Why "just use ChatGPT" is bad advice for your company

The cheapest AI advice in any boardroom is also the most popular: just tell the staff to use ChatGPT. It costs nothing, it sounds pragmatic, and it lets everyone leave the meeting feeling that the AI question has been handled.

It has not been handled. It has been scattered.

Here is the distinction that gets lost. One person using a chatbot gets faster. That part is true, and it is why the advice sounds reasonable: the executive who says it has usually tried the tool, been impressed, and generalised from their own experience. But an organisation is not a person. Forty people quietly pasting things into consumer chatbots is not an organisation using AI. It is forty private, inconsistent, unauditable workflows, and nobody in the building can tell you what was sent where, what came back, or which of it ended up in front of a customer.

I want to be precise about where I stand, because this is not the argument that AI is overhyped. I build with these models every day. They are in the daily work behind the products I ship, from bookkeeping to writing tools, and they are genuinely powerful. That is exactly the problem. Powerful tools with no boundaries do not produce a powerful organisation. They produce fast, confident, invisible mistakes.

The failure modes are specific, and if your company runs on “just use ChatGPT,” some of them are already happening.

Your data is leaving the building. The moment a staff member pastes a customer list, a contract, or a pricing sheet into a free consumer tool, that information has left your control, under terms nobody in your company has read, in a country nobody checked. Nigeria now has a data protection law with real penalties. “A junior staff member pasted it into a chatbot” is not a defence you want to rely on, and the uncomfortable part is that you currently have no way of knowing how often it is happening. In every conversation I have had with business leaders about this, the honest answer to “what has already been pasted where” is that nobody knows.

The answers are confidently wrong in ways your staff cannot check. These tools do not hesitate. They produce polished, fluent, plausible text at the same speed whether they are right or wrong, and the polish is the trap. A wrong number in a scruffy handwritten note invites a second look. The same wrong number in a clean, well-formatted paragraph sails through. The people most likely to lean on the tool hardest are the people least equipped to catch it when it is wrong, because if they could catch it, they would not have needed it.

The capability walks out the door. Left unmanaged, one person in the office gets genuinely good at this. Their prompts, their workarounds, their sense of what the tool can and cannot be trusted with, all of it lives in their head and their personal account. Then they leave, and your company’s entire AI capability resigns with them. Nothing was ever written down because nothing was ever official.

Nothing accumulates. This is the quietest failure and the most expensive one. Forty people solving the same problems separately means the same lessons get learned forty times and kept zero times. There is no shared way of doing things, no process that gets better each month, no system. Individual productivity went up a little. The organisation learned nothing.

So what does the sensible version look like? Not a ban. Banning the tools just drives the same behaviour underground, onto personal phones, where you have even less visibility. And not a grand programme either. I wrote last week about the five questions that replace an AI strategy, and the answer here is the same shape, applied narrowly.

Pick one queue, the place where information sits and waits for a person to read it. Give the tool an official place inside that existing workflow instead of asking people to wander off to a chatbot and paste things. Name one owner whose job is to make it work and to say how it is used. And decide, in writing, what kinds of data may leave the building and what kinds may never. That last one is a single meeting and a single page, and it puts you ahead of most companies on this continent.

Then, only after that one queue works, take the next one.

The deeper issue with “just use ChatGPT” is what it delegates. Whether AI touches your customers, your contracts, and your numbers is a leadership decision. The scattered version hands that decision to whoever happens to have a browser tab open, which means your most junior staff are making it forty times a day, unsupervised, with no record. Adoption done properly is the same decision made once, deliberately, by the people whose job it is to decide.

The companies that get this right will not be the ones whose staff used chatbots first. They will be the ones where the organisation, not just the individuals in it, got smarter.

I write one of these every week, on AI and African business as they actually work. If your company is somewhere between “we should use AI” and knowing what that means on a Tuesday morning, this is the conversation I keep having, and I am happy to have it with you.