AI Tools I Stopped Using After 30 Days and Exactly Why

I canceled three AI subscriptions in the same week last month, and the reasons weren't what I expected when I signed up. None of them were bad tools. That's the weird part. They all did exactly what they promised. I just couldn't make them stick in my actual workflow.

The Writing Assistant That Made Me a Worse Writer

Jasper was first. I'd been paying $49/month for about six weeks when I noticed something uncomfortable. My drafts were starting to sound the same. Not the same as each other — the same as everything Jasper produces. That slightly enthusiastic, bullet-point-friendly, aggressively helpful tone that works great for landing pages but makes blog posts feel like they came off a conveyor belt.

The tool worked fast. I'll give it that. I could generate a 1,200-word article outline in maybe 90 seconds. But I was spending more time editing out the Jasper-ness than I would have spent just writing the thing myself. Every draft needed surgery to remove phrases like "let's dive in" and "here's the thing" and "Ultimately."

What I didn't expect: the tool was training me. After a month, I caught myself writing "simplify your workflow" in a draft I'd done completely on my own. That's when I knew I had to stop. The $49 wasn't the problem. The problem was I could feel my own voice getting flattened.

I switched to using Claude for brainstorming only. Never for drafting. Never for anything that might end up published. Huge difference.

The Meeting Transcription Tool That Created More Work

Otter.ai seemed like the obvious choice. Everyone recommends it. The transcription accuracy was genuinely impressive — maybe 94% on clear audio, which is better than I expected. I used it for about 40 client calls over a month.

Here's what nobody tells you: accurate transcription creates a new problem. You now have 47 minutes of text to read instead of 47 minutes of audio to skip through. The summaries were okay but missed the actual important stuff about 60% of the time. They'd highlight "discussed project timeline" when the real takeaway was that the client casually mentioned they might not renew.

I found myself watching recordings anyway to catch the tone of voice, the hesitations, the things that matter in client work. So I was doing double work — reading transcripts AND reviewing footage. The tool added a step instead of removing one.

What I do now: I just take handwritten notes during calls like it's 2019. Turns out the act of writing helps me remember, and I don't need a searchable database of every conversation. I needed to pay attention the first time.

The Kick: The Image Generator That Got Me in Actual Trouble

This is the one that taught me something I couldn't have learned from reviews. Midjourney is incredible at making images. It is terrible at making images you can legally use without stress.

I generated about 200 images over a month. Used maybe 15 of them in actual client work. Then I got a message from a client's legal team asking about the licensing status of a hero image I'd created for their campaign. I sent them Midjourney's terms of service. They sent back a four-paragraph explanation of why those terms weren't sufficient for their liability requirements.

The issue isn't that the images are illegal. The issue is that nobody can definitively say they're not derived from copyrighted training data, and some clients care about that a lot. I spent three hours that week finding a stock photo to replace something I could have found a stock photo for in the first place.

For personal projects? Midjourney is magic. For anything with a client's name attached? I stopped using it entirely. The conversation with legal wasn't worth the time it saved.

The Pattern I Didn't See Coming

All three tools failed for the same reason, and it took me a while to figure it out. They optimized for the wrong moment. Jasper optimized for the first draft when editing is where I spend real time. Otter optimized for transcription when understanding is the actual bottleneck. Midjourney optimized for creation when approval is what takes forever.

The AI tools I still use — Claude for thinking through problems, Perplexity for research, Descript for podcast editing — they target the actual slow part of my workflow. Not the part that feels slow but the part that actually eats my day when I measure it.

I think most AI tool hype focuses on speed of output because that's easy to demo. "Look, it wrote 1,000 words in 8 seconds!" But the 8 seconds was never my problem. My problem was always the 3 hours after that, and none of these tools touched it.

I've got two more subscriptions I'm probably going to cancel next month. Haven't decided yet. One of them might actually be useful — I just can't tell if I'm using it because it helps or because I feel like I should be using something.

Heads up: Some links in this post may be affiliate links. I only recommend tools I've personally tested. Opinions are entirely my own.