I added it all up one Sunday and actually winced. I was paying for nine AI tools. Nine. Some of them I hadn't opened in a month. One of them I genuinely could not remember signing up for.
So I did the unglamorous thing. I cancelled half of them. Not as a stunt — as basic hygiene. And the surprise was that my work didn't get worse. It got simpler, and a little cheaper, and a lot less noisy.
This is the math, the method, and the rules I now use so I never end up there again.
I deleted half my AI subscriptions because most of them did the same job, just with different logos. The fix: keep one strong general AI assistant, keep one or two genuine specialists, cancel everything that overlaps, and add a hard rule that a new tool has to replace one before it's allowed in. Tool count crept up because subscribing is easy and unsubscribing is friction.
Nobody decides to pay for nine AI tools. It accumulates. A free trial here. A "this looks cool" there. A tool a friend swore by. Each one was a small, reasonable yes.
The problem is that subscriptions are designed to be easy to start and quietly forgotten. There's no monthly nudge asking "did you actually use this." So they just sit there, charging, like a gym membership for your software.
When I listed them out, the overlap was embarrassing. Three of them could write. Two could summarize. Two did basically identical research. I was paying multiple times for the same capability because I'd bought them at different moments and never compared.
What made it sneaky is that no single subscription was expensive. Each one was a small monthly number that felt trivial in isolation. But trivial numbers stack, and they stack silently, and a year of them is real money you never consciously decided to spend. Nobody would write one cheque for all nine at once — you'd balk. The damage is done a little at a time, which is precisely why it's so easy to miss until you force yourself to add it all up on a Sunday.
I gave every tool one question: what does this do that nothing else I pay for can do?
If the answer was "nothing, really, it's just nice," it was gone. That question alone killed four subscriptions in about ten minutes.
Here's roughly how the list sorted out:
| Tool's job | How many I paid for | How many I needed |
|---|---|---|
| General writing/chat | 3 | 1 |
| Summarizing | 2 | 0 (the general one does it) |
| Research | 2 | 1 |
| A real specialist task | 2 | 1 |
Three writing tools collapsed into one. The summarizers vanished because my main AI assistant already summarizes fine. I kept one research tool and one true specialist that did something genuinely different.
If a tool's only advantage is "it's a bit nicer," it's not a tool, it's a tax.
Not every cut was easy. One subscription survived because it did exactly one thing better than anything else, and that thing mattered to me weekly.
That's the test for a specialist. It can't be "slightly better at a general task." It has to be "clearly the best at a specific task you actually do often." A great AI assistant covers 80% of everything. A specialist earns its slot only by owning a slice of the other 20% that you genuinely live in.
Everything in the murky middle — a bit better at writing, a bit nicer interface, a bit faster — is exactly the stuff you should cut. The middle is where money goes to die.
The hard part of this test is emotional, not logical. I liked some of the tools I cut. They were pleasant to use. I'd developed little routines around them. Letting them go felt like throwing out something perfectly good, which is exactly the feeling that keeps people paying for nine subscriptions forever. The reframe that helped: I wasn't throwing away a capability, because the capability lived on in a tool I kept. I was just refusing to pay twice for the same thing because one version had a nicer button. Once I separated "I enjoy this" from "this does something nothing else does," the cuts got easy. Enjoyment is not a billing justification.
I expected to lose something. Here's what actually happened.
That last point shocked me. I assumed more tools meant more capability. The opposite was true. Mastery beats variety, and you can't master nine things you barely open. That gap between what AI tools promise and what they actually deliver is the whole subject of the honest truth about AI productivity tools, and survey work like the Stanford HAI AI Index keeps showing the same thing: adoption is wide, but real value concentrates in a few tools used deeply. It's a smaller version of why I eventually pared my whole approach down to the five AI habits that actually changed my year.
Cutting once is easy. Staying cut is the hard part, because the bloat comes back the same way it came the first time — one reasonable yes at a time.
So I made one rule: a new AI tool has to replace an existing one to get in. One in, one out. No net growth. If I want to try something new, I have to name the thing it's beating and cancel that thing first.
This forces an honest comparison every single time. The new tool isn't competing against "nothing," it's competing against what I already pay for. Most contenders lose that fight, which is exactly the point.
I also put a recurring reminder on my calendar: once a quarter, re-ask the audit question of every tool. Ten minutes, four times a year, and the bloat never builds up again.
The money was the obvious cost, but it turned out to be the smaller one. The bigger cost of subscription bloat was cognitive, and I didn't see it until it was gone.
Every tool you own is a tiny open loop in your head. A thing you're supposed to be using. A capability you're supposedly leveraging. Nine of them meant nine little nags of low-grade guilt: am I getting my money's worth out of this one? Should I be using that one more? None of it loud. All of it draining.
When I cut to three, the noise stopped. I wasn't managing a portfolio of half-used software anymore. I had a small, sharp kit I actually knew, and the mental space that opened up was bigger than the money I saved. I'd been paying not just in dollars but in the quiet background hum of "I should optimize my stack."
There's a productivity myth that more tools equal more power, and it's exactly backwards for most people. Every tool you add is something to learn, maintain, remember, and choose between. Past a small number, the overhead of managing your tools eats the value they provide. You spend your energy deciding which AI assistant to use instead of doing the actual work.
The teams and people I've seen get the most out of AI almost always run lean. One strong general assistant they know cold, a specialist or two for the work they live in, and a ruthless refusal to add more. They're not behind for having fewer tools. They're ahead, because depth beats breadth and because a clear head beats a crowded one. The richest-looking stack is often the least productive.
If you've never added it all up, it's worth running the one-question audit on your own stack this week and seeing how much of what you pay for is redundancy wearing a different logo.
Q: Isn't it risky to depend on one main AI assistant? A little. But depending on nine isn't safer, it's just more expensive and more confusing. I keep one strong general tool plus one or two specialists. That's resilient enough without being bloated.
Q: How do you decide what counts as a "specialist"? It has to be clearly best at a specific task you do often. "Slightly better at writing" is not a specialist. "The only thing that does this exact job well" is.
Q: What if I cancel something and then need it? Then resubscribe. That's the beauty of subscriptions cutting both ways. The cost of cancelling and re-adding is near zero. The cost of paying forever for something you don't use is real.
Q: Did you lose any actual capability? No. Everything the cancelled tools did, my remaining ones do, or I never actually used. That was the whole revelation — the bloat wasn't capability, it was redundancy.
I wasn't paying for nine capabilities. I was paying for three capabilities, six times over, because subscribing is frictionless and quitting isn't. One strong AI assistant, a specialist or two, and a one-in-one-out rule covers almost everyone.
You don't need more AI tools. You need to actually use the ones you have.
Go add up what you pay for AI right now. If the number surprises you, that surprise is the bloat, and it's been quietly billing you for months.
I bought the dream of money while you sleep and chased it for years. Here's what passive income actually costs — and what I do now instead.
I tried every budgeting app and spreadsheet and quit them all within a month. Here's why budgets fail and the boring system that finally stu…
I spent years saving the hardest task for when I 'felt ready.' Doing it first instead quietly fixed my focus, my dread, and my output.
Comments
Sign in to join the conversation
No comments yet. Be the first to share your thoughts!