Brief Your AI Like You Brief Your Team. Then Forget That Advice.

If you brief a team member by saying "can you put on some training", do not be surprised when one person comes back with AI training and another comes back with flying trapeze. Both answered the brief. Neither read your mind.

I say this on stage because it is the single most useful reframe I can give someone who thinks AI is underwhelming. Almost every disappointing AI output is not a tool problem. It is a briefing problem. You typed one line into a chat box, got one line of mush back, and concluded the technology was overrated. You did not manage it. You interrogated it.

So here is the advice everyone needs first. Then here is where that advice quietly stops being true, because the metaphor is more dangerous than it looks.

What carries across from managing people

The discipline of a good brief transfers almost perfectly. Everything you learned about getting quality work out of a person works on AI too.

You set context before you set the task. You would never drop a deliverable on a new hire with no background. You tell them who the audience is, what has happened so far, what good looks like, what to avoid. AI is identical. Onboard it. Tell it who you are, what you do, how you work, what you are trying to achieve. The model is only as good as the context you give it, and most people give it none.

You give structure, not vibes. A clear brief to a person has the goal, the constraints, the format, the deadline, the examples. The exact same structure lifts AI output from generic to sharp. Tell it the role to play, the audience to write for, the length, the format you want back, and one example of the standard you are chasing. Vague in, vague out. This is true of people and machines in precisely the same way.

You expect a draft, then you direct. No good manager accepts the first version of anything important and ships it. You read it, you react, you redirect. The first AI output is a starting position, not a finished product. The skill is in the second, third, and fourth instruction, the same way it is with a talented junior who needs a steer.

You manage up the quality bar. The biggest determinant of what you get back, from a person or a model, is the standard you make clear you expect. Set a low bar and you will hit it. This is why I tell rooms to be very, very, very unreasonable in what they expect of these tools. Push them like a top performer, because the ceiling you set is the ceiling you get.

If you took nothing else from this, you would already be in the top 10 percent of AI users. Brief it like you brief your team and the quality jumps overnight.

But now the harder part.

Where the metaphor breaks

"Brief your AI like your team" is a brilliant on-ramp and a terrible operating manual. Lean on it too hard and it will lead you into mistakes a good manager would never make with a person. Here is where you should deliberately do the opposite of what you would do with a human report.

You can be unreasonable in a way you never could with a person. Tell a junior to redo the work six times in an hour and they will, rightly, resent you. Tell them their first three attempts were not good enough and you have a morale problem. AI has no morale. No ego, no fatigue, no Friday afternoon. You can iterate ten times in ten minutes, throw the whole thing out, and start again with zero relationship cost. The management instinct that says "do not over-correct, do not bruise them" is exactly the instinct to drop. Push harder than you ever could with a human, because there is nobody on the other side to wear it.

You do not build its judgement over time. The whole point of managing a person is that they get better. You invest in a junior so that in six months they need less from you, not more. AI does not work like that. It does not accumulate judgement about your business between sessions unless you deliberately give it memory and context every time. A new hire you brief once; this you brief, in effect, every time. The work is front-loaded into the prompt, not amortised across a career. Treating it like a report you can train up and then trust is how people get burned.

You must verify, not delegate. When you delegate to a capable person, you transfer responsibility along with the task. You can stop checking. With AI you transfer the task but you keep all of the responsibility. It will state something wrong with the exact same confidence it states something right. A good manager learns who they can stop double-checking. With AI, you never stop. The output is yours the moment you use it, and "the AI said so" is not a defence that exists. This is the most important difference, and the one the team metaphor most dangerously obscures.

It has no stake in the outcome. A person on your team cares, at least a little, whether the thing succeeds. That care does quiet work, catching the obvious problem, flagging the thing you did not ask about, pushing back when the brief is wrong. AI will not push back unless you build the pushback in. It will cheerfully execute a bad brief perfectly. So the burden of catching a flawed premise stays entirely with you. You have to be your own devil's advocate, or instruct the model to be one, because it will not volunteer.

The synthesis

Hold both at once.

Borrow the briefing discipline from managing people, because that is what lifts your output from noise to signal. The context, the structure, the standard, the iteration. All of it transfers, and most people are leaving it on the table.

Then drop the rest of the metaphor. This is not a junior who will grow, soften under pressure, share the risk, or care if it is wrong. It is an instrument of extraordinary range with no judgement of its own and no skin in the game. So you brief it like your best team member, and then you push it harder than any human could bear, verify everything it gives you, and never hand it the responsibility along with the task.

Brief it like your team. Then remember it is nothing like your team. The professionals pulling ahead with AI are the ones holding both of those truths at the same time.