AI board report drafting
Also called AI board pack drafting, Board paper summarisation, AI investor update drafting
Can AI write your board report? Experiments show faster drafting, but no study shows better board decisions. What it needs and where it goes wrong.
AI speeds up writing, but we found no study showing it improves board reports or decisions. In an MIT randomised experiment published in Science, professionals using ChatGPT cut time on writing tasks such as short reports by 40% and raised rated quality by 18%. Knowledge workers given AI on consulting tasks did 12.2% more tasks, but were 19 percentage points less likely to get a correct answer on a task outside AI's strengths. A board report's value lies in its numbers and judgement, and drafting speed does not improve either.
At a glance
| Status | Cosmetic |
|---|---|
| What it does | Drafts or summarises board and investor reports from management accounts, notes and previous reports |
| Typical cost | US$21 per user per month, billed annually, for an office-suite AI add-on, cut to US$18 for the first year under a promotion running to 31 December 2026 (vendor list price, checked Oct 2026) |
| Data it needs | Final management accounts, prior board reports and minutes, and a short list of the period's events and decisions |
| Setup effort | Low: a business AI account and a template; the work is gathering final inputs and editing |
| Evidence | Thin. Lab and field experiments on general writing tasks; no study of board reports or board decisions |
What is AI board report drafting?
Board report drafting uses a general AI assistant, or an AI feature inside office or board-portal software, to turn management accounts, notes and earlier reports into a written board or investor update. Typical outputs are a chief executive's report, a finance commentary, a risk summary, or short summaries of a long board pack for directors.
In an owner-led business the board may be just the owners, sometimes with an outside director or investor. The report matters because directors rely on it to oversee the company. In the UK, a director must exercise reasonable care, skill and diligence, and that duty does not move to a drafting tool.
This entry covers drafting and summarising. Producing the underlying numbers is covered under the automated weekly management pack.
How does it work?
You give the assistant the source material: the period's management accounts, last quarter's report, notes on key events, and a template or example of the structure you want. The model writes a draft in that shape, often with sections on performance, cash, people, risks and decisions needed. For long board packs it can also summarise each paper into a few lines.
The model works only from what it is given plus its general training. It does not know which numbers are final, what the chair cares about, or what was promised at the last meeting unless that is in the material. It writes with the same confidence whether the input is right or wrong.
Longer inputs carry more risk. Models use information at the start and end of long documents better than information in the middle, which is where the awkward detail in a board pack often sits.
What it needs to work
Drafting helps only when the numbers and decisions feeding it are settled. It needs final management accounts, a fixed report structure, and someone who knows the business to edit and check the result.
- Final, reviewed management accounts for the period (profit and loss, balance sheet, cash flow and aged debtors), not a draft that is still changing
- The previous two or three board reports and minutes, so the draft follows up on actions and earlier promises
- A fixed template: sections, length limits, and the five to ten KPIs the board tracks, with targets
- A short note of the period's events and the decisions needed, written by the chief executive or finance lead
- A business AI account whose terms keep your inputs out of model training, with access limited to directors and the finance lead
- A named editor who checks every number in the draft against the source accounts before it is circulated
Most use cases in this guide succeed or fail here, on the data underneath them, before any AI is involved.
What does the evidence show?
The independent evidence is about writing in general, not board reports. In an MIT randomised experiment published in Science, with 453 professionals, ChatGPT cut average time on tasks such as short reports and analysis plans by 40% and raised rated quality by 18%. A field experiment with 758 knowledge workers on realistic consulting tasks found AI users completed 12.2% more tasks and worked 25.1% faster on tasks within AI's strengths. On a task outside them, they were 19 percentage points less likely to reach a correct solution.
Summarising accurately is the weak point. Human reviewers of AI book summaries found omissions and unfaithful claims, and AI checkers did not reliably catch the unfaithful claims. One vendor's leaderboard, updated May 2026, shows hallucination rates of about 3% to 15% for well-known models when summarising short documents, and higher for some reasoning models.
Time saved does not automatically reach the bottom line: a Danish study linking chatbot use to official labour market records ruled out effects on earnings or hours above 2% two years after ChatGPT's launch. We found no study of AI-drafted board reports.
AI-drafted vs written-from-scratch board report
| Compared on | AI-drafted, human-edited | Written from scratch |
|---|---|---|
| Time to first draft | Minutes once inputs are gathered | Hours |
| Error profile | Fluent errors: altered or omitted facts that read well | Gaps and typos that are usually visible |
| Judgement and emphasis | Even-handed by default; misses what the chair cares about | Reflects the writer's view of what matters |
| Cost model | Per-user AI subscription | Senior staff time |
| Best for | Routine monthly updates with a stable structure | Bad news, major decisions or sensitive matters |
Status and evidence
Cosmetic: the independent evidence shows faster, more polished writing, which changes how quickly the report is produced rather than the revenue, cost or cash the board oversees. Evidence is thin: no study examines AI-drafted board reports, and a large Danish study rules out effects of chatbot use on earnings or hours above 2% two years after ChatGPT's launch.
Where does it go wrong?
