AI you can audit — Options & a Path

A single-file strategic options deck for a small audit firm standing up two service lines:

1. AI-augmented internal audit — governed, citation-grounded AI over the firm's own workpapers and licensed standards.
2. Auditing AI — assurance over clients' AI systems, built on the same evidence spine.

The deck was produced as a consulting artifact for a small audit firm (midmarket clients, Claude for Teams seats, Microsoft 365) and is published here as a public showcase edition — the client-identifying details are removed and the archetype is described generically.

Live demo

Open the live GitHub Pages deck.

Open it

ai-enablement-deck.html is the entire deliverable — HTML, CSS, JS, data and vendor libraries in one file (~2.4 MB), zero network dependencies. Open it in any modern browser, or serve it locally:

python -m http.server 8642
# → http://localhost:8642/ai-enablement-deck.html

Keyboard: P pauses motion (also a button, bottom-left). Print/PDF works from §12. The deck is authored as a dark cinematic piece — the theme stays dark by design (the retired light palette exists in code but was never production-checked).

What's inside

Twelve sections: the four pains → one root cause → an option spectrum (BUY + T0–T4, six cumulative tiers, lightest to heaviest) → total-cost modeling → the trade-off space (cost × quality × auditability, with an honestly empty corner) → the 12-field defensible audit trail → drift monitoring (PSI) → citation-grounded RAG with an on-prem n8n blueprint → a calibrated LLM-as-judge → the WORM evidence spine → a cloud-readiness baseline → the recommended path and a 30/60/90.

The thesis in one line: the cheapest, fastest, most defensible AI a small firm can run is the seat it already owns, governed at the harness layer — and the T1 gateway/observability layer is the unavoidable investment on every path.

How it was built

• Single self-contained HTML file; no build step required to view (the JS bundle is pre-compiled with esbuild).
• GSAP + ScrollTrigger + SplitText for scroll choreography, Lenis smooth scroll, Three.js WebGL backdrop, all vendored inline.
• An OKLCH design-token system with runtime assertions, dark + light themes, prefers-reduced-motion support, and print styles.
• All section content lives in one declarative DATA object; renderers read from it, so figures are editable in one place.

License

Original work in this repository is licensed under the MIT License © 2026 Rahim Siddiq. Bundled third-party components, including vendored libraries and any referenced vendor assets or code, remain under their own applicable licenses, notices, and terms; this MIT grant does not relicense them.

Honesty notes

• Figures are illustrative unless tagged as sourced. Pricing, tok/s, box counts and maturity levels are planning shapes, not quotes.
• Not legal, audit, or tax advice. Regulatory citations (FRE 502, SEC 17a-4(f), GLBA 16 CFR 314, EU AI Act Art 12, PCAOB QC 1000, SR 26-2, …) were verified against primary sources as of 2026-08-16; re-verify before relying on them.
• Vendor product facts (Claude for Teams/Cowork/M365 connector, model pricing, OSS versions) were current as of the same date — this space moves fast.
• Anything tagged ILLUSTRATIVE is a scenario, not a claim about a real client.

Repo layout

ai-enablement-deck.html   the deck (everything in one file)
README.md            this file