes and
those related to pricing are ideally starting points because they are usually well
established but often lack sophistication. A pricing maturity assessment conducted by
BCG and the Professional Pricing Society revealed that more than 50% of all industrial
goods companies still use Microsoft Excel to build their primary pricing tools and 25% of
B-to-B companies use static one-size-fits-all pricing with limited inputs and infrequent
updates. These companies are ripe for the kind of steps change that AI can provide.

So large companies often have a patchwork of pricing processes and AI can enable them
to raise and then scale their level of sophistication. So, these are revenue benefits from an
AI transformation. So, companies with AI initiative in areas other than pricing and
companies with AI initiatives that are focused on pricing. So, you see that here it is 1.7
times the change when they are using AI.

A MID BHI survey showed that large companies that undertook AI driven pricing
transformation achieved more than $100 million of revenue improvement, 70% more
than companies that focused on other areas. So, companies can win many pricing battles
with AI. Across industries companies can tackle pricing complexities using advanced
analytics approaches that take advantage of AI. Consumer packaged goods companies for
example can use insights from AI based products to rethink pricing for their overall brand
portfolio. Refine their pack price architecture and improve their mix management by
doubling down on products and channels that have higher margins.
CPG companies can also use AI based tools to improve the efficiency of their promotions
and tighten their trade terms. B2B companies can employ AI based analytics to mine rich
transaction data to find quick wins in terms of incremental price differentiation and
improved discounting. They can also use AI based tools to determine price metrics, set
price levels and manage price implementation. Companies should identify the battles
with the clearest and fastest upside relative to the investments and begin their AI pricing
transformation with those. The journey to intelligent pricing.

Generalized set of pricing based on static rules to pricing tailored in real time based on
demand, competition, customers' willingness to pay, all run by advanced algorithms and
AI. So, this is how this looks like. On the x-axis we have tailored one by one and then
here it is general one side fits one size fits all. On y-axis we have systematic and
automated, that is the change can be fast. Here it is ad hoc and manual, wait and see.

Now when we when the change is fast and it is tailored then it is called as intelligent
pricing. So, price is calculated in real time based on multiple variables such as demand,
competition, customers' willingness to pay using advanced algorithms. Now let us look at
what this personalized pricing is. So, when we are into ad hoc and manual processing and
but tailored by one, so that is personalized pricing. Every single customer gets a specific
price to match his or her maximum willingness to pay.

The third is rule based. Again, here it is general and one size fits all and it is ad hoc and
manual processing. So that is rule based. Minimum price discrimination based on static
rules that is price on weekends higher than weekdays. At this place fourth, where it is
systematic and automated fast change, but it is about general and one size fits all.

So that is dynamic pricing. All the customers receive the same price, but prices vary over
time. So, this is human generated, mostly transparent pricing. Here it is machine
generated, mostly non-transparent pricing. Properly strategized and implemented.

Intelligent pricing can increase margins and support growth. To reach that end here are
the few steps to keep in mind. The first is think big, start small and scale fast. Second is
build a business case and the third is build on industry norms. Now what is this think big,
start small and scale fast.

Use approaches like design thinking workshops to brainstorm an ideal end state for your
intelligent pricing strategies. Then put in place a proof of concept or several proof of
concept to test ideas for a small product segment or market. The third is scale
successfully, scale successful proofs of concepts across the enterprise as quickly as
possible. The second step is to build a business case. Lead with the business case
showing the potential revenue and margin growth from intelligent pricing.

Also highlight the benefits and risk related to the hallow effects on customers
relationships. Sales teams, partners and customers will all significantly benefit from
intelligent pricing in the short and the long term. The third step is to build on industry
norms. The trend towards real-time data-driven and more tailored pricing is inescapable.
The test for companies will lie in their ability to both masters as extracting value in the
B2C space and also scale it to transactions with their B2B and B2C ecosystem partners.

