September 14, 2026 · by David Gilbert · 4 min read · Tech & AI
A genuinely large share of the software I work with for clients has, over the past while, quietly grown some kind of AI feature bolted onto a tool that worked completely fine without it, and that nobody I've actually spoken to specifically asked for. I've got a fairly clear, honest theory about why this keeps happening, and what it means for how clients should actually evaluate these features as they appear.
The Pattern I Keep Seeing Repeatedly
A familiar piece of software — accounting, scheduling, basic communication tools — gets an update, and suddenly there's a new AI-powered button or panel somewhere in the interface, doing something that's mildly interesting in a demo and genuinely unclear in actual day-to-day practical value. Sometimes it's genuinely useful once you dig in properly. Often, honestly, it isn't, at least not yet, for most actual users.
My Honest Theory on Why This Keeps Happening
Competitive pressure, more than genuine, validated user demand, seems to be the real driver behind most of this. If a competitor ships an AI feature, even a mediocre, half-baked one, there's real commercial pressure to ship something comparable quickly, regardless of whether actual users specifically asked for it or have a clearly defined need it's solving. The feature becomes a checkbox on a sales sheet before it's necessarily a genuinely, carefully validated improvement to the actual product.
Why This Matters for How You Should Evaluate New Features
"This software now has AI" tells you almost nothing useful on its own about whether that specific feature will genuinely help your specific business. Plenty of these additions are genuinely half-finished, evaluated and shipped under real competitive pressure rather than thoroughly validated against actual user need first. Evaluate each new feature individually and honestly, on its own clearly demonstrated merits, rather than being swayed by the mere presence of "AI" as a checkbox label.
How I Actually Evaluate These for Clients
I ask what specific problem the new feature is solving, whether my client actually has that specific problem in their own day-to-day work, and whether the feature solves it better than however they're currently handling it already. If the honest answer to any of those is genuinely unclear, I recommend ignoring the new feature entirely for now and reassessing it again in six months or so, once it's had time to mature properly.
The Features That Have Actually Proven Genuinely Useful So Far
The ones solving a clear, specific, pre-existing pain point — automated categorisation that's saved genuine time on a previously tedious manual task, drafting assistance for genuinely repetitive written content, summarisation of long documents that were previously eating real time to read manually. The pattern among the useful ones is consistent: a clear, specific, pre-existing problem, with the AI feature solving it better than the previous manual method actually did.
The Features That Have Mostly Been Forgettable Noise
Vague "AI insights" panels with no clear, specific action attached to them. Chatbots bolted onto tools that didn't genuinely need one and that nobody seems to actually use after the first curious try. Generative features producing content nobody asked for and nobody, in practice, actually uses afterwards.
The Honest Bottom Line for Anyone Choosing Software
Don't choose, or stick with, software purely because it has AI features, and don't dismiss software purely because it doesn't yet have any either. Evaluate the actual, specific features in front of you, on their own real merits, against your own specific, actual needs. The "AI" label is currently doing far more marketing work than it's reliably doing genuine, validated functional work, and that gap is worth staying clearly, deliberately aware of as you evaluate what's actually in front of you.