AI supports Annual General Meeting (AGM) preparation and governance administration, but live meetings carry greater risk. Learn where AI adds value and where caution is required.
Key takeaways
- AI supports company secretarial work around an AGM, including preparation, document review and first-pass drafting.
- Written resolutions and other structured governance documents are practical opportunities for AI-assisted drafting, subject to human review and local requirements.
- Using AI around an AGM is different from allowing AI to record or transcribe the live meeting.
- Live AI capture introduces confidentiality, privacy, retention, discoverability and data governance risks.
- AGM minutes are formal corporate records. AI assists with drafting, but qualified people remain responsible for their content and approval.
- A workflow that works in one country may not satisfy requirements in another.
- Companies need clear rules governing approved tools, inputs, users, storage, retention, review and sign-off.
Who invited the AI note-taker?
AI note-takers have become a familiar presence in everyday business meetings. They record conversations, generate transcripts, summarize discussions and pull out action items with little effort from the people in the room.
The AGM is different.
Shareholders are voting. Resolutions are passed. Quorum is established.
Discussions may involve confidential, commercially sensitive or personal information. The minutes will also become part of the company’s formal corporate record.
Allowing AI to record or transcribe that meeting creates an additional record that would not otherwise exist.
That raises some immediate questions. What exactly is being captured? Where is that information going? How long will it remain there? Who can access it? Could it later become subject to disclosure or production?
Most importantly: is creating that additional record necessary in the first place?
AI still adds value to the AGM process. But much of that value sits around the meeting rather than inside it.
An AGM is not just another call
Formal meetings have always required tighter control over information and records.
Depending on the meeting and organization, that can mean restrictions on recording, electronic devices or even the notes participants create and retain.
The guiding principle is simple: not everything said in a governed meeting needs to become a permanent record.
AI introduces a new dimension to that principle.
A tool that records or transcribes a meeting automatically creates a detailed digital record. That record may then exist separately from the carefully prepared minutes the company intended to preserve.
AGM and board minutes are formal corporate records. They are not simply a convenient account of what people remember discussing. Minutes are not transcripts.
| Ordinary business meeting | AGM or board-governed meeting |
| Notes are usually internal and informal | Minutes form part of the formal corporate record |
| Errors can often be corrected informally | Errors may require formal amendment |
| Retention may be discretionary | Retention may be legally mandated |
| Confidentiality exposure may be limited | Privilege and confidentiality obligations can apply |
| One jurisdiction may apply | Multiple jurisdictions may need consideration |
The transcript point is especially important.
A transcript tries to capture everything. Good minutes do not.
Minutes require someone to exercise judgment about what belongs in the formal record. That includes decisions about how dissent and sensitive discussions are documented.
An AI tool readily generates thousands of words without understanding why 200 carefully chosen words may be better.
That is not necessarily a failure of the technology. It is simply not the same job.
So where does AI actually help?
Treating governance as a reason to ban every new tool would miss the point.
AI adds value to AGM-related work when companies separate administrative assistance from governance judgment.
Before everyone enters the room
Plenty happens before an AGM starts, from preparing agendas and tracking deadlines to summarizing documents and coordinating logistics.
AI supports that work without becoming part of the formal meeting itself.
The usual controls still apply.
Companies need to consider what information the tool receives and who can access it. Someone also needs to review the output before relying on it.
But there is an important boundary here: AI is helping prepare for the meeting. It is not creating a parallel record of the meeting.
When the work does not require a meeting
Some governance work happens without a live meeting at all.
Where written resolutions or other structured governance documents are appropriate, AI supports first-pass drafting, organizes source information and helps prepare materials for review.
The same boundary still applies. AI should not determine whether a written resolution is legally appropriate, what approval threshold applies or whether the document satisfies local requirements.
Those decisions remain with the qualified people responsible for the entity.
Used this way, AI reduces administrative work without creating an unnecessary recording or transcript of a formal meeting.
After everyone leaves
AI also adds value once the AGM is over.
A company secretary or chair may provide selected materials for first-pass drafting. AI organizes that information into a starting structure.
The word selected matters. Those materials should also be provided only through approved tools and in line with the company’s confidentiality, data governance and retention requirements.
The tool is working from materials provided by the people responsible for the meeting. It is not quietly recording every word spoken in the room.
A qualified person then reviews and edits the draft before the minutes are approved. Human accountability remains with the person responsible for the final record.