The future of intelligent pricing is one that reaches end to end across the value chain.
When this happens, companies will realize the full spectrum of benefits well beyond
steering sales and improving margins and revenues. It will increasingly also be leveraged
as a powerhouse to foster enduring customer relationships, ecosystem partnerships and
ultimately fuel innovation, competitiveness and growth. Now let us understand ways in
which AI can assist in pricing. Identify then eliminate the most unproductive customer
discounts and segments, freeing up more financial resources and time for those that
contribute to profits.

Stopping the leakage of revenues due to sub-optimal expensive customer investment has
value at any time. Companies’ first priority remains ensuring strong safety protocols for
employees and customers. They can then bring order to the discount chaos. To shore up
profitable revenues during a downturn, companies will need to avoid unnecessary
spending on uncontrolled and unproductive discounts. They should focus instead on high
return high return investments in their most valuable customers and identify the growth
opportunities with the level or shrinking customer base.

Companies often overlook it due to the long legacy of discounts, rebates and other
investments that have not been subject to the kind of scrutiny that management gives to
more explicit costs. Plugging the leaks entails such rigorous scrutiny. Determining which
investments to make, at what levels and what time and for which customer is a complex
task. Which investments, where what time and for whom. So many companies create
models of their transaction waterfalls even going as far as analyzing past actions to learn
where they might be unintentionally leaking profits.

However, the traditional methods does not answer important questions like are we
making investments in the right form? Two, how can we optimize or eliminate
unproductive discounts? Three, where should we free up investment to allocate to certain
types of customers? An effective approach using segmented statistical analysis to look at
the effect of a discount or other customer investment across all variables like product,
geography, customer segments and so on. This analysis can become the baseline for an
artificial intelligence supported engine to continuously analyze results and inform trade-
offs for different types of investments. It shows that transactions levers contribute most to
the profit, which one destroys value and what the pricing waterfall should look like in
order to maximize profitable growth. A recent BNN company research brief provides an
excellent example of how AI can be used to determine the effectiveness of discounts by
customer segments and the type of discount. The brief mentions how focused analysis of
discounts can help stop revenue leakage due to suboptimal expensive customer
investment.

One building product distributor has benefited from this analysis. For years the company
used a wide array of discounts across a diverse product and customer portfolio. Because
there was so much overlap and noise in the data, standard business analysis could not
confidently flag which investments had a positive return and which were wasted. The
company applied segmented statistical analysis to isolate which investments and at what
levels and with which customers drove the most incremental profit. For many products
and customer segments different types of discounts produced a negative return on
investment.

In other cases, different discounts produced a surprisingly high return on investment. By

optimizing investments this distributor could get 50% more profit out of the investments
using the following tactics. First, immediately eliminate investment in low return
customers, namely those that did not respond to incentives in a profitable fashion. Next is
to end ineffective discount programs. Of the five major discount programs the company
ran only two had a statistically significant effect on profits and the rest of the three were
not worthful.

Eliminate discounts on products that resulted in low profit growth or outright loss.
Rebalance the spending across the remaining viable discount programs. Using the money
saved from these steps increase spending on the highest ROI investments. So, the two
graphics summarize the key the briefs key findings. So, this is figure 1 on the left-hand
side.

For a building product distributor certain customer segment had similar average discounts
but the ROI obtained from each type of discount varied widely. So, you see that on this x-
axis we have this quality quantity discount the various types of discounts, quantity
discounts, customer tier discount, customer group discount, special discount and
discretionary discount. And then there are segments, and this is segment 1. So, the quality
discount total weightage was 4.7; customer tier discount minus 0.4 and for customer
group discount 1.0; special discount minus 0.3 discretionary discount minus 0.2. So, the
total weighted was -0.2; -0.2; 0.1; 0.4; 0.1; and 0.2. And on the right-hand side of the
figure 2 the building product distributors identified a large lift in profit due to optimizing
discount.

