Largely yes, with reservations. Drafting minutes or a decision log after a meeting, including recognizing action items and sharing these with those involved, is a task AI can largely handle today if a recording or a usable account of the meeting is available. What remains is checking for accuracy: did the text correctly capture who decided what and who is taking on which action.
Three axes are decisive: structuredness, judgment latitude, and volume.
Structuredness scores a 3. A meeting has a fixed form — agenda items, speakers, decisions, actions — but the content differs every time. AI recognizes the pattern well (who says what, what is agreed), but must determine anew each time what is relevant to record and what was noise.
Judgment latitude also scores a 3. A choice is needed: is a remark a decision, a suggestion, or just a thought no one picks up? Anyone who has attended a meeting where three people talk over each other and someone changes their position halfway through knows this distinction is not always obvious. A language model can usually assess this well based on context, but not flawlessly.
Volume scores a 4. Companies with a culture of meeting — weekly team meetings, project updates, steering committees — produce a lot of this kind of report. That very volume makes it attractive to automate: the same kind of task, repeated over and over, with limited variation in form.
Customer contact and physical presence both score a 5, the most favorable point on the scale. This is not because AI necessarily manages this effortlessly, but because this task requires neither. Drafting minutes happens after the meeting, not during it; there is no live interaction with a customer and no physical action needed. That makes the task in principle deployable anywhere a meeting is recorded or summarized.
Creativity scores a 2, the least favorable point in this profile. Minute-taking is largely reproducing, not inventing. There is little room, and little need, for personal style or interpretation — in fact, a minute-taker who adds too much of their own input does their job worse. That is precisely why this task lends itself well to AI: the required result is an accurate representation, not an original piece of text.
Error costs and compliance both score a 3, a middle position. An incorrectly recorded action item costs time to fix and can lead to confusion about who was supposed to do something, but rarely leads to directly measurable damage. In some sectors this is different: for a decision log that serves as formal evidence — for example in a works council or during an audit — errors weigh more heavily and precise checking of every word is needed. There the compliance axis shifts toward a higher score and the role of the human approving becomes heavier, not lighter.
A management team meets weekly for an hour. A recording or transcript goes to a system that drafts a decision log: who was present, what decisions were made, who has which action with which deadline. That draft is automatically shared via the team's collaboration platform. A participant — often the one who led the meeting — reads it through and approves it or sends it back with a correction. That last step, judging the account, remains human work. The same logic applies to management reporting for management or shareholders: AI drafts a first version, a human checks the content before it goes out.
At a company without a recording or reporting culture — where meetings run informally and no one takes notes — the basis on which AI can work is missing. The precondition is explicit: a recording or a clear account of the meeting. Without that input there is nothing to summarize. Also in meetings with a lot of technical jargon, sharply alternating speakers talking over each other, or decisions that are only made implicitly without anyone stating them, the chance of misinterpretation increases and human oversight becomes more important than in a meeting with a tight agenda.
What changes is not that minute-taking disappears, but that the first draft is no longer typed by a human while listening. That time is freed up for something else: actively participating in the meeting itself, or monitoring the follow-up of tasks and responsibilities delegated to a team. The check on the outcome remains, and in some meetings — where decisions carry legal or financial weight — that check remains just as heavy as before. How this shift plays out in a specific sector depends on how much meeting structure is already in place; see for example what AI can take over in wholesale for a sector view in which meeting volume and documentation obligations play a role.
Whether minute-taking is a task that can easily be shifted in your company depends on how your meetings are currently recorded and how heavily the decisions made in them weigh. Related tasks such as monitoring project progress against a schedule or organizing and scheduling a review cycle share a similar mix of structure and human judgment. The free quickscan from FTE TO AI consists of twelve questions, requires no account, and gives an indication of what portion of the hours in your profile can be taken over by AI today. The full work scan, which calculates the work of an entire company task by task into FTE capacity, is still under construction.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.