Let AI help with the blank page, but do not let it decide what becomes history.
When the guardrails already exist
AI is easier to govern when nobody has to guess how it should be used.
Which tools are approved? Who can use them? What information can they receive? Where are outputs stored? When are they deleted?
Those decisions belong in the workflow before AGM-related information enters an AI system.
The strongest processes limit AI use to approved users, approved tools and predefined inputs. Review and retention responsibilities should also be clear.
Clear guardrails are also much easier to explain later if somebody asks how the corporate record was created.
AI helps write the first draft. It does not get the last word.
When does AI cross the line?
The harder question is what happens when AI moves from supporting the AGM process to participating in the meeting itself.
Live recording or transcription changes the equation because it creates a detailed record of what was said during a formal meeting. Information that might otherwise have existed only in participants’ working notes and the approved minutes instead becomes part of a persistent digital record.
That record may sit on a third-party platform. It may contain confidential information, personal data or commercially sensitive discussions. Depending on the tool, it may also be retained or processed in ways the company has not reviewed.
Creating that additional record may have consequences beyond the meeting. Depending on the jurisdiction and circumstances, transcripts and other AI-generated materials may become relevant to litigation, regulatory inquiries, investigations or disclosure obligations.
For a multinational group, the analysis becomes even more complex. Rules and expectations around privacy, data transfers, record retention and disclosure differ across jurisdictions.
Live recording should never happen by accident
An AGM should not be recorded or transcribed simply because someone’s AI assistant joined automatically.
If live AI transcription is permitted, the company should make that decision deliberately. The process may require policy authorization, participant disclosure and local legal review.
This is partly a confidentiality issue, but it is also a data governance issue.
Once an AI tool captures the meeting, the company needs answers to practical questions. Where is the transcript stored? Who has access? How long is it retained? Can it be deleted? Is the data used to train or improve the system? Could the record later need to be disclosed or produced?
Those questions are much easier to answer before the AGM than after someone discovers a complete transcript sitting on a third-party platform.
The fact that a tool can record the meeting is not a reason to let it.
A transcript should not become the minutes by default
Recording everything creates another problem: it encourages companies to confuse completeness with accuracy.
Minutes are selective for a reason.
They need to capture the business of the meeting and create an accurate corporate record. They do not need to preserve every aside, unfinished thought or poorly phrased comment.
That selectivity becomes especially important when discussions involve disagreement or sensitive subjects.
AI can summarize what was said. It cannot independently decide how dissent should be characterized or whether a discussion should appear in the formal record.
Those decisions require legal and contextual judgment. They should not be handed over to AI.
There is also a practical consequence to creating more information than the company needs.
A comprehensive transcript becomes another document that must be secured, retained or potentially produced later.
Sometimes the better governance decision is to create less.
Keep sensitive discussions out of the AI workflow
Live AGM and board discussions often contain information companies work carefully to protect. That makes feeding meeting content directly into an AI system a very different proposition from using AI for controlled administrative work before or after the meeting.
Privileged legal advice is an obvious example. Sensitive commercial negotiations or discussions involving personal data may also require tighter controls.
Those boundaries need to be clear before sensitive information enters the tool. Trying to recover control afterward may already be too late.
Approval of an AI platform for general business use does not automatically make it appropriate for live AGM or board content. Formal meetings need their own controls.
One global AI policy does not fit every entity
Multinational companies standardize processes for good reason. A common workflow can make governance easier across dozens of entities, so applying the same approach to AI may seem logical.
Headquarters approves a tool, establishes a global AI policy and rolls the same controls out across the group.
The problem is that AGMs do not take place in one legal environment.
Corporate law, data transfer rules and evidentiary standards vary between jurisdictions. Local requirements also shape recording, retention and the formal approval of minutes.
AI technology does not recognize those borders. The same platform may handle information for entities in London, Singapore or São Paulo using the same underlying technology. The legal and governance requirements still change with the jurisdiction.
For global teams, the goal is consistency without pretending every entity follows an identical process. A global policy establishes the baseline. Local review determines how it applies in each jurisdiction.
This becomes especially important before AGM season, when multiple entities may hold formal meetings within a relatively short period.
Standardize the controls. Validate the workflow locally.
What should an AI policy for formal governance actually cover?
A useful AI policy should answer the questions people face across the governance workflow.
People should know what information belongs in an AI system, what stays out and who makes that decision.
For multinational groups, a workable policy needs to address three things: what is allowed, who is accountable and what requires local validation.
