Age of Agentic commerce: The personalization paradox

Insights from our retail roundtable, part 3

At our recent roundtable with retail leaders, one question kept getting more complicated the longer we discussed it: how much should the system know?

As conversational tools become more embedded in retail, personalization gets harder to get right. If a customer asks for packing advice for a holiday, should the assistant remember that they bought luggage last year? If someone searches for running shoes, should their previous sizing data shape what comes up?

The answer, leaders agreed, depends entirely on context.

When personalization adds value

In high-frequency categories, personalization may not add much. Many purchases are habitual, intent immediate and contextual.

In apparel or specialty retail, though, personalization may be the whole point. Some retailers have spent years building fitting expertise across their store network, backed by loyalty data that ties the majority of purchases to individual customers. That kind of accumulated knowledge, how people actually shop and fit, is worth more than inventory alone.

The tension sharpens as AI becomes more central to the shopping experience. If generic AI tools start offering size recommendations based on pooled data, how does a brand hold onto what it knows? If the fit logic gets standardized across platforms, what's left to compete on?

The conversation moved from competitiveness into trust. Hyper-personalization can feel intuitive or intrusive, and the line between them shifts depending on context, category, and how well it's communicated. Retailers are learning that personalization requires calibration, applied deliberately rather than switched on by default.

Protecting proprietary advantage

Underneath all of this sits a harder question: what do you actually own?

AI tools now draw on enormous amounts of public and semi-public data. Digital sovereignty, once a concern mostly for governments and large enterprises, is becoming a live question for retailers too. So what stays yours?

Participants pointed to a handful of things they believe hold up: in-store expertise, category-specific training, loyalty-linked purchase history, operational data that never leaves the building, supply chain knowledge built over years.

In categories where fit, advice, or configuration drive the decision, that expertise can be a real competitive edge. But only if it's been turned into something a system can actually use.

The real risk is that retailers fail to bring their own knowledge into the AI-mediated environment they're now operating in. Staying distinctive requires becoming legible to the tools shaping how customers shop.

Platform alignment: coherence versus flexibility

As experimentation picks up, so does the complexity of the stack underneath it.

Some organizations are going deep with a single provider, pulling analytics, AI tooling, creative software, and cloud infrastructure under one roof. The case for it is coherence: fewer connections to manage, shared governance, data that flows cleanly between systems.

Others prefer picking the best tool for each job, assembling across communication, AI modeling, storage, and analytics.

The trade-off is becoming clearer. A mixed stack gives you flexibility but demands more orchestration. Going all-in with one provider smooths the friction but builds in dependency.

What matters most is making the choice deliberately. The cost of an incoherent architecture compounds over time.

Managing expectations in the age of LLMs

A less expected challenge came up over the course of the evening: perception. Customers arriving at a conversational interface bring assumptions from their experience with general-purpose AI. Some expect something close to human reasoning. Others brace for the rigid logic of an old-school chatbot.

When expectations run ahead of what the tool can do, disappointment follows. When they're low, even modest capability can land well. Closing that gap is now a product problem, not just a technical one.

This matters most where personalization is carrying real weight. A fit recommendation that's off, a suggestion that feels generic, a size it should already have known: in specialty retail, those moments don't just frustrate customers, they chip away at the expertise the brand has spent years building. Conversational AI earns its place by solving specific problems reliably. The skill is in choosing which ones to go after first.