AI for Board Papers
Board papers are the most consequential documents most organisations produce: they exist to let busy, accountable people make good decisions quickly. Having produced them at board level for years, I can tell you the work splits cleanly in two — assembling and structuring information (which AI now accelerates dramatically) and exercising judgement about recommendation, emphasis and sensitivity (which remains entirely, non-negotiably yours). Confusing those halves is how careers get damaged.
Why this task matters
A board paper isn’t a report; it’s a decision instrument. It must compress months of work into pages a director can absorb the night before, state a recommendation someone will be accountable for, and survive being read hostilely in a year’s time. That’s why they take so long: every sentence is load-bearing. It’s also why the drafting help matters — the assembly work routinely swallows the week you needed for the thinking.
The traditional workflow
- Gather source reports, financials and prior papers
- Draft the body: background, discussion, options, implications
- Distil the executive summary (usually last, usually rushed)
- Craft the recommendation wording with the sponsor
- Review cycles for accuracy, tone and political sensitivity
How AI can help
Draft
- First-pass body sections from your source material and dot points
- Options sections in parallel structure so directors compare like with like
Summarise
- Source reports and prior papers into background sections
- The finished paper into a disciplined one-page executive summary — often better done early, not last
Analyse
- Consistency: does the paper contradict the figures in its own attachments?
- Gap-checking against your board’s template: implications, risks, consultation, resourcing
- Reading level: ‘what would a director unfamiliar with this program not understand?’
What must stay human
The recommendation is yours — its substance, its wording, its timing. Political and stakeholder sensitivity is yours: what to emphasise, what needs a conversation before it appears in writing. Every figure is yours to verify against source, because ‘the AI drafted it’ has never once satisfied an audit committee. And the paper’s authority is the author’s and sponsor’s accountability — that cannot be delegated to software.
Traffic light assessment
Summarising your own source documents into background drafts. You hold the sources and can verify line by line; the content is assembly, not judgement.
Drafting discussion and options sections, executive summaries. Genuinely useful, but every claim and figure must be source-verified, and emphasis is a judgement you re-make on every paragraph.
The recommendation, sensitive commercial or personnel content, anything for a confidential session. Judgement, accountability and confidentiality converge here. Board material only ever touches tools explicitly approved for it — for many organisations, that means none.
Example prompt
For assembling a background section from sources you’re permitted to process:
Draft the Background section of a board paper from the attached project report and the previous paper on this topic. Audience: directors who read quickly and dislike padding. Constraints: 350 words maximum, strictly factual, no adjectives of enthusiasm, no recommendation language (that comes later and isn’t yours to write), and flag any figure that appears in both sources with different values rather than choosing one. Output: three short paragraphs — origin of the initiative, progress to date, current position. [paste sources]
The risks
Board packs are the most confidential routine documents in the organisation: strategy, financials, personnel, legal exposure. The tool-approval question isn’t a formality here — for many organisations the correct answer is that board material never enters generative AI at all, and you must know your organisation’s position before a single paragraph is pasted. Beyond confidentiality: verify every figure against source, watch for smoothed-over ambiguity, and never let AI wording of a recommendation stand — directors are expert detectors of sentences the author didn’t quite mean.
A better workflow
The current way
- A week of assembly steals the thinking time
- Executive summary rushed on the last day
- Review cycles catch assembly errors late
The AI-assisted way
- AI drafts background and structure from verified sources (where policy permits)
- You spend the reclaimed days on options, recommendation and the pre-conversations
- Executive summary drafted early, refined as the paper matures; consistency checks run before review
What improves
- Assembly hours become judgement hours — the trade the paper always needed
- Papers reach review cleaner, shrinking the cycle count
- The recommendation gets authored, not squeezed in
- Your name goes on work you’ve verified line by line
Key takeaways
- Board papers split into assembly (AI helps) and judgement (never delegated)
- Confidentiality rules decide everything — many organisations rightly exclude board material from AI entirely
- Every figure is verified against source; audit committees don’t accept ‘the AI wrote it’
- The recommendation, emphasis and sensitivity are permanently human work
- The prize is reallocated time: from assembling the paper to thinking the decision through