Define what is allowed
Start with the tools themselves.
Companies should identify which AI platforms are approved for AGM-related work. The policy should also state clearly that general approval of a tool does not automatically extend to live meeting content.
Meeting type matters too.
An internal team call does not carry the same governance considerations as an AGM, board meeting or committee meeting. The policy should distinguish between ordinary operational meetings and formally governed meetings.
Live recording and transcription require an explicit position of their own. Employees should not have to infer from broader confidentiality language whether AI tools are allowed to capture a formal meeting.
The default position should be clear: live AGM or board content stays outside AI systems. Any exception requires deliberate authorization and appropriate local review.
Put a name next to the responsibility
AI governance becomes slippery when everyone is responsible for it in theory.
Formal records need an owner.
The policy should identify who reviews AI-assisted drafts and who approves the final minutes. It should also define the sign-off process across entities.
Retention needs the same clarity.
Companies should know where AI-generated drafts and source materials are stored. They should also determine how long those materials remain there and when they are deleted.
These responsibilities should belong to named roles rather than “the business” or “the meeting organizer.”
When a record matters, somebody needs to own it.
Leave room for local rules
A global policy is the starting point, not the final answer.
Local counsel or a qualified company secretary should validate how the global policy applies to an entity’s AGM. That review should consider corporate law, data transfers, record-keeping and local evidentiary requirements.
The same process should identify any governance information that requires tighter controls than the global policy provides.
This gives multinational groups a workable balance.
Headquarters does not need to reinvent AI governance separately for every entity. But it also should not assume a policy written for one jurisdiction travels perfectly across borders.
The policy should work when nobody has time to read it
There is another test worth applying to all of this.
Imagine the AGM is approaching and someone wants to use AI to prepare materials, draft a document or process information connected with the meeting.
The people responsible should be able to answer three questions immediately:
- Is this tool approved for this work?
- Is this information allowed to enter the system?
- Who is responsible for reviewing the output?
If those answers are unclear, the workflow is not ready.
Good governance does not mean turning every new technology into a compliance exercise. It means deciding where the technology belongs before sensitive information enters the workflow.
People should know who owns the minutes, where AI fits into the drafting process and what happens to AI-generated materials afterward.
That is what turns an AI policy from a document into an operating process.
For formal meetings, that difference is worth getting right.
Frequently asked questions about AI and AGMs
Can AI be used during an AGM?
Companies should distinguish between using AI around an AGM and allowing AI to capture the live meeting.
AI is well suited to controlled administrative work such as preparation, document organization and first-pass drafting.
Live recording or transcription is different. It creates an additional record and introduces confidentiality, privacy, retention, data governance and potential disclosure concerns.
The default should be to keep AI out of the live meeting unless its use has been deliberately authorized and locally validated.
Can AI draft AGM minutes?
Yes. AI is useful for first-pass drafting when it works from selected materials provided by the company secretary or chair. A qualified person then reviews, edits and approves the final record.
AI speeds up the drafting process. Accountability for the minutes stays human.
Should AGM minutes include a full transcript?
No. Minutes and transcripts serve different purposes.
Minutes create a focused corporate record rather than a word-for-word account of everything said. Unnecessary transcripts introduce additional confidentiality, retention and data governance concerns.
Can multinational companies use one AI governance policy globally?
A global policy establishes common controls, but it does not eliminate the need for local validation.
Corporate law, data transfer requirements and evidentiary standards differ between jurisdictions. Companies should validate AGM-related AI workflows at the entity or jurisdiction level before deployment.
What should companies do if they do not have an AI policy for formal meetings?
Do not improvise during the AGM.
Companies should first establish which tools are approved, what information they may receive and who is responsible for their use. The policy should also cover live recording, review, sign-off, retention, deletion and local validation.
Until those rules exist, keep AI out of the live meeting.
Put AI to work without giving up control
AI has a useful role in corporate governance. It supports preparation, organizes information, assists with structured governance documents and gives company secretarial teams a faster starting point for drafting.
That value does not depend on capturing the live meeting. Recording or transcribing an AGM creates an additional record, along with additional questions around confidentiality, privacy, retention and disclosure.
The strongest approach is deliberate: use AI where it reduces administrative work, keep live formal meeting content outside the workflow by default and leave governance judgment with the people responsible for the entity.
Managing governance across multiple entities? Talk to us about keeping corporate requirements clear, compliant and under control.
