The scope of AI analysis is agreed upon by drafting a written scope document, securing Chair and board approval, and communicating it to every director before any feedback is collected. The document defines what data the AI will process, what it will produce, and what it will never do. Settling these terms first protects candour, which depends on directors trusting how their words will be used.
For the Company Secretary, this is a governance task, not a technical one. The aim is to ensure that every director knows, in advance and in writing, how their feedback will be analysed. The sections below set out what the scope should contain, who must sanction it, how to communicate it, and how to manage exceptions.
What should an AI analysis scope document include?
An AI analysis scope document should state the purpose of the evaluation, the data the AI will analyse, the outputs it will generate, the confidentiality rules that apply, and the limits of automated interpretation. It should be short, precise, and written for directors, not for technologists.
A defensible scope document covers six elements:
- Purpose and mandate: why the board evaluation is taking place and how the findings will be used.
- Data inputs: which responses are in scope, such as questionnaire ratings, free-text comments, and any interview notes, and which are excluded.
- Analytical outputs: the themes, patterns, and recommendations the AI will produce, and the level at which results will be reported (board, committee, or individual director).
- Confidentiality and attribution: whether comments are anonymised, who can see raw responses, and the minimum group size before results are reported.
- Data handling: where data is stored, how long it is retained, and whether it is used to train any model.
- Human oversight: who reviews the AI output before it reaches the board, and how errors or misreadings are corrected.
The last element matters most. A rigorous evaluation treats AI as an analytical aid that surfaces patterns, while experienced human judgement interprets what those patterns mean for the board’s strategy, culture, and dynamics. The scope should say so plainly.
Who needs to approve the AI analysis scope?
The Chair should own and approve the AI analysis scope, with the full board confirming it before evaluation begins. The Company Secretary typically drafts and coordinates the document, and legal counsel or the data protection officer should review it for privacy and regulatory compliance.
Approval works best as a short, sequential process:
- The Company Secretary drafts the scope in partnership with the Chair.
- Legal counsel and the data protection officer confirm that data handling meets applicable privacy law and internal policy.
- The Chair approves the document and presents it to the board.
- The board formally notes or endorses the scope, with the decision recorded in the minutes.
Board-level endorsement is not a formality. A scope imposed by management invites suspicion; a scope the directors have themselves approved becomes a shared agreement. Where the evaluation includes assessment of the Chair, the Senior Independent Director or another independent director should oversee that element so the Chair is not approving the terms of their own review.
How should directors be told what AI will analyse?
Directors should receive a plain-language briefing, in writing and before they submit any feedback, that explains what the AI will analyse, how their responses are protected, and what will appear in the final report. The same information should be repeated at the top of the questionnaire itself.
Effective communication has three parts:
- A pre-evaluation briefing issued by the Chair or Company Secretary, summarising the scope document in a page or less.
- An opportunity to ask questions before the evaluation opens, either in a board meeting or through a short call. Directors who raise concerns early are far more likely to answer honestly later.
- A point-of-entry reminder on the evaluation form, restating confidentiality terms and what the AI will and will not do with free-text comments.
Avoid technical language. Directors do not need to understand model architecture; they need to know who sees their words, whether those words can be traced to them, and whether a machine or a person draws the conclusions. Frank answers to those three questions build the trust that honest feedback requires.
What happens when feedback falls outside the agreed scope?
When feedback falls outside the agreed scope, the AI analysis should not process it. The Company Secretary should flag the material, refer it to the Chair for a decision, and handle it through the appropriate channel, which may be a separate conversation, a confidential discussion, or a formal governance process.
Out-of-scope feedback arises more often than expected. A director may use a free-text field to raise a concern about a specific colleague, an allegation of misconduct, or a matter that belongs under a whistleblowing or conduct policy. Feeding such material into general analysis risks breaching confidentiality, distorting themes, and exposing the organisation to legal risk.
A sound protocol includes:
- A defined list of excluded categories, such as personal grievances, legal matters, and conduct allegations.
- A human review step in which the Company Secretary screens responses before analysis.
- A documented escalation route to the Chair, or to the Senior Independent Director where the Chair is involved.
- A record of the decision, so the board can demonstrate that sensitive material was handled consistently.
Where out-of-scope feedback points to a genuine governance concern, the correct response is to address it directly, not to discard it. The scope protects the integrity of the evaluation; it should never become a means of suppressing legitimate issues.
How often should the AI analysis scope be reviewed?
The AI analysis scope should be reviewed at least once a year, before each evaluation cycle begins, and again whenever regulation, technology, or board composition changes materially. A pre-cycle review keeps the document aligned with current law and current practice.
Several triggers warrant an interim review:
- New or amended data protection and AI regulation in the jurisdictions where the board operates.
- A change of platform, vendor, or analytical method.
- A change of Chair or Company Secretary, or a significant shift in board membership.
- Director feedback after the previous cycle indicating discomfort with how data was handled.
Continuous performance tracking, now possible with modern evaluation platforms, makes regular review more important, not less. When board performance is monitored throughout the year rather than once, the volume and sensitivity of the data increase, and the scope must keep pace. Reviewing it in 2026 and each year thereafter, and recording the outcome in board minutes, gives regulators and investors clear evidence of disciplined oversight.
How The Board Practice helps with scoping AI analysis
The Board Practice brings more than 19 years of board effectiveness methodology, refined across over 120 board performance programmes and multinational boards, to the question of how AI should be used in evaluation. Its work is built on close partnership with the Chair and on candid, forward-looking counsel.
- Tailoring the scope of analysis to your board’s specific dynamics, culture, and regulatory context.
- Combining AI-powered analysis with experienced human interpretation, so findings reflect strategy, relationships, and leadership, not data alone.
- Supporting the Company Secretary in briefing directors and building confidence in the process before feedback is collected.
- Identifying both strengths and development areas, with a multi-year plan monitored in cooperation with the Chair.
To explore how a rigorous, well-scoped evaluation can strengthen your board, learn more about board effectiveness evaluations or contact The Board Practice to discuss your requirements.
Related Articles
- How do you prepare historical board evaluation records for AI analysis?
- How do you record human edits to an AI-generated board evaluation report?
- What happens when a CEO leaves suddenly?
- What are the common corporate governance failures?
- How do you create a governance improvement roadmap using data generated by your board platform?