So, that is per customer annual profit. So, profit with no discount, net impact of
traditional discounting, profit with traditional discounts, eliminate discounts to low ROI
segments, eliminate low ROI discount types. Then there are eliminating discounts to low
ROI product categories, profits after eliminating non-productive investments, optimize
across discount types, optimize overall discount level and optimize after profits after
optimizing overall discounting. So, these are high confidence and then you have medium
confidence. Automating pricing rules with AI in revenue management system to increase
the total revenue.

Boston Consulting Group, the BCG found that 95% of successful digital transformation
initiatives utilized one or more revenue growth levers. 77% of the given digital
transformation's financial impact was achieved through the combined use of 6 revenue
growth levers. Improving pricing optimization with advanced techniques including AI
has the potential to deliver a 5% increase in total revenues and the 6 levers are as follows.
The first lever is targeted selling programs. The second lever is sales channel
optimization.

The third lever is sales force activation. The fourth lever is contract improvement. The
fifth is price optimization and the sixth lever is revenue assurance. So, what is this first
lever that is targeted selling program? Sales of strategic products or services or sales to
specific customer segments can be increased by for example, bundling components of
existing offerings to meet the needs of target customers. Converting one-time sales to
subscriptions or improving the effectiveness of targeted promotions and loyalty
programs.

Sales channel optimization. The efficiency of the sales coverage models can be increased
by, for example, expanding the size and capabilities of the inside sales function. To better
target underserved customer segments, clarify direct sales role such as hunters versus
farmers or partnering with third parties to increase penetration in markets with limited
sales presence. Sales force activation. The effectiveness of the sales force can be
improved by for example, increasing inbound marketing, improving account planning
and sales funnel governance and optimizing incentive structures for exceeding sales
targets.

Contract improvement. So, contractual agreements can be standardized, and profitability

improved by for example, in introducing big data tools and process. Improvements to
simplify and accelerate decision making in contract reviews and negotiations or aligning
service agreements with delivery capabilities. Pricing optimization. The sales process can
be accelerated, and more value captured through standardized pricing and rigorous
pricing discipline. For example, by automatic pricing rules in revenue management
systems or enforcing contractual pricing changes.

Revenue assurance. Revenue leakage that is the failure to collect revenue earned from
contracted services or products sold can be identified and stopped by for example,
applying machine learning and automation tools to quickly find leaked revenues and fix
systemic billing issues. So, these are the six revenue levers and their impact. So, as you
can see from the left-hand side there are these are the six revenue levers targeted sales
program increasing sales of strategic product or services. The second is sales channel
optimization that is increasing the efficiency of sales coverage model. The third is sales
force activation that is enhancing the effectiveness of the sales force.

The fourth is pricing optimization that is standardizing pricing and improving pricing
discipline. The fifth is contract improvement that is improving the profitability of
contractual agreements and the sixth is revenue assurance that is identifying and stopping
revenue leakages. So, this shows the typical financial impact, and this is the average time
of impact. So, in the first the impact was 3 to 5 percent and that happened within 6 to 12
months. In the second, that is the sales channel optimization, the total financial impact
was 2 to 3 percent and that happened within 3 to 12 months.

For sales force activation the total financial impact was 2 to 5 percent and the total
average time to impact in months was 3 to 12. For the fourth one, that is pricing
optimization, the total financial impact was 2 to 5 percent again and the average time to
impact was 3 to 9 months. For contract improvement it was 1 to 2 percent and that was
between 6 to 18 months. While for revenue assurance the total financial impact was 3 to
5 percent and that happened in 3 to 6 months. So, this light green is low range, and this
dark green is high range.

So, this is the high range. Again, these are the high ranges while these greens they are
low range. So, it means that in the initial 3 months the impact was less and then it started
increasing. Now, implementing the revenue levers. Phase 1 is rapid assessment.
Identifying, valuing and prioritizing revenue levers according to which ones will lead the
most value and can be implemented quickly.

