What to Tell Your Team About AI

You are planning to automate something. Your team knows, because people always know, and the version circulating is worse than the truth.

Nobody will ask you directly. What happens instead is that the person who understands the process best becomes quietly unhelpful during the mapping, and the project takes twice as long for reasons nobody can name.

Silence is the worst option

Most owners avoid the conversation because they do not yet know the answer, and would rather not promise anything they might have to take back. That is reasonable and it backfires.

In the absence of information people assume the worst version, because the worst version is what they have read. And they act on that assumption for months before anybody says anything out loud.

You do not need certainty to have the conversation. You need to say what you know, what you do not know yet, and when you will know it.

What the evidence actually shows

Stanford’s AI Index for 2024 reviewed the research on assisted work and found it both speeds up task completion and improves the quality of the output, with the largest gains going to people who were previously less skilled at the task.

That last clause is the useful one and it is worth saying to a team directly. The effect is not that your best person becomes slightly faster. It is that the floor comes up, so somebody two years in produces work closer to somebody with ten years of experience.

For a business bottlenecked on one experienced person, that is the whole opportunity. For the people in that business, it is a genuinely different message from the one they are expecting.

Say these four things

What you are actually automating, specifically. Not a strategy. This process, these steps, starting next month. Specificity is what stops speculation, because vague statements expand to fill the available anxiety.

What happens to the time. This is the question underneath every other question and it deserves a direct answer. If the plan is to take on more work with the same team, say that. If the plan is to reduce headcount, you owe them the truth, and hiding it costs you the cooperation you need to build the thing.

What is not changing. Usually the majority of the job, and usually the part people care about. Naming it explicitly is more reassuring than any general statement about how AI is here to help.

Who to raise problems with. A named person, and an explicit statement that reporting a fault is welcome rather than obstructive.

The question you will be asked

If not out loud, then in the room afterwards: does this mean fewer of us.

Answer it honestly. If the answer is no, say why in a way that is checkable: we are not replacing anybody, we are taking on the work we currently turn down. If the answer is yes, say that too, with a timeframe.

People handle bad news considerably better than uncertainty. What they do not forgive is being told nothing was changing and then discovering it was, and that costs you the next three projects as well as this one.

Involve the person who does the job

Not as a courtesy. Because the project fails without them.

They hold every exception, every awkward client, every case the official process does not cover. No map is correct without them, and a system built from the documented process rather than the real one is wrong on day one.

Ask them to be the person who decides when it is good enough to run. That changes the relationship entirely: they are not having something done to them, they are the one signing it off. It also produces a better system, because they will refuse things a manager would have accepted.

What actually reassures people

Not the announcement. Three things that happen afterwards.

• The first thing you automate is something everybody hates. Data re-entry, chasing, filing. If the first project removes a task nobody wanted, the argument makes itself and no amount of messaging is needed.

• Somebody raises a fault and it gets fixed. This is the moment a team decides whether the system is theirs or something imposed on them.

• Nobody loses their job in the first year. Whatever you said, this is what gets believed.

The staff who will surprise you

Two groups behave differently from expectation, and knowing which is which is worth a great deal.

The person you assumed would resist is frequently the one already using these tools privately, because they are the one doing the most repetitive work. Ask before assuming.

The person who resists most is usually somebody whose value in the business rests on being the only one who knows something. That is a real concern rather than an irrational one, and it is not addressed by reassurance. It is addressed by giving them a role in the new arrangement that is genuinely worth having.

What not to say

Four phrases that reliably do damage, all of them well intentioned.

“It will free you up for more strategic work.” Nobody believes this, because in most businesses the freed time gets absorbed by more of the same work. If you mean it, name the strategic work specifically. If you cannot name it, do not say the sentence.

“Nothing is going to change.” Something is going to change, that is the point of the project, and saying otherwise means you will be caught in it.

“This is just a tool.” Dismissive of a real question. People are not asking about the tool.

“Everyone is doing this.” An argument for you, not for them, and it reads as a decision already made without their input.

Say it to the whole team at once

Not individually, and not through managers relaying a summary.

Told separately, people compare what they each heard and the differences become evidence that something is being concealed. Told through a layer, the message arrives in five versions with five degrees of confidence behind it.

One room, one message, questions taken openly, including the one about headcount. Fifteen minutes, and it saves months of the quiet version of this conversation happening without you.

Have it early

Before the project starts, not after it is built. A conversation before is a consultation. The same conversation afterwards is an announcement, and everybody can tell the difference.

Sources

• Stanford Institute for Human-Centered Artificial Intelligence, Artificial Intelligence Index Report 2024, 15 April 2024.

AI Optimize starts every build with the person who does the work today, because they hold the exceptions and no system is right without them. That work sits under Workflow Automation.