Insights
The AI tool you bought three months ago is already behind. Here's why that shouldn't scare you off.
If you bought into an AI tool this year, there's a good chance a better, cheaper, or faster version of it exists today. That's not a fluke — it's the pace of the category right now. Tools and tool-expertise are turning over on something like a 90-day cycle.
Most owners hear that and conclude AI is too risky to bet on. I'd draw a different lesson. The tool changing doesn't mean the problem it was solving went away. Missed calls after hours, estimates nobody chases, invoices sitting too long — those bottlenecks don't churn. They're permanent until you fix them.
That was the short version, and it stands. What it deserves now is the working-out: why the churn is structural rather than a phase, what it actually costs an owner in practice, and how to buy — and keep — tools sensibly inside it.
Why does the category turn over this fast?
Three gears, all turning at once. The underlying models — the engines nearly every AI tool is built on — improve in leaps a few times a year, and each leap makes some products dramatically better and others suddenly pointless. The platform companies absorb features: a capability that was a whole startup's product last quarter shows up as a checkbox inside the big tools the next. And the cost floor keeps dropping — what was expensive to run in January is cheap by fall, so pricing that made sense at signing looks inflated two quarters later.
None of that is anyone cheating you. Everything else you buy for the business — a truck, a compressor, a phone system — depreciates over years. This category depreciates over quarters, and nobody hands you a depreciation schedule at signup. Once you see the three gears, the churn stops reading as risk and starts reading as terrain.
What does churn actually cost me?
Be precise about this, because the costs are real but they're not the ones the fear suggests. The tool you bought does not get worse — it does exactly what it did on day one. What moves is everything around it. The real costs, in descending order of what I see hurt owners:
The re-decision tax. Every quarter brings a fresh round of "should we switch?" — and evaluation time is owner time, priced at your rate whether you count it or not. The abandonment spiral: buy tool, lose confidence when something shinier ships, stop investing in setup, get poor results, conclude AI doesn't work — pay for the lesson twice. Stale configuration: a tool nobody revisits drifts out of tune with your business even as its category improves around it. And occasionally, real strandings — a vendor pivots or folds, and whatever lived only inside their platform leaves with them.
Notice what's absent from that list: "bought the wrong tool." In a fast category, there is no permanently right tool. There are right outcomes, rented from whatever's currently best underneath — and the rest of this piece is about buying that way on purpose.
How do I buy sensibly inside the churn?
Flip the object of the decision. Stop buying tools; start buying outcomes. Before any purchase, write one sentence: "This exists so that ___ stops leaking." Answer the calls we miss. Chase the estimates we send. If you can't complete the sentence with a leak you've actually counted, the purchase is a mood, and churn punishes moods hardest — a tool bought for a reason survives its category moving; a tool bought for a demo doesn't.
The one-sentence discipline also fixes the most common evaluation mistake in the category: comparing tools to each other. Feature-versus-feature comparisons reward whoever shipped most recently — in a 90-day category, that's a coin toss you re-flip forever. Comparing each tool to your written sentence rewards whichever one actually stops the leak, which is the only contest that pays you.
Then apply three churn-proofing tests. Portability: if I leave in a year, what walks with me — my data, my process, or just my logbook of payments? Connection over platform: does it work with my phone system and CRM, or does it want to become them? Platform migrations are the single most expensive way to experience churn. Commitment honesty: in a 90-day category, a multi-year contract mostly transfers churn risk from the vendor to you. Month-to-month costs a little more and is usually worth it.
Should I switch tools when something better ships?
Usually not, and here is the pivot the whole piece turns on: outcomes don't churn. The question was never "is my tool still the best in category" — it's "is my outcome still being delivered." If the missed calls are still getting caught and the estimates still get chased, your tool is doing its job, whether or not it demos well anymore. Switch when the outcome degrades, when the price stops matching what the leak actually costs you, or when the vendor's decline threatens your data. Don't switch because the new one's demo is prettier. The demo isn't the outcome.
There's a fourth honest trigger: sometimes a leap underneath genuinely changes what's possible — the new generation doesn't just poll faster, it handles the conversation your old tool fumbled. Those moments are real, they're worth catching, and knowing which quarter's noise contains one is precisely the judgment the category demands.
Which brings it home. The owners who get this right stop buying tools and start assigning the churn: someone — you on a quarterly review, a sharp employee with a standing hour, or someone whose actual job is watching this category — owns the question "is what's underneath still the right thing." The turnover becomes a line item somebody watches instead of a surprise you keep discovering. Watching it is the work I do, and which outcomes deserve a watcher first — that's the call that pays. When the reading's done, bring me the bottleneck: book a conversation.