𝓤𝓷𝓲𝓽𝓮𝓭 𝓝𝓮𝔀𝓼

Conviva is a Business Reporter client

For three decades, every visitor saw the same website. Now AI is taking personalisation to the next level.

Pretty soon, you’ll click an ad on Instagram and something magical will happen. Somewhere between you pressing “Shop Now” and a page appearing on your screen, a version of that site will be assembled just for you. Your shopping history, sizes and behaviours all factored in before a single pixel loads. You will never know what someone else sees when they visit the same page.

This is not the promise of some distant AI future. It is, according to people who build the infrastructure behind the world’s biggest consumer apps, already underway. And the implications for shoppers are significant.

For most of the internet’s history, personalisation has been a clumsy recommendation engine. Amazon made it easier to repurchase fabric softener when you bought laundry detergent. Netflix recommended Mission: Impossible III after you watched Mission: Impossible II. But underneath it all, the fundamental architecture of a website or app was the same for everyone: a fixed set of steps to take consumers down a predetermined path.

That era, say people working at the frontier of consumer technology, is coming to an end. And in its place are AI agents making thousands of micro-decisions, in real time, for every individual user, at every moment of their experience.

The scale at which this is already operating is difficult to comprehend. Conviva, whose intelligence platform sits across some of the world’s largest consumer apps and websites – among them the NFL, QVC and Fox – analyses more than five trillion events per day. What that data reveals about how people actually behave online is, to put it plainly, at odds with almost everything the industry has assumed for 30 years.

The standard model of online behaviour has always been the funnel: a tidy sequence of steps a user takes from arriving at a website to completing a purchase. Enter the website. Browse. Click on a product page. Add to cart. Checkout. It is elegant, measurable and, according to Conviva’s data, largely fictional. 67 per cent of users do not follow a linear path. They loop back, hesitate, compare, abandon and return hours or days later with different intentions entirely.

The funnel, in other words, compresses a rich and deeply personal decision-making process into five identical checkmarks. In doing so, it hides everything that would allow for true personalisation.

“What we’re seeing,” said Keith Zubchevich, president and CEO of Conviva, “is that consumer behaviour is far more individual than the industry has ever treated it. People don’t move through experiences in the same way. They have different windows of decision-making, different triggers, different points of friction. The question is whether technology can finally catch up to that reality – and the answer is that it can.”

The practical consequences of this shift are, once you understand them, everywhere. Consider the apparently simple problem of a shopper who adds a coat to their basket and then disappears. The conventional response – a reminder email after 24 hours – assumes the person simply forgot. But Conviva’s pattern data shows that for a particular type of shopper, the hesitation had nothing to do with forgetting. The delivery cost, revealed only at checkout, had triggered a period of comparison shopping that typically resolves itself within four hours. An email sent at the 24-hour mark, by which point the person has either bought elsewhere or moved on, is not a reminder. It is a postcard from a missed opportunity.

This is where AI agents enter. Rather than applying the same rules to everyone, agents can now be trained on population-scale behavioural data to identify which intervention, at which moment, is most likely to be useful for a specific individual.

The shipping cost that will cause one customer to abandon an order will not bother another. The stock notification that brings a particular user back to a site becomes irrelevant if sent to someone who was never planning to return. What agents make possible, in theory, is an internet that has finally learned to tell the difference.

There is, of course, a version of this story that feels less like progress and more like surveillance dressed in helpful clothing. The same granular behavioural data that allows an app to send a cart recovery email at precisely the right moment also allows it to identify – and potentially exploit – the moments when a consumer is most psychologically susceptible to a nudge. The line between personalisation and manipulation has always been thin. AI agents, operating at scale and speed no human team could match, do not necessarily make it thicker.

The industry’s answer to this concern is that agents which genuinely improve an experience generate better outcomes for everyone. An app that fixed the broken payment gateway you encountered, before you had to encounter it three times in a row, is not manipulating you. It is simply working properly, for once.

Whether that distinction holds as the technology matures is, like most questions about AI, not yet answered. What is not in doubt is the direction of travel. The generic internet – the one built on the assumption that a single experience, applied consistently to millions of people, was good enough – is passing. What is arriving in its place is something truly personal.

For more information, visit conviva.ai.