Am I Allowed to Use AI at Work?

Two thirds of office professionals have used AI at work when they believed it was against the rules. Twelve hundred and fifty people, across four countries, at large companies, in jobs that have nothing to do with IT.

That is not a few rule breakers at the edges. That is most of the room. And more than half of them were told informally to stop, which means somebody noticed. So let us answer the actual question. Am I allowed to use AI at work.

Almost everything written about this answers a different question than the one being asked. You are not asking whether AI is good for the economy. You are asking whether you are going to get in trouble on Thursday.

And there is a second thing underneath it, which is that the pressure runs in both directions at once. Fifty eight percent of managers say AI skill now affects how they judge someone. Twenty nine percent of employees think it does.

So the people being judged have no idea they are being judged, and the people doing the judging have not said so out loud. Forty eight percent of employees admit they have exaggerated how much they use it. Forty five percent feel pressure to use it without feeling confident with it.

That is a lot of people performing a competence nobody has offered to teach them. There are only four situations you can be in, and the answer is different in each one. One. There is a written policy and it says yes.

Then you are allowed, within whatever it says. Read the whole thing once. Most of these policies are two pages and most of that is about data, not about permission. The part worth finding is which tools are approved, because a policy that says yes to one tool is not saying yes to all of them.

Two. There is a written policy and it says no. Then you are not allowed, and using it anyway is the situation two thirds of people are in. I am not going to tell you that is fine. If it goes wrong, the policy is what everyone will point at, and you are the one who typed.

What you can do is ask why. A lot of these bans were written in twenty twenty three, in a hurry, after the Samsung story went around. Engineers there pasted internal source code into a chatbot. The company banned the tools outright.

That was a reasonable response to that problem. It is not a permanent answer, and plenty of those bans have never been looked at since. Three. There is no policy at all. This is where most people actually are, and it is the one that feels safest and is not.

No policy does not mean permission. It means nobody has decided, so if something goes wrong the decision gets made afterwards, about you, in hindsight. The fix is boring and it works. Ask, in writing, and keep the reply.

One line to your manager. I would like to use a chatbot to draft first versions of my reports. No client data goes in. Is that alright. Whatever comes back, you now have an answer instead of an assumption. And if the answer is no, you have saved yourself the problem.

Three things make that one line work. Name the tool, name the task, and name what you are keeping out of it. A manager who hears I want to use AI has to imagine the worst version. A manager who hears I want to use it to draft an internal summary, with no names in it, has almost nothing to object to.

Do it in email or chat, not in a corridor. The point is not the permission. The point is that the answer exists somewhere other than your memory. Four. You are the one who decides. If you run the place, or the team, the question is not whether to allow it. People are already using it. The question is whether they are doing it where you can see.

A policy that says no, with no enforcement and no alternative, does not stop the usage. It moves it onto personal phones, personal accounts, and out of your sight entirely. You will get a better outcome from naming two approved tools and one hard rule than from a ban nobody follows.

Two tools, because one is a single point of failure and five is a shopping list nobody reads. One rule, because people remember one. And say what happens when somebody gets it wrong. A policy with no stated consequence is read as either toothless or terrifying, and people guess which, usually badly.

And there is one thing that is true in all four situations. Nothing goes into a chatbot that you would not email to a stranger. Not a client list. Not somebody’s personal details. Not a contract, not medical information, not anything with a name attached that is not yours.

Take the names out and send the part you actually need help with. Most of the time the useful part of the question survives that perfectly well. That rule holds whether you are allowed or not, because it is not about permission. It is about what leaves the building.

So, the honest summary. If there is a policy, read it, and find out which tools it names. If there is no policy, ask in one sentence and keep the reply. If you are the one who decides, decide out loud, because the usage is happening either way.

And whatever the answer turns out to be, the names come out before anything goes in. You are not being reckless for wanting to use this. Two thirds of the people around you already are. You are just the one who wanted to know where the line was.

Subscribe. No permission required.