Advisory · Personal capacity
I work with finance teams and firms on where AI creates real leverage and where it quietly introduces risk — from workflow design to governance and the judgement layer that can't be automated.
AI has made financial analysis cheap. It has not made judgement cheap. The value has moved to whoever can tell whether an output is right, defensible, and worth acting on — and this work is about building that verification into how your function operates.
The four tests
Before an AI-enabled workflow earns a place in a finance function, it has to survive the same four questions — regardless of the tool, the vendor, or how impressive the demo was. These are the questions I bring to every engagement.
Test 01
Leverage
Where AI-enabled finance workflows remove avoidable effort without hiding how the work was produced.
Real leverage compresses the mechanical work — gathering, reconciling, drafting, restating — while leaving the reasoning legible. The failure mode is a workflow that looks faster but buries how the answer was reached, so no one can stand behind it later. The point isn't speed for its own sake; it's freeing experienced people to spend their attention where it actually changes the decision.
Test 02
Risk
Where speed or scale creates new control, governance, and professional exposure.
The same automation that saves time can quietly move risk somewhere no one is watching — into an unreviewed model, an unlogged prompt, a control that used to be a person and is now an assumption. Scale multiplies small errors before anyone notices them. Mapping that exposure honestly, including the professional and regulatory kind, is what keeps an efficiency gain from becoming a liability.
Test 03
Verification
How outputs are checked, challenged, and evidenced before they inform a decision.
Production is cheap; verification is the scarce skill. A workflow is only as good as the step that tests its output — the check that can catch a plausible, confident, wrong answer before it reaches a decision. That step has to be designed, resourced, and evidenced, not assumed. When verification is real, an output arrives with its challenge already attached: here is how we know this holds.
Test 04
Judgement
Who remains accountable for the call — and what they need to know before making it.
A model can produce the answer, but it cannot answer for it. Someone still signs. That person needs to understand the assumptions, the limits, and the confidence behind an output well enough to defend the decision it shaped. Designing that sign-off — making sure the accountable person is equipped rather than merely nominal — is where an engagement earns its keep.
Engagements
Every engagement is scoped to your function and delivered in a personal capacity. Most begin with an audit and move into governance design; briefings can stand alone.
Engagement 01
A workflow-level review of where AI already sits, or is proposed, in your finance function — run against the four tests. You get a clear map of where the leverage is real and where the exposure is, with the trade-offs named rather than glossed.
Format — Fixed-scope review · findings & risk map · prioritised recommendations
Engagement 02
Turning findings into something operational: the verification steps, controls, and sign-off framework that let an AI-assisted output shape a real decision defensibly. Built to fit your existing governance, not to replace judgement with process.
Format — Governance & sign-off framework · verification steps · control documentation
Engagement 03
Grounded sessions for boards, leadership, and finance teams on what AI genuinely changes about financial work — and what it doesn't. No hype, no vendor pitch: the shift in where value sits, and what it means for your people and your risk.
Format — Board or team briefing · tailored to your context · Q&A
Process
A typical audit-and-governance engagement moves through five steps. The shape adapts to your function, but the sequence — orient, map, test, design, hand over — stays the same.
01
We agree what we're looking at and why — the function, the pressure it's under, and how AI is being used or considered today. Scope is fixed up front so the work stays focused and the cost is predictable.
02
We trace the workflow end to end: where the work is produced, where AI already sits, where it's proposed, and where the hand-offs and controls actually are — as opposed to where the org chart says they are.
03
Each step is run through leverage, risk, verification, and judgement. This is where the real picture emerges: which gains are genuine, which are cosmetic, and where exposure has quietly moved.
04
We design the verification steps, controls, and sign-off framework that make the workflow defensible — fitted to your existing governance, so an accountable person can stand behind the output.
05
You get documented findings, the governance materials, and a briefing for the people who own the decisions — so the work continues to hold up after the engagement ends.
Who it's for
Less useful if you want a tool recommendation or a generic AI strategy deck. This is about your workflows, your controls, and who is accountable when an output shapes a real financial decision.
A considered next step