The AI tools I actually use.
A lot of shops are bolting AI onto things that don't need it. Here's what I reach for, and more importantly, what I don't.
Everyone's adding AI to everything right now. Chatbots on the homepage, AI product descriptions, AI search, popups that guess what you want before you've clicked anything. Sometimes it makes sense. Often it's a feature nobody asked for, added so the shop can say it has AI. I have strong opinions about which is which, and almost all of them come down to one question: is the AI solving a real problem, or decorating one?
What I actually reach for.
A few uses show up again and again as genuinely worth the money. AI-written product descriptions with a human review pass. Site search that understands what someone means instead of matching exact keywords. Automated tagging and categorisation for a big content or product library. None of these are exciting. They're boring, back-office jobs. That's exactly why they work: they save real hours every week and they move numbers you can actually measure, like search-to-cart rate or how long it takes to publish a new collection.
Product descriptions are the most underrated one. If you run a webshop with 300 SKUs or more, writing decent copy for each product by hand is genuinely miserable work, and it's usually the job that gets skipped. A well-prompted model, fed your existing tone and the product specs, gets you maybe 80% of the way there in a fraction of the time. A human then does the last 20%: fixing the claims that aren't true, cutting the fluff, adding the one detail the spec sheet left out. That last pass is not optional. The model will confidently write that a jacket is waterproof when it's only water-resistant, and that's the kind of mistake that ends up in a complaint or a return. So it's a draft engine, not a publish button.
Smart search is the other one I'll happily build. On a plain keyword search, someone typing 'warm jacket for winter' gets nothing if your product is called a 'thermal parka'. Search that understands intent connects the two. For a shop with a wide catalogue, that's the difference between a sale and a bounce, and unlike a chatbot it works silently in the background. Nobody has to know it's there.
The best AI feature is one your customers use without noticing it's AI.
What I push back on.
Chatbots on webshops almost never work as well as owners hope. They need constant feeding and maintenance, they frustrate people who just want a size chart or a delivery time, and they get expensive to run once you have real traffic. Most of the questions a shop gets are the same ten questions. A well-organised FAQ, a clear returns page and honest navigation answer those faster than a bot that makes someone type a full sentence and then wait. I've watched people rage-click a chat bubble closed. That's your support tool actively costing you goodwill.
Same story with fully AI-generated blog posts and landing copy. Search engines are getting better at spotting the pattern, and readers can feel it even when they can't name it. It reads smooth and says nothing. If your brand voice matters, and for most shops with any personality it really does, AI-generated content needs a heavy human edit before it stops sounding like every other page on the internet. By then you've often spent more time fixing it than you'd have spent writing it yourself.
How I decide what's worth it.
When a client asks whether to add some AI feature, I run it through a short checklist before I quote anything.
- Is there a real, named problem? 'Descriptions take my team two days a week' is a problem. 'We want to look innovative' is not.
- Can I measure whether it worked? If there's no number that should move, there's no way to know if it's earning its keep.
- What does it cost per month at real volume, not demo volume? API calls add up fast once you have traffic.
- What happens when it's wrong? A wrong product tag is cheap to fix. A chatbot giving a wrong returns policy is not.
- Would a simpler, non-AI fix do the same job? Often a better FAQ or a fixed list of search synonyms beats a model.
If a feature can't get past those five, I say so, even when the client was hoping I'd just build it. Talking someone out of spending money is a strange way to run a studio, but it's exactly why the people I work with come back.
The honest takeaway.
AI is a tool, not a strategy. It earns its place when someone has thought carefully about the specific job it's doing and what it costs when it slips. Used that way, on the boring jobs, it quietly saves you time and money. Bolted on to look modern, it just adds one more thing that can break. If the brief is 'make us look innovative', that's not a problem any model can fix, and I'll tell you that before you've spent a euro.