August 2, 2026 · by David Gilbert · 3 min read · Tech & AI
I get asked a lot about the AI tools I use and recommend, far less often about the ones I tried, liked initially, and then quietly stopped opening altogether. I think the second category is actually more useful to talk about honestly, so here's one of mine, and what it taught me about evaluating these tools properly rather than just chasing the next interesting feature.
What It Promised
It was a tool built around automating a specific chunk of client communication — drafting follow-ups, summarising project status, flagging things that needed my attention. On paper, exactly the kind of repetitive admin task I'm always looking to hand off. I was genuinely enthusiastic for the first month or so, recommending it to a couple of clients, integrating it properly into my actual workflow rather than just testing it superficially.
What Actually Happened
The drafts it produced were fine, technically — grammatically correct, reasonably well-structured, nothing actually wrong with them. But I found myself rewriting a meaningful chunk of every single one anyway, because the tone consistently missed the specific relationship I had with whichever client it was drafting for. Not wrong exactly. Generic in a way that, once I noticed it, I couldn't stop noticing.
The Moment I Actually Noticed the Problem
I caught myself spending almost as long editing a draft into something that actually sounded like me as I would have spent writing the thing from scratch with a faster, simpler tool I already trusted. That's the exact moment a "time-saving" tool quietly becomes a net time cost, and it's easy to miss because each individual edit feels small and reasonable in isolation, right up until you actually add up the total time across a week.
Why I Didn't Notice Sooner
Partly sunk cost — I'd invested time setting it up properly and didn't want to admit that investment hadn't paid off. Partly the general hype around AI tools generally, which made it feel like abandoning one was somehow a step backwards rather than just an honest evaluation. It took a deliberate, slightly uncomfortable moment of actually tracking my own time for a week to notice what should have been obvious much sooner.
What I Actually Do Now Instead
I went back to a simpler, more general-purpose AI assistant for first drafts, and I personally shape every single one before it goes anywhere near a client, rather than expecting a specialised tool to nail the specific relationship-aware tone automatically. Less impressive-sounding as a setup. Genuinely faster and more reliable in actual daily use, which is the only metric that was ever supposed to matter.
The Broader Lesson
I think a lot of people, myself very much included, adopt new tools based on what they promise rather than periodically, honestly checking what they're actually delivering once the initial enthusiasm wears off. It's worth doing that check every so often, deliberately, on every tool you've adopted — not because new tools are bad, but because the gap between "exciting at first" and "still genuinely earning its place three months later" is real, and it's surprisingly easy to not notice you've drifted from one into the other.
No Hard Feelings Toward the Tool Itself
It's a perfectly competent product, and I'd guess it genuinely works well for plenty of people in different situations to mine. This isn't really a story about a bad tool. It's a story about a mismatch between what I actually needed and what it was built to do — and a reminder to keep checking, honestly, that the mismatch hasn't crept back in somewhere else I haven't noticed yet.