Ajay Pundhir · Independent Advisory
I help regulated enterprises ship AI that lasts — and stands up when someone asks how it was built.
Independent advisory practice. Not a notified body.
The Failure Modes
Three patterns I have watched undo AI programs in regulated industries. Each one is avoidable. None of them is avoided by accident.
01
A name-brand firm builds you a roadmap. Six months later the partner rotates, the deck moves to SharePoint, and your team is back to translating slides into systems. I do not rotate. I do not outlive the work.
02
The model ships. Then someone asks for the risk file, the data lineage, and the human-oversight log. There is no file. We rebuild from memory under deadline. I write the artefacts as we go — so the audit gap never opens in the first place.
03
A regulator turns up two years after launch. The model has drifted, the original team has moved on, and the documentation describes a system that no longer exists. I leave you with frameworks and ledgers your successor can defend — long after I am gone.
The Canvas
15 questions. 5 pillars. One score that reveals your weakest link. The same diagnostic used across enterprise organisations—now available as a self-guided tool.
Deploying AI agents? The assessment detects this from your answers and adds 7 agentic-readiness questions automatically—mapping your autonomy level from L0 to L4.
Free
No login required · 8–10 minutes
$197
Powered by the same AI behind the Situation Room
$497
Used across 12+ organisations · Weakest-pillar scoring · Board-ready output
5 Frameworks · 5 Companion Tools
Five frameworks I have refined across enterprise engagements. Each one pairs with a free tool from the AI Exponent product portfolio — so the engagement leaves you with something durable.
Seven capability dimensions for autonomous AI agents.
Most organisations cannot run agentic AI safely yet. A7 names the seven dimensions that decide whether you can — from data infrastructure to human oversight protocols.
Companion tool
Pressure-test retrieval with RAG-BenchA working governance structure in 90 days, NIST-compatible from day one.
Most governance frameworks demand maturity you do not have. MVG starts with three questions you can answer in two weeks and grows to full NIST coverage without replacing what you built.
Companion tool
Operationalise it with RiskForgeThe measurable business advantage of AI that people actually trust.
IBM IBV: organisations investing in AI ethics report 34% higher operating profit. Trust Premium is how I quantify and capture that advantage inside an engagement.
Companion tool
See the AI Exponent product portfolioA structured way to inventory, score, and reduce hidden liability in your AI portfolio.
Like technical debt, ethical liability compounds silently. Every AI system shipped without bias testing, fairness audits, or governance review adds to a ledger that grows until the bill arrives.
Companion tool
Track open-source liability with LCCWhere shortcuts in AI design quietly accumulate downstream cost.
Ethical Debt is the lens for surfacing the design decisions that look efficient on day one and become expensive on day three hundred. It pairs with the Liability Ledger to make the cost visible early.
Companion tool
See the AI Exponent product portfolioHow I Work
Pick the door that matches the size of your problem. They all run on the same frameworks.
The AI Exponent product portfolio is open-source. Run RiskForge against your AI register, or pressure-test retrieval with RAG-Bench. No call required.
See the tools →A 30-minute self-serve diagnostic across five pillars. Returns the one weakest link that decides whether your AI strategy survives contact with reality.
Take the Canvas →Independent A7 readiness, phased build to L1 / L2, and an annual durability review. The verdict is honest enough to include "not yet." Capped at four to five engagements per year.
See the engagement →The signature engagement · now open
An independent A7 readiness verdict, a phased build to L1 or L2, and an annual durability review — for organizations crossing from assistive AI to autonomous agents.
Most engagements start with the AI Strategy Canvas — the free diagnostic that surfaces whether agentic transformation is the right next move. Tools, frameworks, and verdict — one ladder.
The Practitioner
I am Ajay Pundhir. Fifteen years inside enterprise AI — Nokia India, Samsung Research, American Express ML Labs, and now G42 / Presight in Abu Dhabi. Four countries, four regulatory environments, one repeated lesson: the part of AI that fails first is rarely the algorithm. The frameworks and tools on this site exist because I needed them on real engagements and they were not on the shelf.
Nokia India · Samsung Research · American Express · G42 / Presight
