AI Board Decision Memo Template
A copyable AI board decision memo: the decision, options, assumptions, evidence and its limits, owner, approval and stop conditions, and a review date.
Direct answer
Direct answer: AI board decision memo
An AI decision paper should let a board approve, reject, or stage a proposal in one reading. It should state the exact decision being sought, the real options including doing nothing, the assumptions the case rests on, what the evidence shows and does not show, who owns the outcome, the conditions attached, what would make the organisation stop, and when the decision returns. The template below keeps all of that to two or three pages.
The question
“How do we present an AI decision with clear options, uncertainty, ownership, and conditions for revisiting it?”
Who it is for
Executives and sponsors writing an AI paper for a board, committee, or investment approval, and company secretaries setting paper standards.
What this gives you
An editable decision memo with assumptions, options, approval conditions, and a review date.
- 01 A two-to-three-page memo structure that fits a standard board pack
- 02 A consistent record of what was decided, on what evidence, and with which conditions
- 03 A built-in review date and stop conditions, so the decision returns when it should
How to use the template
- □Write the decision sought as one sentence the board can answer yes or no.
- □Always include "do nothing" or "wait" as an option, with its cost.
- □Keep demonstrated, measured, vendor-claimed, and assumed figures visibly separate.
- □Put open questions in the conditions section, not in the narrative.
- □Agree the stop conditions before approval, when nobody yet has a stake in continuing.
Common weaknesses in AI papers
- □Capacity freed up is counted as cash saved without saying what happens to the time.
- □A vendor demonstration is presented as evidence of performance on the organisation’s own work.
- □The paper approves a whole programme when only the first stage has evidence.
- □No single owner is named; accountability is shared between technology and the business.
- □There is no date or trigger for returning the decision to the board.
After the meeting
- □Record the decision, any amendments, and each condition with its owner and date.
- □Log the review date in the board forward plan.
- □If the initiative is stopped, record what was learned. A well-documented stop is a good outcome.
Copy and adapt
AI decision memo
AI DECISION MEMO Title: Sponsor: Date: Decision sought (one sentence the board can answer yes or no): ... 1. CONTEXT What problem or opportunity this addresses, and why now. 2. WHAT THE SYSTEM WILL DO What it reads, recommends, decides, or changes. Which actions need human approval, and who can stop it. 3. OPTIONS A. Proceed as proposed B. Proceed with a smaller first stage C. Wait / do nothing (include the cost of waiting) Recommended option and why: 4. ASSUMPTIONS List each assumption the case depends on, and how it will be tested. - ... 5. EVIDENCE Demonstrated: Measured (baseline, sample, period): Vendor-claimed (not yet reproduced): What the evidence does not show: 6. COSTS AND BENEFITS One-off costs: Running costs after year one: Cash savings / capacity / risk reduction (state which): 7. RISKS AND FAILURE MODES Most likely ways it goes wrong, how we would know, and the response. 8. OWNERSHIP Accountable executive: Operating owner: Committee receiving evidence: 9. APPROVAL CONDITIONS Condition | Owner | Date ... 10. STOP CONDITIONS Results or events that would pause or end the initiative: ... 11. REVIEW Date or trigger for returning to the board: Evidence to be presented then:
Evidence
What this draws on
Regulated banking AI at Aveni
Architected enterprise banking AI with conduct-risk workflows, evidence generation, human review, escalation, evaluation, versioning, and release controls. Part of the Aveni team in the first FCA Supercharged Sandbox cohort; the FCA lists Aveni as an accepted firm, which is not an FCA endorsement or certification.
Inspect the source ↗Six companies founded, 2008–2024
Kwippy, Jaja.tv, Octo.ai, ExpressMOJO, Boom, and Neul Labs, spanning social software, interactive media, ML and analytics, logistics, blockchain infrastructure, and applied AI. Advisory and investment records are not counted as founded companies.
Read more →GenAI and Agentic AI Playbooks
21 open chapters in ten languages under Apache 2.0: a business-leader guide to GenAI adoption followed by an engineering and operating guide to agent systems.
Inspect the source ↗Scope and limits
What this is not
- —The template structures a decision; it does not supply the judgement, the evidence, or legal and financial advice.
- —Adapt section names and approval thresholds to your organisation’s own delegation framework and paper standards.
If the need is different
Common questions
Answers before you commission
What should an AI board paper include?+
The exact decision sought, options including waiting, the assumptions the case depends on, evidence separated into demonstrated, measured, vendor-claimed, and assumed, costs after year one, failure modes, a named owner, approval conditions, stop conditions, and a review date.
Is there an AI investment approval paper template?+
Yes. The plain-text memo on this page can be copied into a board or investment-committee template. Section 6 asks the sponsor to state which benefits are cash, which are capacity, and which are risk reduction.
Why include stop conditions?+
Because they are easiest to agree before anyone has a stake in continuing. They turn "we will keep an eye on it" into specific results or events that bring the decision back.
How do we record an AI leadership decision?+
Record the decision, any amendments, each condition with its owner and date, and the review date in the board forward plan. If the initiative is later stopped, record what was learned.
Related