Every pitch deck this year has a slide about AI. Half of them are wrong.
Key takeaways
- AI is genuinely useful in three places: rapid prototyping, behavioural analytics, and personalisation at scale.
- It’s mostly useless for the thing everyone talks about: replacing the actual design decisions.
- Clients don’t need an “AI strategy.” They need to know which three tasks got faster this year, and why that matters for their budget.
What’s actually happening
Layout generators can now spit out a working homepage in minutes. Analytics tools read scroll depth and rage-clicks and tell you where a page is losing people, without a human combing through session recordings for an afternoon. Chatbots handle first-line support questions well enough that a small team can look like a bigger one.
None of that is theoretical. We use versions of all three on live projects at FearNaught.
What’s theoretical is the idea that AI is now “doing the design.” It isn’t. It’s doing the parts of design that were always mechanical, freeing up time for the parts that were never mechanical to begin with: judgement, taste, knowing which client instinct to follow and which one to gently talk them out of.
Where it actually helps
Prototyping speed. A first-round concept that used to take two days of layout work now takes an afternoon. That doesn’t mean the final product ships faster; it means more rounds fit inside the same timeline, which usually means a better final product.
Behavioural analytics. Heatmaps and funnel data used to require someone to sit and interpret them. Now a tool flags “63% of mobile users abandon at the shipping step” without being asked. That’s a real time saving, and it catches problems a human reviewer might miss on page 40 of a report.
Content variants. Testing five headline options against a segment used to mean five separate campaigns. Now it’s one build with variants swapped automatically. Useful for e-commerce and lead-gen sites specifically; less useful for a five-page brochure site that isn’t running that kind of traffic anyway.
Where it doesn’t
Brand voice still comes from a person who’s sat in the room with the client and understood what they actually mean when they say “we want to feel premium but approachable.” AI-generated copy for that brief reads like every other AI-generated copy for that brief. You can tell. Your customers can tell.
Same with layout taste. A generator can produce a technically correct homepage. It can’t tell you that the hero section feels like every other hero section shipped in the last eighteen months, because it was trained on exactly those homepages. That judgement call is still a human one, and it’s the one that actually differentiates a brand.
There’s also a trust cost nobody talks about enough. Visitors are getting sharper at spotting generic, AI-flavoured design and copy, the same sameness that’s fuelling human-first web design. Lean too hard on automation and you end up looking like everyone else who leaned too hard on automation.
What to ask your agency
If an agency’s pitch is “we use AI,” that’s not a pitch, that’s a tool list. Ask what specifically got faster or better because of it, and what still needs a person. If they can’t answer both halves of that question, they haven’t actually thought it through, they’ve just added a slide.
Bottom line
AI earns its place in a web project the same way any tool does: by making a specific task measurably better. Prototyping, analytics, and personalisation clear that bar right now. Brand voice and design judgement don’t, and probably won’t for a while. Build your project plan around that split, not around the hype.