Choosing AI Tools: The Meeting That Decides Nothing

Meetings about AI adoption run on product names. Which one costs less, which one clears the security review, what the competitor down the road is using. The deck comes together nicely. A comparison grid takes half a day. Procurement signs it off. Then you check in six months later and the only people logging in are the three who were already using something before the meeting happened.

This is written for people responsible for getting AI used inside an organization. It is also for anyone who sits in that meeting and has started to suspect they play the same role every time. The subject is not which tool wins. It is whose work changes, so it turns into a question about you fairly quickly.

The problem with a question that has an answer

"Which tool should we buy?" has one genuine virtue: it can be answered. Three candidates go in, one comes out. Budget follows, a rollout date lands on the calendar, the meeting concludes. It behaves like a decision.

And nobody's job changes. That is the catch. Through the whole selection process, every person in the room keeps the same responsibilities, the same authority, and the same things they are on the hook for. A question that is easy to close usually moves nothing.

When adoption turns out to be flat, the explanation tends to land on the staff: not curious enough, not skilled enough, too busy. But what got handed out was a login. Nobody handed out a new way of deciding things. Low adoption is the ordinary result of that, not a mystery.

Rewrite one line of the agenda

Here is what we suggest instead. Change one line before the next meeting.

From "which tool are we adopting" to "with this tool, whose decision changes?"

The wording barely moves. Answering it is much harder. To say whose decision changes, you have to already know who decides what. Suppose you hand summarization to AI. Somebody was reading those documents and deciding something on the strength of them. Their job shifts from reading to checking. Whether anyone on the team is good at checking is not a row in the comparison grid.

If the room cannot answer, that is not a failed meeting. You have learned that the information does not exist yet. More tool comparison will not produce it.

Six questions you can put on the agenda

When work gets redesigned around AI, these are the six we put on the table in corporate sessions. None of them contains a product name.

1. Who does the thinking?
2. Who makes the call?
3. Who carries the responsibility?
4. What do we hand to AI?
5. What do we do with the time this frees up?
6. What do we want to keep human?

Only the fourth looks like a tool question, and you cannot answer it without the other five. The order matters too. Start at "what do we hand to AI" and you slide straight back into reading feature lists.

The fifth is the one that gets skipped. Automate without deciding where the freed time goes, and it gets absorbed by whatever is nearest. That is how a team ends up faster with nobody feeling any relief.

A note on the term "AX"

This kind of redesign increasingly gets filed under "AX." The term does not have a settled definition yet. In Japan it is still under discussion in a government working group, and what has been published is a secretariat draft. We do not write "the ministry defines AX as" in articles or in proposals, because that would not be true.

An unsettled label is not a problem here. What the meeting needs to settle is not the name for this, but whose decision changes. That question works fine without a label.

You cannot guess whose decision changes

Working through the six questions, teams tend to stall in the same place. They do not actually know how people across the organization relate to AI right now. Ask, and the names that surface are the loud few.

RELACTIS is the tool we built for that gap. Twenty-four questions, about four minutes, and you get seven factors (Creation, Crossover, Autonomy, Advocacy, Systemization, Verification, Engagement) and which of fifteen types you sit closest to. It costs nothing and needs no signup. One person can take it to see their own shape. Run it across a team and you get the distribution by role.

Mapped onto the six questions, it looks something like this.

Read a distribution and the empty cells stand out more than the full ones. An organization with nobody in the checking role that automates summarization will circulate wrong summaries at speed. That is a seat nobody is sitting in, not a shortcoming in anyone. Once you can see it, you can appoint someone, borrow someone, or leave that step manual for now.

One limit worth stating. The assessment reads the shape of how someone engages, not the reason behind it. Why a particular person relates to AI the way they do is not something twenty-four questions will tell you. That part they have to say themselves, and the distribution is what gets that conversation started.

One thing for tomorrow

Next time an AI item reaches your agenda, change the heading. Not "tool selection," but "with this tool, whose decision changes?"

If the room goes quiet, that is the most useful fifteen seconds of the day. It also settles what you do next: go and look at how people across the organization actually relate to AI, rather than guessing at it.