how to onboard ai into a team when your ai only knows you

this morning i taped with a man trained in social psychology who spent decades delivering big projects inside large tech and media companies.

the episode is called your team isn't failing because of the tools. and he opened with one question that i haven't stopped thinking about.

he asks every person on a team to imagine this.

"think of your team as though we're a single person doing its best work on its best day to keep all of its promises and achieve all of its goals."

then... what would that person be like?

the closer everyone's answers are, the more effective the team. if the designer describes a careful perfectionist and the sales lead describes a fast closer who ships and fixes later, you've found your problem. it isn't the tools. it's that nobody agrees who the team is.

he says he validated it against gallup's q12 engagement survey across a hundred and sixty teams over two to three years. and he says it compresses the whole forming, storming, norming slog from months into half a day.

what comes out the other side is simple. a written list of dos and don'ts. how this team works. what it won't tolerate. what good looks like here.

then he made the point that turned it into an ai conversation.

every ai platform is built to learn an individual. your assistant learns your writing, your preferences, your shortcuts. mine learns mine. none of them learn the team.

so you end up with five people, five ais, five slightly different versions of how the company works. the tools make each person faster and the team more scattered.

his fix is the same one you'd use for a new hire. you don't hire someone and hope they figure out the culture. you onboard them. you hand them the dos and don'ts.

do the same with the ai. feed the team's list in as a prompt, the same one for everyone. now when someone asks it for help, the guidance is consistent with how the team operates, not just how that one person likes to work.

"that makes it a more active partner."

i like this because it's cheap and it works tomorrow. you don't need a new platform. you need a shared document that everyone's ai reads first.

picture it on a small sales team. one person's ai is tuned to their voice, another's to theirs. now two versions of the company are talking to the same customer. a better model doesn't fix that. agreeing what the team sounds like, and writing it down, does.

most founders asking how to onboard ai into a team start with tools. which model, which seat licence, which integration.

start with the question instead. who is this team on its best day? get the answers to match. write the dos and don'ts. then give that to every human and every machine you bring in.

you can run his question in your next team meeting. it takes ten minutes. everyone writes their answer alone, then you read them out. the gaps will be obvious, and a little uncomfortable. good. that discomfort is cheaper now than six months from now, when five people and five ais have each built their own version of the company.

long term, i think the companies that win with ai won't be the ones with the most tools. they'll be the ones where every tool knows the same truth about how the team works.

i've put my own version of that playbook in the ai method, including how i hand context to the machine.

so if you asked everyone on your team to describe that single person at its best... how far apart would the answers be?