Can company secretaries trust AI with confidential boardroom conversations?

For the past three weeks, I have been speaking with company secretaries about CosecScribe and how AI can help them improve productivity and efficiency. The response has been interesting.

The problem is clear. Company secretaries spend significant amounts of time listening to meeting recordings, identifying speakers, reviewing discussions and eventually turning all of that information into properly structured minutes.

When I speak to company secretaries about what it would mean to take minutes off their plate, many of them tell me the same thing: if the minutes are out of the way, they can spend more time on compliance, governance and other important responsibilities, and even have a better work-life balance.

But there is another question that comes up almost every time:

What happens to the information in the meeting?

This is a very legitimate question.

Company secretaries sit in some of the most sensitive meetings within an organisation. Board meetings can involve trade secrets, financial information, business strategies, acquisitions, investments, internal decisions, corporate directions and other information that is not meant to leave the room. In many cases, this information cannot simply be handed over to another person or organisation without the appropriate authorisation.

So when we ask a company secretary to upload a meeting recording to an AI-powered platform, it is understandable that they ask: Who can access this? Where is it stored? Does the AI remember it? Can someone else see it? What happens if the platform is attacked? And what compliance framework protects us?

These aren't objections we should dismiss. They are exactly the questions that should be asked.

The challenge is that AI is still relatively new, and there is understandably a lot of uncertainty around how these systems actually work. So I want to explain how we think about this at CosecScribe.

AI doesn't automatically "remember" your board meeting

One common misconception is that once you give an AI system information, it permanently remembers everything you have ever given it. That isn't how AI systems inherently work.

For an AI system to retain information and make it available later, the application needs to be deliberately designed with some form of storage, memory or retrieval system that allows that information to be retained and accessed. That distinction matters.

At CosecScribe, we don't treat your meeting recordings as something that should simply sit on our servers indefinitely. Audio recordings, notes and uploaded files are not retained on our end as a permanent archive. Where a user chooses to save generated transcripts or minutes, those documents are stored securely and encrypted.

Our approach is simple: we should only retain what is necessary for the service we are providing, and customers should have control over what they choose to keep.

But what about hackers?

This is another important question. No serious technology company should tell you that a system is completely impossible to attack.

The right question is whether appropriate security measures have been put in place to reduce risk, protect information and respond to threats.

CosecScribe is built with security in mind, including infrastructure-level protection against threats such as DDoS attacks and safeguards designed to protect stored information. Where information is retained, encryption provides another layer of protection.

But security isn't a single feature you switch on once. It is an ongoing responsibility. As CosecScribe grows, our security, privacy and compliance frameworks must grow with it.

AI is coming into the boardroom

Whether we like it or not, AI will become part of how organisations work. The adoption curve will differ from company to company. Some organisations will adopt it quickly. Others will take a more cautious approach. But the direction is clear.

AI is already being used to assist with writing, research, analysis, documentation and administrative work. The boardroom will not be isolated from that transformation.

The question shouldn't only be, "Should company secretaries use AI?" It should also be, "What does responsible AI adoption look like for company secretaries?" That is the question we are trying to answer with CosecScribe.

We want to build AI with company secretaries, not around them

Building a generic AI product and then trying to sell it to company secretaries is very different from building an AI assistant specifically for the company secretarial profession.

The latter requires understanding the actual work. It requires understanding confidentiality, corporate governance, compliance and why a seemingly simple meeting recording can contain information that is extremely sensitive to a company and its stakeholders. Most importantly, it requires understanding the responsibility that comes with being in that room.

This is one of the reasons CosecScribe exists. We are not simply building a transcription tool. We are building an AI assistant for company secretaries. And the product has to evolve alongside the profession, not just in terms of productivity, but also in terms of privacy, security, governance and responsible AI adoption.

The future isn't about replacing the company secretary

It is about removing the work that prevents the company secretary from doing the work that matters most.

Listening to hours of recordings. Identifying speakers. Reconstructing conversations. Writing minutes late into the night. These are necessary tasks, but they don't have to consume the majority of a company secretary's time.

If AI can responsibly take care of the administrative burden, company secretaries can spend more time doing what they were actually appointed to do: governance, compliance, oversight and supporting better corporate decision-making.

That is the future we are building towards at CosecScribe. And because this is a profession where confidentiality and compliance aren't optional, we believe the technology serving company secretaries must be built with those principles from the beginning.

AI is coming to the boardroom. Our responsibility is to make sure that when it gets there, it is built for the people who understand the boardroom best.

This piece first appeared on Medium.