COLUMN
Generative AI license renewal.
If you can’t state usage, cut these three first
Renewal is due, and the only line you can write is “people are using it.” Before anything else, look at three things: who used it, for which task, how many times. Any team or license that can’t answer comes out of the renewal. Cut first, then measure. Only what you can measure goes into next year’s renewal.
“People are using it” is not a renewal case
A year after rollout, the renewal request goes up. Product name and amount are easy. The usage line stays blank. To fill it, someone asks the floor, “are you using it?” The answers all come back the same: “yes, we use it.”
That sentence gives management nothing to decide with. A case nobody can decide tends to go through at last year’s seat count. The following year, the same blank is still there.
What you cut first is not opinion. It is who used it, for which task, and how many times. Any scope where that can’t be written comes out of the renewal. Measuring what remains and renewing every seat are two different conversations.
What is left the year after “people are using it” gets approved
| What went into the renewal case | What is still there a year later |
|---|---|
| Floor interviews all said “we use it” | No definition of usage. The next case starts from the same question |
| Seats issued were counted as users | Seats issued and people using it weekly don’t match |
| The whole company was renewed in one line | Teams that don’t use it stay in. The improvement target becomes “everyone” |
The third one weighs the most. Keep the seat count without measuring and next year’s case just has more “we still use it” in it. Without a unit to cut, both the diagnostic and the improvement work end up scoped to the whole company.
The three questions that set renewal scope — who, for what, how often
If there are no numbers, use just these three cuts. A team with an unanswered question comes out of the renewal. Measure inside the cut scope. Only what you managed to measure stays in next year’s renewal.
- Who: not seats issued, but people who used it at least once in the last four weeks. Can you name the team?
- For what: instead of search, drafting documents, checking internal policies? Can you name the task? Usage without a task name collapses back into “we use it.”
- How often: times per week. Someone who tried it once and stopped, and someone who has it in their weekly routine, are not the same “user.”
Only teams that can answer all three on the spot belong in the renewal. Teams that can’t either sit out this renewal or measure first and decide later — split according to your own contract and approval rules. What most companies end up doing is keeping the measurement target and the renewal target off the same seat count.
The metrics themselves are covered in a separate article on the three adoption metrics (usage rate, answer quality, correction rate). This article is only about cutting renewal scope when those numbers don’t exist yet.
Take unmeasured scope out of the renewal
Company-wide looks easier. Renew company-wide, and the same blank tends to be there next year. The unit to cut first is not the company: it is a team, a license tier, or a task.
What you cut is not a failure. It is a decision to narrow what you measure. Once who/what/how-often can be written for that narrower scope, that scope — and only that scope — goes back into next year’s renewal.
If you bring in an outside diagnostic, it is also easier to run after the cut (our own Metamon works on the same premise). Collecting company-wide impressions rarely fills the blank in a renewal case.
A renewal case with no usage number — sort out the unit to cut before you submit it
We separate the teams that can answer who / for what / how often from the ones that can’t. That single line is the entry point to our Metamon diagnostic. No internal approval documents needed. One line on the form (a team name is enough if you’re not sure).
Separate the units to cut from the renewal →
References
Our service: Metamon (generative-AI usage diagnostic)
Related: Three structural reasons AI agent projects stall at PoC
This article is a general framing, not legal, regulatory, or contractual advice. Decisions on renewals, budgets, and contracts follow your own internal rules.