Reviewing an AI tool twelve months on, honestly

5 days ago
3 min read

Updated: 3 days ago

Introduction

Renewals happen by default. An invoice arrives, it is a familiar amount, it is paid, and nobody asks whether the thing it pays for is still earning its place. Across a dozen subscriptions this becomes a meaningful recurring cost that no one has examined since the original decision, which was made on expectations rather than experience.

A proper twelve-month review is an hour of work and it is the only point at which you have real information: what it actually cost, what it actually changed, what it broke, and whether the market now offers something better. It is also the moment to notice that a tool has quietly stopped being used, which is more common than anyone expects.

1. Reviewing an AI tool twelve months on requires that somebody schedules it

The reason it does not happen.

A date in the diary, a named owner, and a short agenda, set at the time of purchase. Without that, the review is triggered by a price rise or a budget crisis, which is a worse basis for the decision.

2. Start with whether it is being used

The first and most revealing question.

Active users, frequency, and by whom. Tools frequently survive in the accounts long after they have stopped surviving in the work, and this is the cheapest saving available.

3. Compare against the original baseline

If you recorded one.

The measure you chose at the start, then and now. If no baseline exists, note that as a lesson and use the best comparison available, but the absence should change how you do the next one.

4. Count what it actually cost

Including everything since.

Subscription, additional licences, the administration time, the review burden, and anything that had to be built around it. First-year total cost is almost always higher than the original estimate.

5. Ask the people who use it what they think

Different information from the numbers.

What they use it for, what they work around, what they would miss. This surfaces both the value nobody recorded and the workarounds that indicate a problem.

6. Look at what the market now offers

Twelve months is a long time here.

Alternatives, pricing, and capability that did not exist when you chose. The purpose is not to switch reflexively but to know whether your decision is still the right one on current information.

7. Check whether the terms have changed

They often have.

Pricing structure, data terms, features withdrawn or moved to a higher tier. Vendors change these and the notification is easy to miss, and a tool assessed on terms that no longer apply is assessed wrongly.

8. Decide explicitly, including the option to stop

The point of the exercise.

Continue, change plan, replace, or stop. Recording the decision and the reasoning means the next review begins from a stated position rather than from the invoice.

9. Reset the baseline for the next year

The improvement.

Whatever you decide, record the current measures so that next year's review has something to compare against. This is how the reviews get better rather than repeating the same uncertainty.

Be careful about stopping something whose benefit is diffuse. A tool with no measurable saving may still be preventing errors or holding a process together, and the absence of a number is not evidence of absence of value — though it is a reason to look harder.

Conclusion

Put the review in the diary at the point of purchase, because otherwise renewal is automatic.

Begin with whether anyone is still using it, compare against the baseline you recorded, count the full cost including administration and review time, ask the users what they actually do with it and what they work around, look at what the market now offers, check whether the vendor's pricing or data terms have changed, make an explicit decision that includes stopping as an option, and record fresh measures for next year's review.