A feature on the line between helpful and creepy
Relevance, not surveillance.
AI can tailor what each person sees — recommendations, content, offers — to an audience of one, at scale. Done well, it feels like the brand understands you and saves you effort: the right thing surfaced at the right moment, less noise, more use. Done badly, it feels like being watched, and breaks trust in an instant. The entire craft of personalisation is staying on the helpful side of that line — being relevant without being invasive, useful without being unsettling. We personalise to serve the customer, never to surveil them, because the moment it tips into creepy, every gain is lost.
A personalised interface adapting to a user — relevant recommendations surfacing, with a clear, visible "your preferences, your control" panel beside them. Helpful relevance with the customer in charge. Square aspect ratio, dark studio, focused light, solar-yellow glow, a sense of usefulness and respect.
Helpful, not invasive — the right thing for the right person, with the line to surveillance never crossed.
Personalisation is the oldest promise in digital marketing and the one most often botched. For years it meant one of two failures: too crude to be useful — a clumsy "customers also bought" that suggested the thing you'd just purchased — or too invasive to be welcome, the ad that followed you around the web until it felt less like service and more like being stalked. The promise was always relevance; the delivery was usually either noise or creepiness. What changed is that AI can now make genuine, individual relevance possible at scale — which makes getting the line right matter more than ever.
Done well, personalisation is simply good service made scalable. A great shopkeeper remembers what you like, points you to the thing you'd actually want, and doesn't waste your time with the rest — and nobody finds that creepy, because it's helpful and you understand where the knowledge came from. AI lets a brand offer that same considerate relevance to millions of people at once: the right recommendation, the content that fits, the offer that's genuinely apt. When it works, the customer feels understood and saves effort, and the brand earns the engagement that relevance always rewards.
But the same capability sits one wrong step from unsettling, and that step is usually about knowledge the customer didn't expect you to have. Recommending a coat to someone browsing coats feels helpful; referencing something they never told you, or following them with an offer based on data they didn't realise you held, feels like surveillance. The line isn't really about how much you personalise — it's about whether the customer would be comfortable knowing how you did it. We design on the right side of that line deliberately: using data the customer would expect you to use, being transparent, and treating the test "would this delight them or unnerve them?" as the one that matters.
In this feature
Six things serious AI personalisation gets right.
Relevance that helps
The goal is to save the customer effort, not just to lift a metric. We personalise to surface what genuinely fits and cut the noise — because relevance that serves the person earns the engagement, while relevance that only serves you erodes it.
The creepiness line
The test isn't how much you personalise — it's whether the customer would be comfortable knowing how. We stay on the helpful side of that line deliberately, because the instant it tips into unsettling, every gain is wiped out by lost trust.
Recommendation done right
A good recommendation feels like a tip from someone who knows your taste, not a desperate upsell. We build recommendation that reflects genuine relevance — surfacing what the customer would actually want, not just what you most want to sell.
Content & offers that fit
Beyond products, the experience itself can adapt — the content shown, the message, the offer — to match where each person is. We tailor those dynamically, so the brand meets each customer with what's relevant rather than a one-size-fits-none default.
Privacy & consent
Personalisation lives or dies on trust, so we build it on data used transparently and with consent. Respecting privacy isn't a constraint on relevance — it's the foundation that lets a customer welcome it instead of recoiling from it.
Measured by their benefit
We judge personalisation by whether it genuinely helped the customer, not only by short-term lift. Relevance that serves people compounds into loyalty; relevance that merely extracts from them is borrowed against trust you'll have to repay.
The right thing,
for the right person.
The mental model that keeps personalisation honest is the good shopkeeper. A great independent shop owner remembers your taste, steers you to the thing you'll love, sets aside what they know you'll want, and never wastes your time — and you experience all of it as care, not surveillance, because it is plainly in your interest and you understand where the knowledge came from. The aim of AI personalisation is to scale exactly that feeling, not to scale the opposite one: the sense of being tracked, profiled and pursued by a faceless system that knows things about you it was never invited to know.
Where AI changes the game is the combination of scale and subtlety. Old personalisation worked in crude buckets — broad segments, simple rules — because that was all that could be managed by hand. AI can recognise the genuine pattern in an individual's behaviour and respond to it specifically, across recommendations, content and offers, for millions of people at once. That is what finally makes "an audience of one" more than a slogan: not a customer sorted into a demographic box, but an experience that adapts to what this particular person actually seems to want — which is both the opportunity and, handled carelessly, the hazard.
The whole discipline turns on a single test we apply to every personalisation decision: would this delight the customer or unnerve them? Using what someone is browsing right now to show them better options would delight; surfacing something based on data they didn't know you had, or following them with uncanny precision, would unnerve. The line is not about how clever the personalisation is — it is about whether the customer would feel served or surveilled if they understood exactly how it worked. So we build on data the customer would expect us to use, we are transparent, and we give people real control. Relevance built on trust compounds; relevance built on stealth is borrowed against a trust you will eventually have to repay, with interest.
