The Great Unsubscribe: Why My AI Stack Is Now Two Tools and a Sticky Note
Last month I sat down with a coffee, opened my credit card statement, and counted fourteen AI subscriptions. Fourteen. Some cost $8 a month, one cost $30, and a couple were 'free trials' I'd forgotten to cancel since the spring. Total damage: roughly $187 a month for tools I could honestly describe, without exaggeration, as 'things that add items to lists.'
That number was annoying. What was worse was the spreadsheet I built afterward, tracking how much time I spent maintaining each tool — checking dashboards, re-syncing integrations, re-exporting notes when a vendor changed their export format. Two hours a week, conservatively. That's a hundred hours a year babysitting software that was supposed to save me time.
So I ran the experiment: cancel everything, keep only what survived a 30-day cold turkey test. Here's what happened.
The Audit Nobody Wants to Do
For each tool, I scored three things on a 1–5 scale: how often I actually opened it, whether its output showed up in anything I shipped, and how much upkeep it demanded. The pattern was immediate and a little humiliating.
The tools I loved — the ones with beautiful dashboards, clever automations, AI-generated weekly summaries — scored lowest on usage. I was admiring them, not using them. Meanwhile, the boring ones I'd almost canceled turned out to be load-bearing. My AI writing assistant was in the final draft of every client email. My meeting transcriber was the only reason anyone remembered what we decided in standup.
The rest? Digital houseplants. Nice to look at, quietly draining resources.
What Stayed and Why
After the month was up, exactly two tools earned their keep:
- One assistant I actually talk to. A general-purpose AI chat tool that handles drafting, summarizing, and the hundred small questions that used to mean opening a new browser tab. It's not specialized at anything, which turns out to be the point — it's the one place I go.
- One capture tool. Something that grabs what I read, hear, and think, and makes it searchable later. No AI summaries I never read. No weekly digest emails. Just a reliable memory.
Everything else — the project managers with AI copilots, the standalone summarizers, the 'second brain' apps that required a weekend onboarding course — went back to their free tiers or into the void. Two of them I'll genuinely miss. None of them I'll renew.
The Renewal-Season Gut Check
If you want to run your own version of this before your next billing cycle, here's the framework I used. It takes about 30 minutes:
- Pull the receipts. Not what you plan to use — what you actually opened in the last 30 days. Most tools have usage stats in settings; your bank statement fills in the rest.
- Trace the output. Pick your last three pieces of real work. Which tools touched the final product? Be honest. 'It could have helped' doesn't count.
- Price the upkeep. Every integration you maintain, every dashboard you check out of obligation, every notification you silence — that's a cost the sticker price doesn't show.
- Run the cancel test. Cancel anything that failed the first three checks. If you reinstall it within a month, it passed for real. If you don't, you just found $15 a month and two hours of your life.
The rule of thumb I landed on: if a tool needs a streak to justify its existence, it's a hobby, not a workflow.
The Uncomfortable Part
Here's the thing nobody selling you a productivity stack wants to admit: most of us don't have a tooling problem, we have an avoidance problem. Setting up a new app feels like progress. Configuring an automation feels like work. It's the most socially acceptable form of procrastination available to knowledge workers, and the AI boom has handed us an infinite supply of new things to set up.
There's also a status component, and I say this with full self-awareness. Mentioning that you run a five-tool AI pipeline at a coffee shop in Austin lands differently than saying you use ChatGPT and a notes app. But the people I know who ship the most — the freelance designer in Denver, the engineering lead in Chicago — all converged on the same boring setup. One assistant. One memory. That's it.
I'm not saying your stack needs to match mine. If you're a researcher, your two tools will look different from a recruiter's two tools, and that's fine. The number matters less than the discipline: every tool earns its seat annually, or it doesn't keep it.
The next time a launch tweet promises to '10x your workflow,' remember that your workflow is not the bottleneck. It rarely is.