- Altered figures. A summary that rounds, reorders or misattributes a number reads as smoothly as a correct one.
- The awkward item left out. The covenant breach or lost client buried in the middle of a long pack is exactly what a summary tends to drop.
- Bland prose. Drafts default to balanced, generic language that says little, and directors learn to skim them.
- Decisions not surfaced. A good board report asks for decisions. A model summarising the period rarely knows what needs deciding.
- Confidential papers in the wrong place. Pasting salary reviews, disputes or deal terms into a personal AI account can breach confidentiality terms and data protection law.
- Accountability drift. If directors approve a narrative nobody really wrote, nobody owns it. The duty of care stays with the directors.
Numbers it should move
Ask the demo business
Northgate Supply is a fictional wholesale distributor of catering and packaging supplies: 28 staff, R46.6m in sales over the last year. Its data is generated for this demo, to 30 September 2026.
Gross margin was 22.9% in September, against 24.4% last October. Packaging slipped most, from 26.1% to 22.9%, after a supplier raised prices and our sell prices stayed where they were.
How this was worked out
(Sales minus cost of goods sold) divided by sales, per month. Cost uses the supplier price on the day each line was sold.
Where the data usually lives
Where the records this use case needs are usually kept in a small business. It is not a list of recommended tools.
Frequently asked questions
Can I use AI to write my board report?
Yes, as a drafting aid. Experiments show AI cuts writing time sharply, from 27 to 17 minutes on professional writing tasks such as short reports in one study. But the draft is only as good as the accounts and notes you give it, and fluent summaries can leave out or alter facts. Have someone who knows the numbers check every figure, and write the decisions section yourself.
How much does AI for writing board reports cost?
Usually the price of a business AI subscription you may already have. One office-suite AI add-on lists at US$21 per user per month, billed annually, cut to US$18 for the first year under a promotion running to 31 December 2026 (checked Oct 2026). AI features in specialist board-portal software are typically quote-based. The real cost is the time to gather final numbers and edit the draft properly.
Do I need IT support to use AI for board papers?
No developer is needed to draft a report from documents you upload. You do need someone to set up a business account sensibly: who can use it, which files it can see, and whether inputs are excluded from model training. If the tool connects to shared drives, limit it to the board folder rather than the whole company drive, so it cannot pull in unrelated confidential files.
Are directors still responsible if AI wrote the board report?
Yes. In the UK, a director must exercise reasonable care, skill and diligence, judged against what is expected of someone in that role and the director's own knowledge and experience. Other countries, including South Africa, set their own director duties. Using AI to draft does not change who is accountable for what the report says or for what the board decides on the strength of it.
Can I upload confidential board papers to an AI tool under POPIA or UK GDPR?
Board papers often contain personal information: salaries, disciplinary matters, named customers. POPIA sets minimum requirements for processing it, regulates sending it outside South Africa, and is enforced by the Information Regulator. The ICO's AI guidance covers lawful basis, security and data minimisation under UK GDPR. Remove names where you can and check where your provider processes data. This is not legal advice.
More in leadership & reporting
Sources
- 1Experimental evidence on the productivity effects of generative artificial intelligenceScience (Noy and Zhang), 2023Study
- 2Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and QualityHarvard Business School (Working Paper 24-013), 2023Study
- 3
- 4Lost in the Middle: How Language Models Use Long ContextsTransactions of the Association for Computational Linguistics (via arXiv), 2023Study
- 5Hallucination LeaderboardVectara, 2026Vendor
- 6Companies Act 2006, section 174: Duty to exercise reasonable care, skill and diligencelegislation.gov.uk (UK Government), 2006Government
- 7Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AINational Bureau of Economic Research (Working Paper 33777), 2025Study
- 8Microsoft 365 Copilot for business (pricing page)Microsoft, 2026Vendor
- 9Protection of Personal Information Act 4 of 2013South African Government, 2013Government
- 10Guidance on AI and data protectionInformation Commissioner's Office (ICO), 2023Government
- 11Information Regulator South Africa (home page)Information Regulator (South Africa), 2026Government
This entry is educational. It is not legal, financial or professional advice, and it does not recommend any vendor. Figures attributed to vendors are their own claims.
- 1Experimental evidence on the productivity effects of generative artificial intelligenceScience (Noy and Zhang), 2023Study
- 2Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and QualityHarvard Business School (Working Paper 24-013), 2023Study
- 3
- 4Lost in the Middle: How Language Models Use Long ContextsTransactions of the Association for Computational Linguistics (via arXiv), 2023Study
- 5Hallucination LeaderboardVectara, 2026Vendor
- 6Companies Act 2006, section 174: Duty to exercise reasonable care, skill and diligencelegislation.gov.uk (UK Government), 2006Government
- 7Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AINational Bureau of Economic Research (Working Paper 33777), 2025Study
- 8Microsoft 365 Copilot for business (pricing page)Microsoft, 2026Vendor
- 9Protection of Personal Information Act 4 of 2013South African Government, 2013Government
- 10Guidance on AI and data protectionInformation Commissioner's Office (ICO), 2023Government
- 11Information Regulator South Africa (home page)Information Regulator (South Africa), 2026Government