The Renewal-vs-Cancel Decision Framework
Renew or cancel is the wrong decision, and framing it that way is why most owners renew everything. Nobody cancels the thing that runs the business. The real decision for every software line is Keep, Swap, or Build: keep it because it earns its place, swap it for an AI-native version that does the same job for a fraction, or have an agent build exactly what you need. Three questions tell you which column a line lands in, and this is the framework for running them.
Why renew-vs-cancel fails
Put yourself in the owner's seat at renewal time. The tool does a real job. Someone on the team uses it. Cancelling means that job stops getting done, or lands on a person. So you renew. Every time. The binary framing has a built-in bias toward the status quo, because one of the two options is "break something."
That bias is exactly what the auto-renew depends on. The enemy is not the software. It is the auto-renew, and the middleman margin that depends on nobody asking a better question. The better question is not "should we stop doing this job" but "has the price of getting this job done changed since we signed." For a lot of your stack, the answer is yes, and not by a little. The agency owner in our mastermind who ran this across their stack found AI-native swaps at about a tenth of the cost and agent-built replacements at a tenth of that again. Rules of thumb, not quotes. But that is the scale of movement a binary question will never surface.
The three questions that sort every line
The audit asks the same three questions of every renewal, in order. The order matters because each question narrows the next.
- Is there an AI-native version? Same job, built after the models got good. If yes, Swap is on the table.
- Can an agent roll our own? If the data is reachable and you can name the three features you actually use, Build is on the table.
- Is it a rule or a judgment call? This does not decide whether you build. It decides how: rules go to code, judgment goes to an agent.
If the answer to the first two is no, the line is a Keep, and you ask again next renewal. If either is yes, you are choosing between Swap and Build, and the third question shapes the Build. The step-by-step version of running these on a real list is in how to audit your SaaS renewals in an afternoon.
Column 1: Keep
A line stays when it earns its place. Concretely: there is no AI-native version doing the same job, and either the data is locked inside the tool or the job is so judgment-heavy and so central that handing it to an agent today is not a trade you want to make. Accounting is a common Keep for that reason. Lots of judgment, little clean data to build on, high cost of error.
Keep is not a permanent verdict. It means the price of the alternative has not moved enough yet. Write down why in one line, put the three questions on the calendar for next renewal, and move on. Do not feel bad about a Keep. A stack with the right Keeps and nothing else is a stack you own.
Column 2: Swap
A line swaps when an AI-native version does the same job for a fraction of the price. Reporting, slide decks, transcription, scheduling, first-line support, most of the dashboard category. These are the fastest wins, and they are the ones that fund everything else on the list.
The mechanics: start a trial of the alternative before the renewal date, run both for a cycle, migrate, cancel the old one. Tell the daily user what is happening before you start, not after. Take the Swap column first, sorted by renewal date, because these are the lines where an afternoon of work produces a smaller bill next month.
Column 3: Build
A line gets built when the data is reachable and you can describe the three features your team actually uses. If both are true, the tool is a couple of hours of agent work with exactly those three features and none of the 58 you never click. The rule of thumb from the call: about a tenth of the swap price.
This is where the third question earns its keep. A task board is a good example. Moving a card when a due date passes is a rule. That is code. Deciding which overdue card to escalate to you is judgment. That is an agent. Build the rules as code and the judgment as an agent, and let them talk to each other. Build both halves the same way and you end up with either a brittle script trying to make decisions or an expensive agent doing arithmetic.
One member of our group built her own task board this way a couple of months in, with agents and without a developer. I told her she is now a developer. That is what the Build column does to a business over time: you stop renting the stack and start owning it.
The decision table
| AI-native version? | Data reachable + 3 features named? | Verdict |
|---|---|---|
| No | No | Keep. Ask again next renewal. |
| Yes | No | Swap. Trial, migrate, cancel. |
| No | Yes | Build. Rules to code, judgment to an agent. |
| Yes | Yes | Swap now, Build later. Take the fast win, schedule the build. |
That last row is the most common shape for anything important. One of the agency's lines was literally "swap this year, build next year." You are not doing this overnight. You are chipping away, and a line can move columns as the stack matures.
What about actually cancelling?
Sometimes a line goes to zero. Nobody has logged in for a quarter, the job it was bought for is not a job you do anymore. Cancel it. But notice that is not a decision, it is cleanup, and it is rare. Treating cleanup as the main event is how owners convince themselves they audited the stack when they really just deleted two zombie subscriptions and renewed forty others on autopilot. The real money is in the forty. The seven renewal mistakes covers how that money adds up.
Your next renewal is already on the calendar. It can auto-bill, or it can hit three questions and land in a column. The scorecard that makes this a fill-in exercise is free on the home page.
FAQ
When is cancelling outright the right call?
When nobody on the team has used the tool in the last quarter and the job it was bought for is no longer a job you do. That happens, but it is rare, and it is not really a decision. It is cleanup. For everything else, the job still exists, so the question is who does it and at what price, which is Keep, Swap, or Build.
What if a tool is half rule and half judgment?
Split it. The rules go to code and the judgment goes to an agent, and the two talk to each other. A task board is a good example: moving a card when a due date passes is a rule, deciding which card to escalate is judgment. Building both halves the same way is how you end up with either a brittle script or an expensive agent doing arithmetic.
Is Build always cheaper than Swap?
Cheaper on the bill, usually. Cheaper on attention this month, no. Build costs you a one-paragraph brief and a cycle of running the old tool and the new one side by side. Swap costs you a trial and a migration. Take Swap first because it is faster, and let those savings pay for the Build lines over the following quarters.
What do I do with a Keep line?
Renew it, write down in one line why it stayed, and put the three questions on the calendar for the next renewal. Keep is not permanent. It means the price of the alternative has not moved enough yet. Twelve months from now it may have, and you want the reasoning in front of you when you ask again.