The problem it solves
Most AI failures are not technology failures. The technology often delivered exactly what was specified. The failure was in the specification — in the assumptions about what problem was worth solving, what outcome would look like, and whether the people and processes were ready.
By the time those assumptions surface, the investment has already been made. The question everyone is left with: why didn’t we ask this earlier?
The AI Due Diligence Mandate is designed to answer that question before the cost is incurred.
What eight weeks looks like
The engagement runs in three phases.
Weeks 1–3: Surface
We work through your AI strategy systematically — identifying every belief it depends on being true, from the obvious to the invisible. This includes individual sessions with key stakeholders, team sessions where the full picture can be assembled, and a documented map of everything we’ve found. Vendor claims. Internal assumptions. Strategic rationale. Competitive logic. All of it examined.
Weeks 4–6: Test
Each assumption is stress-tested. We apply the strongest available counter-evidence, model the scenarios where each assumption fails, and rank them by the damage they would cause. Some will prove solid. Some will not. The ones that don’t get documented before more budget is committed to them.
Weeks 7–8: Capability
We don’t hand you a report and leave. We work through the implications for your AI strategy, document the revised thinking, and run your team through the process itself — so you can apply it independently to every significant AI decision that follows.
What you’re not buying
You are not buying an AI strategy. You are not buying a vendor recommendation. You are not buying validation of a direction already chosen.
You are buying the answer to the question that should precede all of those things: what are we treating as certain that deserves scrutiny before we commit more to it?
Scope and duration
The AI Due Diligence Mandate runs for eight weeks. Long enough to do the work at depth across a complex AI investment landscape. Short enough that the deliverable is clear and the engagement has a defined end.
No dependency. No open-ended commitment. No follow-on work unless you want it.
"By the time your AI assumptions fail at scale, the cost is already sunk."
What you leave with
Not deliverables. Capabilities and clarity.
- A complete AI assumption audit Every belief your AI strategy depends on — documented, ranked by risk, and stress-tested against failure scenarios and available counter-evidence.
- A verdict on active AI investments Your in-flight AI programmes and proposed initiatives examined under structured adversarial scrutiny, with the strongest counterarguments on record.
- Identified blind spots before they become sunk costs The assumptions your leadership team had stopped examining — surfaced, named, and assessed before they become expensive.
- A permanent internal due diligence capability Your team trained in the process. Not dependent on external advice to run it again. The methodology yours to apply to every significant AI decision that follows.
Good fit / not a fit
- Good fit: You're committing to a major multi-year AI investment The scale of the commitment makes the cost of being wrong significant. You need more than a workshop — you need the whole picture of what your AI strategy depends on being true.
- Good fit: An AI programme is in motion and assumptions are surfacing in execution The direction was set. Investment is moving. But something in execution is revealing assumptions that weren't tested before the commitment was made. You need to understand which ones are load-bearing.
- Good fit: You want the capability built in, not just the output External due diligence is a one-time intervention. The Mandate leaves your team with the ability to run this process on every significant AI decision that follows.
- Not a fit: Not a fit Organisations looking for validation of an AI strategy already decided, or a consultant to produce a report that confirms a direction already chosen. This engagement is designed to challenge the strategy — not to confirm it.