Reviewing Your AI Spend After Six Months: Questions That Cut Through

Renewal is coming up and nobody can tell you whether the AI tool is working. The people who championed it say yes. The people who never opened it say nothing. The invoice arrives either way.

Six months in, three questions settle it: which workflows are still running on it, how many paid seats were actually used last month, and what breaks if you switch it off tomorrow. Everything else in a renewal review is commentary on those three.

Question one: which workflows survived

At the start you probably listed the workflows the tool was meant to change. Go back to that list and mark each one as still running, quietly abandoned, or never started.

The abandoned ones are the interesting entries. Find out why each stopped. There are usually three reasons and they lead to different decisions.

• It did not fit the work. Fine. Drop it and stop counting it in the business case.
• One person left and it went with them. That is an ownership problem, not a tool problem, and it will repeat.
• The output needed too much fixing. Worth checking whether the setup was wrong before concluding the tool was.

A tool that changed two workflows properly is a better buy than one that half changed nine. Count depth, not breadth.

Question two: which seats went unused

Pull the seat list and last activity date. Not logins, activity. Someone who opens a tool weekly and does nothing in it is an unused seat wearing a disguise. Divide the invoice by the seats that produced approved work and you have what you are really paying per approved output.

Then split unused seats into two piles. People who tried it and stopped, and people who never began. Those are different problems. The first pile is a product fit or trust issue. The second is usually onboarding, or a manager who did not make it part of the work.

If a large share of seats are dormant, the honest move is to reduce the seat count at renewal rather than launch another training push. Buying fewer seats and adding them back when demand appears is cheaper than paying for potential. It is also the renewal case for taking modules one at a time rather than all at once.

This is more comfortable when a tool is sold per seat with modules separable, which is how SicherOne is priced. You can drop a module a department never adopted without renegotiating everything.

Question three: what breaks if you switch it off

Ask this literally, as a thought experiment with a date. If the tool stops on the first of next month, what fails?

Three answers are possible and each means something different.

Nothing fails. Then you are paying for convenience, which is a legitimate purchase, but it should be priced as convenience and defended as such.

Something fails and there is a manual fallback. This is the healthy position. You have real dependency with a known cost of exit.

Something fails and nobody knows how it used to be done. This is the position to worry about, and six months is roughly when it appears. Write down the manual process for those workflows now, while somebody still remembers it.

The questions that sound useful and are not

Skip "are people happy with it". Satisfaction surveys about software measure recency and politeness. Skip "how many hours did it save", unless someone measured the same task before and after with the same method. Retrospective estimates of time saved are guesses dressed as data, and finance people know it.

Also skip comparing to a competitor's marketing. You are not deciding between claims. You are deciding whether the thing in front of you earns its invoice.

Check the boring contract items

Six months in is a good moment to reread the parts nobody reads at signing.

What happens to your data if you leave, and in what format. Whether the price is fixed for the next term or indexed to something. Whether seat reductions are allowed mid term or only at renewal. Whether any usage limits are close to being hit.

If the answer to the export question is unclear, get it clarified in writing before you renew rather than after, along with the questions about where your prompts are stored.

Turn the review into a rule

The single most useful output of a six month review is not the renew or cancel decision. It is deciding what you will measure for the next six months, and writing it down before the period starts.

Pick two workflows. Record how long the task takes and how often it needs rework, using the same method at both ends. That gives your next review something better than opinion to work with.

And write down what you decided

Renewals get repeated by people who were not in the room. A short note saying what you kept, what you dropped, how many seats you bought and why, saves the next person from starting the argument from scratch. It also makes it obvious, a year later, whether the reasoning held up.