Jonathan Taylor and the Quiet War on Choice

A shopper looking for a drill bit, a dishwasher or an industrial valve is rarely suffering from a shortage of products. The shelves have multiplied; understanding has not. Every specification promises precision, every filter promises order, and soon the browser holds twelve tabs and a small private crisis. Jonathan Taylor has spent much of his working life inside that crisis. His subject is not shopping in the cheerful sense. It is the untidy moment when a person has plenty of information and still cannot decide.

Taylor, known as JT around Zoovu, comes to the problem as a software engineer. He studied software engineering and computer science at Durham University, then built a career around technical architecture, personalization and real-time interaction. His work passed through companies including IntraLinks, AGT International, SmartFocus and its later owner Actito. By the time he co-founded Zoovu in 2018, the pattern was clear: machines were getting better at responding quickly, but speed did not make the response useful.

The useful response requires context. A camera lens suitable for one buyer may be wrong for another. A component can match the stated dimensions and still fail a compatibility rule hidden in a manual. A washing machine can be technically excellent and physically too deep for the cupboard. The ordinary shop assistant handles these facts with questions. Online, the catalog usually hands the shopper another row of rectangles.

The intelligence below the conversation

Zoovu began from guided selling: software that asks a sequence of questions, understands the answers and narrows the field. Taylor led the product team as CTO and worked with both consumer brands and complicated business-to-business sellers. The latter became an unexpectedly useful education. An industrial catalog is a stern examiner. Products arrive with variants, dependencies, safety considerations, bundles and enough attributes to make a spreadsheet ask for a holiday.

Taylor has said this difficult beginning became an advantage. Long before conversational commerce acquired its fashionable name, Zoovu was trying to guide high-consideration purchases in technical categories. Those environments punish improvisation. If the underlying product data is incomplete or inconsistent, elegant language merely delivers the wrong answer with better manners.

This distinction separates Taylor's public thinking from the general excitement around chatbots. He does not treat conversation as the achievement. Conversation is the visible portion of a larger system: product information, compatibility logic, ontologies, policies and the commercial context that determines whether a suggestion is sensible. In Milan, after meetings with fashion customers and brands, he condensed the problem into four words: “Garbage in, garbage everywhere.” It is a good engineer's joke because it is only slightly a joke.

The phrase updates the old computing warning for a world in which one bad attribute can travel. An incorrect product fact no longer stays quietly on a product page. It can appear in search, recommendations, a chatbot, email and any agent permitted to reuse it. Fluency raises the cost of a weak foundation because errors now arrive with confidence.

A CTO carries the blueprint upstairs

On January 1, 2026, Taylor became Zoovu's chief executive. The move followed nearly eight years as co-founder and CTO, and more than two decades at the junction of business and technology. It also preserved the company's technical argument. Taylor did not arrive to replace the product thesis with a new slogan. The person who had been drawing the engine was now responsible for where the vehicle went.

The appointment came after Zoovu had expanded beyond guided selling into search, configuration and AI-assisted discovery. Its generative assistant, Zoe, was live on sites including Microsoft, Staples and Bosch. By the company's January announcement, Zoe had handled more than four million shopper interactions, and nearly half of Zoovu customers had either implemented or begun implementing conversational experiences.

Those numbers describe scale but not the daily obligation. An assistant must be fast, accurate, governed and sufficiently modest to stay within what the product knowledge supports. It must also produce a business result. Taylor's vocabulary repeatedly returns to outcomes: conversion, satisfaction, confidence, speed to value. The romance of artificial intelligence fades quickly when the wrong refrigerator arrives.

The stack and the engine

In May 2026, Zoovu acquired XGEN AI, a company focused on search, merchandising, recommendations and personalization. Taylor explained the deal with a mechanical metaphor: “The next generation of product discovery isn't a stack - it's an engine, with AI at its core rather than bolted on.”

A stack is a collection. Search sits here, recommendations there, a chat window somewhere else, and each tool develops its own partial memory of the customer. The acquisition expressed Taylor's preference for a shared model that learns across those encounters. A query in search can improve a recommendation. A guided-selling answer can inform personalization. One set of product facts can govern all of them.

The claim is attractive precisely because the alternative is familiar. Enterprise software estates often resemble old cities: charming from a distance, difficult to cross, and full of infrastructure nobody remembers approving. Unification promises fewer handoffs and a clearer record of why the system made a choice. Yet it also raises the standard. If one engine powers every experience, reliability is no longer a departmental concern. It is the company.

Taylor's version of that engine stretches across two rather different shopping moods. One customer is exploring a fashion collection and wants inspiration. Another is specifying industrial equipment and wants certainty. Taste and technical fit do not ask the same questions. The common layer is an ability to understand intent, represent the product faithfully and explain the path between them.

The right to recommend

There is an ethical idea tucked inside this commercial one. A recommendation asks for trust. The buyer cannot inspect every possible option, so the system is given permission to reduce the field. Taylor argues that brands should retain ownership of this experience: the product data, policies, voice, governance and measures of success. A specialist vendor may build the machinery, but a brand cannot outsource responsibility for the answer carrying its name.

That position has practical consequences. Start with a focused use case where the decision is genuinely difficult and the result can be measured. Ground each suggestion in approved information. Explain why a product was recommended. Make escalation to a person easy. Then watch what customers ask and improve the system. This is less like launching a campaign than maintaining a useful employee, one who works all night but still requires supervision.

Taylor's own career makes him an apt custodian of that approach. The software-engineering background supplies suspicion of magical surfaces. Years in personalization supply an understanding of context. Work with enterprise buyers supplies patience for integration, governance and all the unphotogenic details between prototype and production. His stated temperament is customer-first; his public prose is fond of foundations.

After the search box

Taylor's horizon now extends beyond the familiar chat pane. He has written about voice interaction, avatars, wearable interfaces and the protocols that let AI agents use validated product knowledge. The interface may change repeatedly. The underlying test remains stable: can the system help somebody choose with confidence?

That test is harder than producing an answer. It requires a machine to know what it knows, a company to keep its facts in order and a recommendation to survive contact with the real product. The work is closer to librarianship than fortune-telling. It is also, in Taylor's telling, where much of the commercial value resides.

Online retail spent its first decades putting everything on the shelf. Product discovery is the attempt to make that abundance navigable. Taylor has arrived at the CEO chair just as generative AI offers a new guide through the aisle. His wager is that the guide will earn its place not through personality alone, but through disciplined knowledge and an honest explanation.

There is something pleasantly unfashionable about the wager. The loudest technology often announces the end of old constraints. Taylor keeps pointing at them: the missing attribute, the incompatible part, the customer who still has one unanswered question. Choice becomes useful only when those details are respected. Intelligence, in commerce as elsewhere, is not the ability to say more. It is knowing what matters before speaking.