How does a team change when AI takes on part of its work?
Torsten Schollmayer
AI prepares a draft, and a person has to carry on with it. For that to work, the team must agree who checks the result and who answers for it.
The strip of paper runs through several stations, the way a task moves through a team.
Picture a team that prepares answers to frequent customer questions. Until now, one colleague has gathered the questions, spoken to the people responsible and written a draft. From now on, AI will sort the questions and draft first answers. That changes what she works from and what she passes on to the others.
Decide what AI should work on
Before anyone briefs the AI, the team needs a shared picture of the result. AI can polish the wording of existing answers. It can also propose answers to questions nobody has answered yet. The second job needs far more preparation and checking.
In this example, the colleague could tell the AI to use approved documents only. Where something is missing, it marks the gap. She then knows what to ask the people responsible. The rest of the team can also see which work is done and which is still to come.
Whoever writes the brief needs to know which documents are current and who can answer questions on the substance. That preparation takes time, and the team should allow for it when it shares out the work. Otherwise it expects a finished text from someone who still has to gather the information.
Who does what
01
You define the task
You decide which questions get answered and which documents the AI may use.
02
AI prepares the draft
It sorts the questions and drafts answers from those documents. Gaps stay marked.
03
You approve the answer
You resolve the gaps with the people responsible and decide whether the text can go out.
The example follows a customer answer from start to finish. The team agrees who takes on which task.
Hand over the open questions too
A well-written draft can easily look finished. The next person may not know which documents it is based on. Nor can they tell whether anyone has checked what it says. So a handover needs a brief note on where the work stands.
The colleague might say, for instance, that she has checked the answers on delivery times. On returns, she is still waiting to hear from the person responsible. Whoever picks up the draft can carry on with the checked parts and chase that one question. Nobody has to retrace the whole job.
This doesn't need to be a full report. The next person needs to know what they can use and what they still have to clarify. The quickest way to find out how much detail that takes is to try it on the next handover.
Commitments come from people
A draft from AI is not a commitment to a colleague or a customer. Only a person who knows what they can deliver, and by when, can make one. An approach called Commitment-Based Management gives these agreements a structure. One person makes a clear commitment to another, and both know what it covers.
In the example, the colleague commits to delivering the checked answers by an agreed date. If she is missing information, she raises it before she commits. It might be fine to add the outstanding answers later. Or the team might need a different date. That is for the people waiting on the answers to decide.
Checking and chasing are part of the task, even if the AI produces the draft in minutes.
Walk one task from start to finish
A recurring task is a good place to start. Go through it once with your team: how the question arrives, who gathers the information, who checks the draft and what the next person receives. That process shows which part the AI should take on.
After the first handover, the team can see where explanations were missing or questions had to be asked. Those points shape the next agreements. On the How I work page, I describe who checks and who decides in my own work.