Phase 2 is quick wins. Designing and executing pilots to test the effectiveness of revenue
levers by demonstrating tangible impact. Evaluating lessons learned from pilots and
iterating solutions to maximize value capture. Phase 3 is scaling applying the most
effective revenue levers to new geographies, segments or products. Then capitalizing on
the many insights transactional data can provide by using AI and machine learning to
look for patterns in pricing, volume and mix analysis. The patterns and trending insights
in transaction data include new insights every business can use to become more
competitive.

Unlocking those insights take an AI based approach to interpreting the price, volume and
mix fluctuations often locked within the constraints of transactional data. Combining
transactional data analysis and price, volume and mix fluctuations have proven difficult
and a challenge to combine in a unified intuitive application. One of the companies
having success combining transactional and product mix data using AI is Vindabo. Their
approach is noteworthy in how it solved the usability challenges. So, many other price
optimization vendors have struggled with.

They have been able to deliver real-time price optimization driven by local market
conditions, competitive intelligence and cross-border parameters. AI and machine
learning to capture more revenues and profits by finding how, what are given customers
willing to pay or optimize price across their customers and product mix. Identify blind
spots in pricing, discounts and deal size decisions which are difficult to identify for
customers and products using spreadsheets alone. So, the problem lies in the
spreadsheets.

So, this is where the problem is. AI and machine learning helps pricing managers analyze
whether the existing discounts make sense by correlating deal size to discount made.
Identifying outliers where discounts have been granted due to the negotiating insight of
the customers. Lack of pricing discipline often results from inertia, poor leadership and
affection for the status quo. Especially in the B2B space where everything is negotiated,
and customer relationships are hard baked.

That is difficult to change. The change in ownership creates a golden opportunity to

shake things up. The process of harvesting quick wins allows the buyers to assess talent
and capabilities verifying words with leads. Early opportunities most often show up in
three areas. One is understanding customer willingness to pay.

Two is testing if discounts are rational. And the third is probing the price leakage. When
value is given away through unintended discounts such as relaxed payment terms or
expedited shipping. Many companies leave money on the table because they do not
appreciate what a given customer is willing to pay. So, they are not able to understand
what a customer is willing to pay.
Or how to optimize price across customers and product mix. So, it is about customer and
product mix. And that requires optimization. These blind spots make salespeople timid
about demanding price increases over the life of a contract. To define the scope of this
problem right away, firms can analyze whether existing discounts are rational by
correlating.

By correlating deal sizes to the discount made. This will identify outliers where discounts
have been granted due to negotiating with a human of the customer and not the deal size.
So, in this figure you can see that the this shows correlating discount and deal size can
help determine if a company's discounting behavior is rational. So, here we have small
deal sizes and here we have large deal sizes. Here we have high discounts and here we
have low discounts. Now you see that in this area where the deal size is low the discounts
are also low.

Now this is large. Similar size deals getting highly variable discounts. Similar size deals
but getting different kind of discounts. So, the deal size is high, but the discount is
varying from low to high. Similarly, these the dots represent customer transactions. These
are the various transactions and many small deals with unjustifiable large discount.

So, although the deal size was small, here in this the deal size is small but they are getting
high discount. So, these are the areas of concern. Here the deal size is big and the
discount is less.

So, again that is a problem. So, these two are the problem areas. So, AI is making it
possible to create propensity models by persona and they are invaluable for predicting
which customers will act on a bundling or pricing offer. By definition, propensity models
rely on predictive analytics including machine learning. To predict the profitability, a
given customer will act on a bundling or pricing offer, email campaigns or other call to
action leading to a purchase, upsell and cross-sell. Profensity models have proven to be
very effective at increasing customer retentions and reducing churn. Every business
excelling at omni-channel today relying on propensity models to better predict how
customers preferences and past behaviour will lead to future purchases.

