A lawnmower, a lie, and the death of user understanding

Last week, our elderly neighbours' lawnmower gave up the ghost. I took a look but It was beyond repair. So I set out to find them a replacement: lightweight, easy to start, simple to handle — something suitable for someone who might struggle to lift anything too heavy. And, crucially, available for collection nearby within a few days.

What followed was a dispiritingly familiar dance through a series of interchangable retailer websites. Clicking through endless faceted search filters. None offered ‘lightweight’ so I guessed ‘Battery-powered’ would be a proxy. I dug through specification tables to compare weights. I manually checked local stock for each model.

Not one site asked: Who's using this? What matters to them? What's their real need?

In frustration, I described the same problem to ChatGPT’s agent mode: "An older user needs a lightweight lawnmower. Easy to start, low physical effort, available for collection near SG13 this week."

It responded instantly: understood the context, asked a clarifying question about lawn size, then went about searching, and returned three tailored recommendations — with weights, run times, local availability, and reasoning. All in under 5 minutes.

The contrast wasn't just stark. As someone who’s spent my career in digital trying to improve CX, it was humiliating. An AI with zero memory of my past behaviour — no cookies, no profile, no history — grasped the human reality better than retailers who've served customers like my neighbours for decades.

Digital was quick to morph usability into UX, with Product teams carefully allocating story points to user research and design testing. But we’ve been lying to ourselves: we’ve been terrible at actually understanding customers’ real needs.

We told ourselves we were user-centric. That we listened. That our data gave us insight. But when push came to shove — when a real human need arose — our systems failed to respond like anything remotely human.

The lie wasn't malicious. It was comfortable. It let us fall back on pattern libraries as a shortcut for usability. It allowed us to build interfaces that served databases, not people — and still call it "UX."

Now, AI agents are exposing that lie — not by being magical, but by doing the one thing we stopped doing: Asking the right questions. Actually listening.

The Fermanagh telephone exchange

My friend Cormac grew up in a small village in County Fermanagh, Northern Ireland that still relied on a manual telephone exchange. When we were at university, if he wanted to speak to his parents, he'd call the operator and ask to be put through to them.

But often, the operator already knew they weren't home: "Ah no Cormac, they're over at the shop. I'll put you through there instead."

She didn't see her role as just connecting calls: she understood context. She knew the rhythms of the village, the patterns of people's lives. She could anticipate needs before they were articulated.

That operator did something remarkable: she took intent ("I want to speak to my parents") and overlaid it with context ("but they're not where you think they are"). She she improved their own requests.

Then direct dialling arrived. More efficient. No human bottleneck. You could call anyone, anywhere, instantly.

But something was lost. The intelligence. The local knowledge. The ability to say, "I know what you asked for, but here's what you actually need."

We've spent decades building digital systems that are the equivalent of direct dialling: efficient, scalable, but fundamentally dumb.

We've replaced that knowledgeable operator with search bars, checkboxes and dropdown menus, and called it progress.

Now, AI agents are doing what that village operator once did:

• They're understanding intent
• They're adding context
• They're anticipating needs
• They're saying "I know what you asked for, but here's what you actually need"

The irony is that this was the promise we hired technology for all along. We just forgot the operator's lesson: intelligence isn't just processing requests—it's improving them.

From empathy to efficiency: How UX forgot its soul

Somewhere between the dot-com boom and the mobile revolution, user experience stopped being about users and started being about interfaces. We stopped asking, "What does this person actually need?" and settled for a CRO-dominated world of marginal gains, obsessing over button colours, font weights, and conversion funnels.

User research got pushed into sprints, then sidelined. Personas became PowerPoint props. "User-centred design" became a checkbox. Jakob Nielsen has been saying for some time that UX has taken a wrong turn, and has become “trapped in complacency”. And he’s right.

Take three contrasting customer missions to buy coats at a clothing brand:

• a parent buying a winter coat
• a cyclist needing all-weather protection
• someone replacing a decades-old favourite

All are funnelled into the same sterile faceted search. The system knows nothing about why they're here. It doesn't care.

AI > UI: raising the bar

Agentic search and the growth of AI is exciting for sure, but really it’s highlighting the shortcomings of our existing technologies - and in particular the generic eCommerce UI.

When you ask an agent for help, it doesn't dump a search results page on you. It asks: "What are you using this for?" "Any constraints?" "Should I prioritise availability or price?"

It treats you like a person. Of all things.

