AI Governance · from the CFO seat

Capital, controls, risk, and disclosure — for the AI era.


AI is moving faster than most boards can govern it. This is a working point of view — and a working library — for finance leaders, audit committees, and directors who need a defensible answer before the question arrives.

Atoms
Compute, data centers, energy, supply.

The capital side of AI — and the side software-trained CFOs underweight.

balanced by
Bits
Models, tokens, controls, disclosure.

The operating side — where finance owns visibility, controls, and the board narrative.

The shift

AI didn't make finance softer. It made it more industrial.


Two decades of software-led growth taught CFOs to optimise the bits: subscriptions, cohorts, gross margin, payback. AI quietly drags the balance sheet back toward the atoms — capital commitments, supply, energy, and physical risk.

That doesn't change what the CFO is responsible for. It changes what the CFO needs to be fluent in: capital allocation under uncertainty, control design around non-deterministic systems, and a board narrative that holds up to an audit committee.

The next decade of finance leadership belongs to CFOs who can hold capital discipline and AI fluency in the same hand — and translate both into something a board can sign.

Four pillars

What "AI governance" actually means from the CFO seat.


01 · Govern

Finance controls

Every AI tool touching the close, the forecast, or the reporting pack needs an owner, a control, and a paper trail.

  • Model & tool inventory
  • Human-review checkpoints
  • SOX / internal-control impact
  • Data lineage & PII boundaries
02 · Oversee

Board & audit committee

Directors need a written policy, a stated risk appetite, and a reporting cadence — not vendor demos.

  • AI policy & risk appetite
  • Committee mandate & cadence
  • External-standard alignment
  • Disclosure-ready narrative
03 · Allocate

Capital & infrastructure

Compute is a capital commitment. Build / buy / lease, scenario planning, and unit economics belong on the CFO desk.

  • Build / buy / lease analysis
  • Unit economics under scale
  • Cost, capacity & supply scenarios
  • Energy and physical footprint
04 · Equip

Finance teams

Adoption without controls is the most expensive path. The fastest finance teams build both at the same time.

  • Visibility into AI / compute spend
  • Software + industrial fluency
  • Controls-from-the-start adoption
  • Training that survives turnover
Self-assessment

Where does your AI governance actually sit?


Twelve questions, three pillars, one honest score. Directional, not a certification — answers stay in your browser.

AI Governance · Readiness check
12 questions · ~4 minutes

This is a directional self-assessment to start a conversation, not a certification or an audit. Your answers stay in your browser.

F-01
We keep an inventory of the AI models and tools used anywhere in finance, each with a named owner.
F-02
AI-assisted outputs in the close, forecast, or reporting pass through a defined human review before they're relied on.
F-03
We can trace the data feeding our AI tools, and we've assessed the impact on our internal-control and SOX environment.
B-01
There is a written AI policy and a stated risk appetite that the board or audit committee has seen.
B-02
A board committee has a clear mandate for AI oversight and a regular reporting cadence on it.
B-03
We've mapped our AI use against external standards (NIST AI RMF, ISO/IEC 42001, EU AI Act) where they apply.
C-01
We frame compute and AI infrastructure as a capital commitment, with explicit build / buy / lease analysis.
C-02
We track how AI affects unit economics and gross margin as usage scales, not just total spend.
C-03
We run scenarios on AI cost, capacity, and supply rather than assuming it stays linear or cheap.
T-01
We have tooling that gives finance clear visibility into AI / token / compute spend.
T-02
Our finance team has, or is building, fluency across both software-scaling and industrial / capital disciplines.
T-03
AI adoption in finance is designed with controls from the start, not bolted on afterward.
0 / 12 answered
The library

Frameworks, checklists & board-ready templates.


A working toolkit, built out one piece at a time. Tags show what's live and what's on the way — nothing here pretends to be finished before it is.

Primer · articleIn progress

The industrial-tech CFO

Why the AI era is pulling the finance seat back toward atoms — and what software-trained CFOs need to relearn.

Checklist · downloadPlanned

AI in the close: a control checklist

The control points to put around any AI touching your month-end, forecast, or reporting workflow.

Template · boardPlanned

An AI policy your audit committee can sign

A starting-point policy and risk-appetite statement, written to survive a committee review.

Framing · articlePlanned

CapEx for compute

How to frame compute and infrastructure as a capital commitment — build, buy, or lease — instead of a cloud bill.

Template · boardPlanned

AI risk one-pager

A single-page board reporting pack that gives directors a defensible line of sight on AI risk and spend.

Register · downloadPlanned

Model & tool inventory register

A structure for cataloguing every AI model and tool in finance, with owners, controls, and review dates.

Work with me

Bring this into your boardroom.


Fractional CFO engagements, board and audit-committee advisory, and AI-governance reviews for finance teams that need a defensible answer before the question comes from the board.