Keep, Change, or Cut: How to Review the AI Tools You Already Pay For

I encourage you to do two things before saying yes to an AI tool: Write down a baseline and put a recurring check on your calendar.

What I haven't shared is what to do with that baseline once you've had it up and running for a few months. That's on me. It's also the part that is easy to skip. A tool gets adopted and renewed while the card keeps getting charged.

Best-case scenario: It is being used regularly and providing value. Worst-case scenario: It was used a few times, then abandoned. Six months later, nobody remembers why you bought it.

We're heading into planning season. Before anything new goes on next year's list, I'd take some time to review what's already on this one.

Go back to the number you wrote down

When you ran the tool through the Value dimension, you named what would change if it worked and how you'd measure it. Find that note. If it's in a notebook somewhere, now is a good time to dig it out and then compare it to today.

Some of you will find the number moved in a way that matters. Some will find it barely budged. And many of you won't be able to tell. That last answer is more common than anyone likes to admit.

If you fall into that last camp, then there was never a real baseline. Not to worry, and there's no judgment here. I'm writing this because I've made that same mistake. The fix is unglamorous: Write the baseline down today and look at it again in 90 days.

How a tool turns into an expensive habit

Nobody adopts a tool planning to waste money on it. It happens slowly, one auto-renewal at a time. A few signs I'd watch for:

• You couldn't say what it does that a free option, or something you already pay for, doesn't.
• One person uses it. Or everyone used it for 3 weeks and then stopped.
• You keep it because canceling would mean admitting it didn't work.
• The time it saves goes straight into checking and fixing what it produces.

That last one is sneaky. It's the first of the six risks I mentioned in The Founder’s Case for AI Governance. The AI tool might hallucinate or give you the wrong answer. The problem isn't a crisis; it's a time-consuming activity that fills your calendar with extra scans and fact-checks.

Keep, change, or cut

Every tool gets one of three calls, and it's important to make those calls quickly. Most of the time, you already know.

Keep it when it hits or beats the baseline, people use it, and your risk check is still clean. Then renew it on purpose this time, with a new target for next year.

Change is for the tool that works but has something in the way. Whether it's a lack of adoption, it doesn't fit with how you work, or you bought it for one job ,and it turned out to be better at another. Pick one thing to change and check again in 90 days. Why choose one instead of all three? Because if you change three things, you won't know which one made the difference.

Cut it when nothing moved, when you can't measure it, or when the ground shifted under it, like a vendor rewriting its terms or a new regulation that now applies to you. Cancel, export your data, and jot down what you learned. Doing so will help you improve your next tool evaluation.

I know cutting a tool can sting, especially one you championed. But that's the rollback plan from the Risk dimension, working the way it should.

What does this mean?

This review means starting the new year with a leaner budget and more knowledge about what works in your business. Not what worked in a vendor demo. Your next evaluation will start from evidence instead of enthusiasm.

Also, your team will notice and understand the courage it took to drop something that didn't deliver. Your decisions teach them more about your standards than a policy ever would. And when a client asks how you use AI (they will, if they haven't already), you can tell them what it does and what it delivers.

Your year-end action step

Block 45 minutes this week, before your planning session for next year. List every AI tool you pay for or your team uses, including the free ones people signed up for on their own. Next to each, write the baseline ("none" is allowed), where it stands today, and one word: keep, change, or cut.

Here are some examples of what that might look like:

• The writing assistant that speeds up first drafts but doubles your editing time? Change. Narrow it to one task.
• The meeting tool nobody has opened since spring? Cut.
• The transcription tool that shortened your follow-up time against the baseline you set? Keep, with a new target.

Related articles by Lori Lalonde:

The Founder’s Case for AI Governance

The AI-Ready Leader: Why Your Biggest Risk Isn't Falling Behind on Technology