The AI License Renewal Question Nobody Can Answer

Why does availability look like adoption?

Because the metrics that are easy to collect all live on the vendor's side of the boundary, and the vendor can only see their own product. A license platform can tell you that a document was generated. It has no way of knowing whether that document was sent to a client, used in a pitch, or closed without being read. The one fact it cannot report is the only fact you are buying.

There is also nobody obvious whose job it is to report the other number. IT owns provisioning and can prove it. Procurement owns the contract and can prove it. The question of whether the work in a department changed sits with the department head, who has no dashboard, was not asked, and has a quarter to deliver.

There is a second complication, and it runs the other way. Microsoft's 2024 Work Trend Index reported that 78% of people using AI at work bring their own AI tools rather than using what their employer provides. If that pattern holds in your organization, then some of your real usage is happening outside the license you are about to renew, and none of it appears on the dashboard you are reading.

That makes the seat report wrong in both directions at once. It counts people who have access but change nothing, and it misses people who changed how they work using something they pay for themselves. The second group is the more interesting one, because it tells you which tasks people found worth the effort when nobody was mandating anything.

What should you measure instead?

The useful reframe is to stop asking how many people are using the tool and start asking what came out. Every metric below on the left is worth keeping. None of them should be reported without the column on the right sitting next to it.

What the dashboard shows	What it measures	What to put beside it
Seats provisioned	How many people could use it	How many produced a piece of finished work with it in the last month
Weekly active users	That the tool was opened	Which named deliverable came out of the session
Prompts per user	Volume of attempts	Whether the output entered the workflow: shipped, sent, filed, billed
Self-reported time saved	How people feel about the tool	Where that time went afterwards, named by the manager who redirected it
Training completion	That a seat holder attended a session	Whether a task that person owns is done differently now than it was in January
Licensed tool usage	Activity inside your contract	Which unlicensed AI tools are doing real work in the business, and on what

The right-hand column is harder to collect. That is the point. It is also collectible: a monthly question to team leads asking which of their recurring deliverables changed shape this month will get you further than any seat report, and it takes about ten minutes per team.

The work-changed test

There is a single question that sorts this quickly. Pick a recurring deliverable your team produced last week. A campaign brief, a client report, a board pack, a monthly forecast. Then ask whether the steps that produced it are different from the steps that produced the same deliverable a year ago.

If the answer is that the steps are identical and someone used an AI model somewhere inside one of them to go faster, you have a tool in the building. If the answer is that a step was removed, merged, or handed to a different person because the AI models made that possible, you have changed the work.

Both are legitimate. Only the second one has a path to the P&L, because only the second one changes what the organization can produce in a week.

Run that question across six recurring deliverables and you get a far better renewal case than any usage report, in both directions. Where steps have changed, you have named the value in language a CFO will accept. Where nothing has changed after a year of access, you have found the places where a redesign is the missing piece, and you know it before you commit to another term.

Why the link is called Usage

Average Robot describes the distance between what a leader expected AI to deliver and what it is delivering as the AI Profitability Gap™, and the chain that explains it runs Investment, Usage, Direction, Skill, Reinvestment, Return.

That second link was originally named Access, and it was renamed on purpose. Access is a purchasing state. It is finished the moment IT provisions the accounts, and it can be verified by anyone with admin rights in an afternoon. Usage is an operating state, it is never finished, and nobody owns it by default. Procurement's job ends at access. The distance opens in the space after that, where somebody has to decide what the tools are for and make sure the work in front of people changes.

This is why the first two links look healthy in almost every stalled AI program. The money was spent. The seats were issued. The evidence that both happened is abundant and easy to produce, which is exactly why it gets mistaken for evidence of progress.

Before you sign the renewal, try answering one question in the meeting: which deliverables in this business are produced differently because of this tool, and who here can name them? If the room can answer that, the renewal is easy and the number of seats barely matters. If the room cannot, another year of licenses will not change it, and the distance is somewhere past procurement.

The AI Profit Readiness Assessment is a free way to find out which link in the chain is the one holding your return, and how we help sets out what the work looks like once you know.