How many quiet days make a Copilot seat reclaimable?

Every tool that promises to find your unused Microsoft 365 Copilot licences is really doing one thing: applying a threshold. Somebody decided how many quiet days make a seat dormant, and every pound in the reclaim figure hangs off that decision.

Which means the first question about a reclaim number is not "how much?" It is "after how long?"

Both ends of the dial are wrong

Set the threshold tight — a fortnight, say — and the list fills up with people on annual leave, people on a project that happens not to be Copilot-shaped this month, and the finance team in the week before year end. Send that list to managers once and you will not be allowed to send another one.

Set it loose — six months — and the list is honest but arrives far too late to matter. You will be reclaiming seats from people who stopped using Copilot two renewals ago.

The useful range sits between those, and exactly where depends on things only you know about your organisation.

Three things the threshold has to survive

Annual leave. Any window shorter than the longest ordinary absence in your organisation will generate false positives on a predictable cycle. In most places that argues for at least 45 days, and for not running the reclaim pass in August.

Seasonal roles. Finance at year end, HR in the recruitment season, legal around a transaction. These users look dormant for a quarter and then become the heaviest users in the tenant. If a role has a season, the threshold has to clear it or the role needs excluding from the pass by name.

The once-a-month user who matters. Somebody who uses Copilot four times a month, each time to get through a document that would otherwise have cost them a day, is getting more value per pound than a daily user running meeting recaps. Frequency is a proxy for value, and it is not a very good one. Treat it as a prompt to ask, not as a verdict.

Bands, not a binary

The single biggest improvement you can make to a reclaim process is to stop treating it as a yes/no and split the population three ways.

• Active — leave alone. Nothing to do and no conversation to have.
• Low usage — a coaching list, not a reclaim list. These people have the licence and have touched it. Something is in the way: they have not seen it used in their own workflow, or the content they work with is not in a shape Copilot can help with. Twenty minutes with a peer who does the same job fixes more of this than another all-staff webinar.
• Dormant — the reclaim candidates, and still only candidates.

Running the low-usage band separately is what stops the whole exercise reading as a cost cut. It gives you something to do for the people you are not taking a licence from, which is most of them.

Reclaiming is a conversation, not a script

The mechanics are trivial — a licence is unassigned in seconds. The part that goes wrong is social.

Tell people before, not after. A short note to the user and their manager, naming the threshold, saying when the seat will be released and how to keep it, costs you a week and buys you the right to do it again next quarter. Expect a proportion to reply "actually I need that" — which is fine, and is itself a finding.

Keep the exclusions explicit. Executives, new joiners inside their first month, and anyone in a seasonal role should be out of the pass by rule rather than by somebody remembering. Write the rule down next to the threshold.

And re-measure afterwards. If the same names are dormant two quarters running despite a conversation, the licence is not the problem and a reclaim is now the obvious answer rather than a contested one.

Why the threshold has to be yours

A vendor default is a guess about an organisation the vendor has never seen. The moment somebody in a renewal meeting asks "why 30 days?" and the answer is "that is what the tool came with", the whole figure is negotiable.

The answer you want to be able to give is duller and much stronger: we used 60 days because our longest standard leave is five weeks, we excluded the four seasonal roles listed here, and on that basis this is the monthly spend. Nobody argues with that.

Copilot Insights makes the threshold an input rather than a hidden assumption: you set the dormancy window, it bands every user as active, low usage or dormant against it, and it prices the dormant band in money per month and per year. The scan is read-only and runs on activity metadata, so you can re-run it with a different threshold and see what moves before you commit to a number.

Related: why assigned is not adopted, where adoption actually stalls, and what to take into the renewal meeting.