The serious version of personalisation is measured by the customer's benefit, not just the short-term lift. It is easy to build personalisation that squeezes a little more from people in the moment — an aggressively targeted offer, a manipulative nudge — and watch a metric tick up while trust quietly drains. We judge the work by whether it genuinely helped: did it save the customer effort, surface something they were glad to see, make the experience better? Relevance that serves people earns loyalty and repeat engagement, which dwarfs any one-time gain. Personalisation done for the customer is an asset; personalisation done to them is a slow liability wearing the costume of a win.
A storefront experience adapting to a shopper — relevant suggestions and fitting content surfacing, with a visible, friendly preferences control. Helpful relevance with the customer in charge. Wide cinematic 21:9 crop, dark elegant studio backdrop, solar-yellow accent, a sense of usefulness and respect.
The good shopkeeper, at scale — relevance that helps, built on data the customer expects and controls.
Five questions we ask before personalising anything.
The line isn't how much you personalise — it's whether the customer would be comfortable knowing how. Would this delight them or unnerve them? That question decides everything.
AI personalisation reaches across the estate. It works hand in hand with Conversion Engineering, where relevance is one of the most powerful levers on conversion, and it overlaps with CRO's discipline of testing what genuinely works. Within the AI pillar, it draws on content generation to produce the variants a tailored experience needs, can be powered by a custom recommendation model where the data justifies one, and shares conversational AI's core principle: technology in service of the customer, with trust as the non-negotiable foundation. Relevance is the reward; trust is the price of entry.
We personalise the way a good shopkeeper would — using what the customer would expect us to use, surfacing what genuinely helps, keeping people in control, and judging the work by whether it served them. The brief is relevance that earns trust, not relevance that spends it. That combination of genuine usefulness, a firm creepiness line and privacy treated as foundational is exactly why personalisation, done properly, deepens a relationship rather than straining it: it is the right thing, for the right person, offered the right way.
A storefront showing relevant, helpful recommendations beside a clear customer preferences panel — useful, transparent, in the shopper's control. The moment relevance stays on the right side of the line. Contemporary, shallow depth of field, dark desk, solar-yellow screen glow.
Relevant and welcome — more useful, more conversion, and not a whiff of creepy.
Representative scenario · not a named client engagement
A retailer wanted personalisation but feared the creep factor — until relevance built on trust lifted results without it.
The retailer knew personalisation could lift results but had held back, wary of the creepy reputation it carried — they had no wish to become the brand that unnerved its own customers. Their instinct was right: aggressive, opaque personalisation built on data customers don't expect you to have is exactly how trust gets spent. The challenge was to capture the genuine benefit of relevance — less noise, better suggestions, a more useful store — without ever tipping into the surveillance feeling they, and their customers, disliked.
We built relevance the considerate way. Recommendations drew only on the customer's own browsing and purchase history — data they would naturally expect a shop to use — and surfaced it helpfully rather than pushily. Content and offers adapted to what each shopper had actually shown interest in, never to data they hadn't knowingly provided. And we gave people clear, easy control over their preferences, so the personalisation was transparent rather than mysterious. Every decision passed the test: would this delight the customer, or unnerve them?
Engagement and conversion rose — and complaints didn't, because customers found the relevance genuinely useful.
The result validated the cautious approach completely. Engagement and conversion both rose as customers found more of what they wanted with less effort — but just as importantly, the feared backlash never came, because nothing the brand did felt invasive. Customers experienced the personalisation as good service, the way they would a thoughtful shopkeeper, and trust in the brand strengthened rather than strained. The relevance helped precisely because it never strayed past what people were comfortable with the brand knowing — which is the entire difference between personalisation that compounds loyalty and personalisation that quietly corrodes it.
We wanted the benefit of personalisation without becoming the creepy brand that follows you around. Revolutionize built relevance our customers actually find useful — using only what they'd expect, with full control. Engagement went up, complaints didn't, and it genuinely feels like good service rather than surveillance.
What serious AI personalisation actually involves.
Personalisation is an investment in relevance that deepens the customer relationship, and it is scoped to the experience it touches. Engagements range from focused recommendation or dynamic-content work on one part of a site to a personalisation programme spanning the whole experience, and the scope rises with the breadth of what adapts, the data involved, and the rigour of the privacy and control built in.
The two largest variables are the breadth of the experience that adapts and the state of the underlying data. Personalising recommendations on one page is contained; tailoring content, offers and journeys across a whole experience, with the data and consent infrastructure that demands, is larger. We always build the privacy, transparency and control in as part of the work — never as a bolt-on — because they are what make the relevance welcome rather than alarming.
Every engagement includes the full discipline: defining genuinely helpful relevance, building on data the customer expects and consents to, designing transparent controls, applying the delight-or-unnerve test throughout, and measuring by the customer's benefit — so the personalisation earns trust rather than spending it.
We scope every programme to the experience rather than to a price list — which is why we don't publish rate cards. Every engagement begins with a free 30-minute scoping conversation, and we will tell you honestly where personalisation will genuinely help your customers and where it risks unsettling them. We would rather personalise less and keep their trust than chase a lift that quietly costs you the relationship.
When you're ready
Be relevant. Not creepy.
Tell us where you'd like your experience to feel more relevant to each customer — and where you worry it might tip into creepy. We'll respond within 24 hours with an honest read on how to capture the benefit of personalisation while staying firmly on the helpful side of the line.
Begin the conversation →