The rough ground: How will agentic retail actually arrive?

Infrastructure’s the easy bit, but the rough ground is what comes after: the scrappy, human side of progress. False starts, crossed wires, the odd explosion. The part nobody demos on stage.

Delegation isn’t a software update; people don’t patch habits overnight.

Early adopters

The first movers live at the high-frequency, low-emotion end of retail—where convenience already beats choice and online ordering has flattened the experience.

Groceries, pharmacy, household essentials. Repetition, not reflection. Rich loyalty and transaction data already in place.

High-emotion categories—fashion, gifts, travel—will drag their heels. Those purchases are little mirrors of who we think we are. Handing them to an algorithm doesn’t feel like convenience; it feels like giving something up. Agents will still help by comparing options and collating reviews, but the final call stays human.

Trust will grow one transaction at a time but will break faster than it grows. A botched grocery order is recoverable whilst a mis-booked holiday is not. That tolerance gap will shape the early years. It’s not obvious where the first true delegation flows will surface. The logical endgame is the personal assistant that finally eats life’s drudgery—just not yet.

What's more likely is new players inserting themselves into household decision-making and optimising for cost, nutrition, allergies, schedule chaos, balanced per household. Retailers won’t surrender direct shopper access without a fight. Once delegated flows feel trusted, expect agentic behaviour to appear inside retailers’ apps—baked into the journey rather than layered on top.

Inside retail

Adoption inside retailers is less tech than politics. Every organisation has borders: CRM “owns” the customer, marketing the message, e-commerce the platform, and data the insight.

External delegation cuts straight through all of them; the moment an outside agent clicks "buy" on your behalf, the why shifts upstream. The data isn’t in the retailer’s vault anymore; it sits with the agent (and by extension, with you).

That’s not a systems issue, that’s a turf war.

For product teams, delegation means volume, whilst for CRM, it means opacity. It’s the aggregator problem reborn for the AI era—the same wedge search and marketplaces drove between brand and customer. Only this time it isn’t a website, it’s a trusted agent the user has a relationship with.

There is an upside we tend to miss; shoppers will share far more with an agent than they ever would with a retailer—diet, household context, values, whims. Agents will use that data to choose between near-identical options. That opens room to merchandise on micro-axes that rarely win alone: sustainability, local sourcing, brand ethics, even something as trivial as colour alignment or cultural sentiment.

Individually, none of these swing a purchase, but while two products look the same to a human, they may not look the same to an agent. When an agent’s choice quietly lines up with someone’s beliefs, a routine transaction turns into a small moment of positive engagement.

Retailers face a choice; lean in and make products and services easy for agents to find, understand, and buy. They can treat delegation like a new distribution channel and compete for algorithmic attention.

Or, get in the game—build (or partner with someone who can build) agents so customers can delegate inside the brand experience.

Most retailers will try some of both, and in the short term, that tension—between openness, control, and new kinds of influence—will be the roughest ground of all.

Context and constraint

Delegation will spread the way most ideas spread: unevenly, and sometimes the wrong way first. In places like the U.S., speed and novelty still outrun caution, so expect open pilots and the occasional spectacular mistake.

Across Europe and the UK, stronger privacy norms, and frameworks like the AI Act, the Data Act, and the DUAA, will slow rollout and deepen trust. In parts of Asia where mobile payments and conversational commerce are already routine, delegation may arrive quietly and just work.

It’ll start in the shallow end: groceries, small orders, low-stakes convenience, but once agents handle real money or complex intent, mistakes go public. Regulators follow quickly and law trails behaviour—until a headline writes the first draft.

Security is the other early brake. We’ve been here before: the first time e-commerce asked people to type card numbers into a browser, adoption lagged until trust caught up.

Delegation poses a harder question: not “Is this site safe?” but “Can I trust this agent to spend on my behalf?” Encryption and protocols will solve the mechanics quickly; perception will take longer as trust still spreads one safe transaction at a time.

The experimental phase

We’re in an extended beta: promise, friction, and unplanned lessons. Most “agentic” systems still behave like assistants rather than delegates, they suggest, build baskets, and ask for confirmation.

Less “Go do it," and more “Mind if I go ahead?”

New protocols that let agents take the next step are kicking off a fresh round of design work. How do you show intent? How do you explain reasoning? How do you set boundaries without burying people in settings?

This is where trust UX becomes its own craft—the etiquette of delegation. Winners won’t just have smarter models, they’ll have clearer manners: agents that feel safe before they try to feel clever.

The near horizon

Agentic retail won’t arrive overnight, but the groundwork is being laid already. For now, the sensible move is quiet prep: clean up product data, expose intent-level APIs, and consider how your brand looks through an agent’s eyes.

When delegation finally scales, the systems that already speak clearly—to people and to agents—will move first. Everyone else will still be in a workshop, drawing boxes, while the future slips through someone else’s checkout.

Reach out to our mobile experts to get ahead of the curve and explore what agentic could look like in your mobile product.