The AI hype is annoying — especially when basic features are missing

The AI hype is annoying. Especially when basic features are missing. I see it every day, in almost every tool I use professionally: "Now with AI!" — and at the same time my export doesn't work, the API stutters, or a broken filter logic costs me ten minutes that I've lost again every Monday for the last three years. That's not just irritating, it's disrespectful to the people who are supposed to work with the software.

An example baked into every CRM

Over the last few months I've evaluated three CRM systems for different purposes. Every single one prominently advertises AI-driven text generation. You can have it draft emails, generate summaries, analyse sentiment. Quite nice. I wouldn't miss it if it disappeared, but there it is.

What none of these three systems can do properly: a classic data import. Drop a CSV in, do the mapping, done. Honestly the first thing a CRM should manage. In one tool I had to deliver column names in a specific order. In another there was no preview. In the third, accented characters were mangled and the support reply was: "That's on the roadmap." For months.

On the roadmap there were also fourteen AI features I never asked for.

The underlying problem: wrong priorities

I have great respect for AI. I build products myself in which AI plays a sensible role. But right now I see a lot of teams building AI features because investors like them and product demos with AI look pretty. The actual product — the thing customers originally bought the software for — doesn't get better in the process, it gets worse. Because every developer hour that flows into "write me an email" buttons doesn't flow into the CSV import.

The result is a fairly absurd state of affairs. Tools that can answer questions you rarely ask. Tools that, at the same time, fail at the most trivial function you need every day.

I know the commercial logic behind it. AI is good for pricing. You can sell AI features as a new tier. You can show them in keynotes. A reliable import is hard to market. It's just what an adult product simply has to have.

What I've learnt

I now evaluate software differently. I look at the basics first. Exports, imports, backup, search, API. If the foundation is clean, everything else is a bonus. If the foundation wobbles, the prettiest AI button is just a poster nailed to a crumbling facade.

The result is that I increasingly choose fewer big, loud vendors and more small, quieter products. There's often more genuine product work in those than in the heavily-lit startups currently flooding the timeline. Not always, but often enough that it's a pattern.

What I'd advise software vendors

If I had product responsibility today, I'd run a deeply unfashionable quarter. No new AI feature. Nothing. Instead I'd pull twenty tickets out of the backlog that have been bothering customers for years. The missing filter field in the list view. The import that's been bad since V1. The error message nobody understands. The search that can't find accented characters.

I'd run that for three months and at the end publish a changelog that honestly says: "We didn't produce any new headlines this quarter. Instead we lifted the product you actually use." I guarantee the NPS scores for that quarter would beat all the AI quarters that came before.

My personal takeaway

AI is a serious technology and it will change software for good. That's exactly why this hype annoys me so much. It hides what AI can really do, because the label gets slapped on wherever no better marketing fits. Whoever builds a solid product today and uses AI where it brings real benefit, wins. Whoever lets themselves be carried by the hype and ignores basic features, loses in two years. The market does the maths in the end, and it's pretty sober about it.

I'm staying patient. But I'm not blind.