With the help of AI, the model can automatically identify the customers which are prone
to buying multiple product, single products and those who would not buy at all. And so,
based on the profitability, sellers can target and offer different customers with different
bundle pricing and also at the same time different discounts. So, these are customer
profensities. The following is a dashboard that show how propensity model works.

So, this is the customer profensities and it is shown here on this dashboard. Price
optimization and price elasticity are going beyond industries with limited inventories
including airlines and hotels proliferating into manufacturing and services. All marketers
are increasingly relying on machine learning to define more competitive, contextually
relevant pricing. Machine learning apps are scaling price optimization beyond airlines,
hotels and events to encompass product and service pricing scenarios. Machine learning
is being used today to determine pricing elasticity by each product, factoring in channel
segment, customer segment, sales period and the product's position in the overall product
line pricing strategy. We will take the example of Microsoft's Azure Interactive Pricing
Analytics per configured solution.

So, PCS is an Azure cloud application providing a set of tools to recommend prices for
wholesale and retail products based on elasticity estimates from transaction records of the
past sales. It is targeted at mid-size companies with small pricing teams who lack
extensive data science support for bespoke pricing analytics models. The Pricing
Analytics PCS is an economical, full-featured application driven by the sales transaction
history of your business. It discovers products whose prices deviate from optimum and
suggests pricing promotion for sets of related products. A pricing manager can examine
sales predictions as forecasts, demand curves and cross-product effects as a function of
recommended price change then track the effect of the price changes over time on a
dashboard.

Given the abundance of historical data, pricing seems ripe for the application of machine
learning. However, there are many confounding effects like holidays and promotions.
Machine learning tradition prioritizes prediction performance over causal consistency.
But pricing decisions need high quality inference about causation.

Causation is most reliably established using randomized control experiments. But

experimentation, also known as A-B testing is extremely difficult to set up on existing
operations systems. Observational approaches do not require infrastructural changes but
tools that guard against the many statistical pitfalls are essential. The core concept is that
of price elasticity of demand. A measure of how sensitive the aggregate demand is to
price. Self-price elasticity is the percentage lift in sales of a product if you discount it by
1%.

Most consumer product have elasticities in the range of 1.5 to 5. Products with more
competitors are easier to substitute away from and will have higher elasticity.
Discretionary products are more elastic than staples. Elasticity, that is consumer
responses to price varies by current price points, time, sales channel, location, customer
segments from product to product and by other considerations. The PCS includes a model
that computes an elasticity for each combination of item, location, channel, segment and
week.

Optimal price is any price that optimizes some business objective. The most natural
business objective is the gross margin. The difference that is price minus marginal cost.
Suppose you sell products with negligible marginal cost that is electronic media then you
should lower or raise price until you reach elasticity of minus 1. The elasticity minus 1
you are in equilibrium. You use the dollar worth of sales quantity every time you raise
price enough to make a dollar from the higher prices and your gross margin stays the
same.

More frequently you have substantial marginal cost. Cost to make one more piece or
acquire it from the supplier. Then the constant minus 1 in the calculation is replaced by
price upon gross margins also known as inverse learner index. For example, suppose your
marginal cost of orange is dollar 4 and you are selling them for dollar 5 making the gross
margin of dollar 1. Your inverse learner index is minus 5 as you should increase prices at
lower elasticities.

So, where A is more than minus 5 and decrease them otherwise. These are the following
example is from Microsoft Azure’s interactive pricing analytics pre-configured solutions.
So, this is how this Microsoft Azure's interactive pricing analytics pre-configured
solution works. So, there are site names and channel names and customer segments. So,
there you can tick and and then find out and compare. So to conclude in this module we
have first given a brief introduction of how AI is transforming the pricing strategies in the
industry.

Then we have studied about the price journey towards intelligent pricing and some steps
to achieve the goal. Finally we have gone through the various applications of AI that can
assist in pricing and the same thing will be continued in the next module. These are the
nine references from which the material for this module was taken. Thank you.