And crucially — it has access to your life data: your context, your history, your preferences. Not just what you bought, but why you bought it. Not just your location, but your routine. Not just your job title, but your daily reality. Even if you’ve not given it to an agent, through a few engagingly written prompts it can gather insights that most brands would kill for. Yet most aren’t even trying to develop this type of understanding.

These agents will soon set a new standard for customer interactions: understand me, evaluate my needs, genuinely help me choose. Customer expectations are going to increase quickly.

To bring this to life, let’s evaluate the experience of using a major European trade supplier’s website v using an agent to use the website.

I can ask my agent: "I'm an electrician. I need a 10-way, IP2XC-rated consumer unit. Available for collection tomorrow within 10 miles of AL1.". It will use the language of my query to gather insights, then go out and get results then return a shortlist: models, specs, pricing, stock levels, click-and-collect options and recommendations based on my specific needs. Done.

Now let’s compare it with a visit to the supplier's website. Search "consumer units." Scroll. Filter by IP rating (buried in a 15-item faceted search bar). Check 10, 11, 12, 13, 14-way boxes. Click each product to check local stock. Repeat. Divine most appropriate/shortlist and compare.

The agent-led experience is listening to a human with a job to do. The UI is asking the user to interact with a database. The agent understands that the user is working - that this isn’t fun. They don’t want a discovery experience - they want task resolution. The website meanwhile acts like the user is engaged in an exciting shopping experience. It's not just inefficient — it's tone-deaf.

The reckoning: brands are being out-listened

This is nothing new. For years, brands have told themselves: "We know our customers." "We have data." "We personalise."

But what did that personalisation look like? Remembering you prefer women's clothes over men’s or kids?? Showing you the same jacket you viewed yesterday? That's not personalisation. That's recognition. And soon it won’t be anywhere near enough.

Now, AI agents are doing what brands should have done:

• Building dynamic profiles
• Inferring intent
• Acting on context
• Asking questions

Users are going to get used to being understood. And are going to reject brands that fail to do so. Google results that once seemed adequate now feel primitive. Generic category pages look like relics. And brands that once felt "convenient" now seem indifferent. People will get used to good service quickly, and won’t settle for less.

Know your game — or lose your customer

Maybe the future of commerce isn't one-size-fits-all, but rather some distinct paths. And every brand needs to choose.

1. Functional commerce: the efficiency game

You're selling solutions, not experiences. Tools. Supplies. Utilities. Your customer isn't here to browse. They're here to solve a problem and leave.

If this is you, accept it. Stop pretending your UX is "engaging." It's not. It's a chore.

Your job now: make it invisible.

Build rich, structured data. Open APIs. Clear specs. Real-time stock and pricing. Make it effortless for AI agents to find, compare, and buy on behalf of your customers. Because they will. And if you're not machine-readable, you'll be invisible.

2. Experiential commerce: the relationship game

You're selling choice, identity, delight. Fashion. Luxury. Lifestyle. Here, the process is part of the product. If this is you, double down on real understanding. Not fake personalisation. Not retargeting. Empathy.

Your product teams should be laser focused on:

1. Why does this customer keep coming back? What need are we serving for them?
2. What do they value? Speed, exclusivity, sustainability?
3. How do they want to feel when they buy?
4. How do we identify our future brand evangelists?

Because if you can't answer that, an AI agent will — and it'll offer a better experience.

The battleground?

The middle ground — where brands pretend they're experiential but deliver commodity UX.

And maybe this is where the reckoning will hit the hardest: where there is the biggest mismatch between user intent and brand response.

The comeback starts with a question

The solution isn't to slap a chatbot on your homepage.

It's not to "go AI-first" or hire a prompt engineer.

It's to remember how to listen.

Start with "what does this customer actually need, right now?"

Then design for that.

If you're in functional commerce: optimise for speed, clarity, and machine readability. Help agents help your customers.

If you're in experiential commerce: Invest in deep user research. Build adaptive, contextual experiences. Make the interaction worth staying on your site.

And if you're in the messy middle? Pick a side. Because users won't do it for you.

The future isn't AI vs. humans. It's listening vs. indifference.

I’m excited about this shift: it’s a great time to remind ourselves of the rules, the techniques and user-centred disciplines we’ve forgotten.

The brands that survive won't be the ones with the flashiest tech. They'll be the ones that start understanding their customers best: not as data points, but as people.

The reckoning isn't coming. It's already here. AI is reshaping expectations now. So the question is whether your brand will remember how to listen before your customer walks away.