For decades, retailers and brands have optimised for one thing: the customer. Every decision - product, price, promotion, placement - has been about influencing human choice. That model is now changing. A new decision-maker has entered the shopping journey, the AI agent. And with it comes a simple but profound shift: Loyalty needs to be earned twice – once from the customer and once from the AI agent acting on their behalf. This is the new era of shopping, an era marked by agentic commerce.
The scale of change is already visible. ChatGPT now has 900m[1] weekly average users and Gemini has 750m[2]. Traffic to retail websites from generative AI assistants has surged by 4,700% year-on-year, with visitors arriving via agents spending more time, browsing more deeply, and converting more effectively[3]. Consumers are rapidly delegating discovery, comparison and even purchasing decisions to AI systems that can search, filter and transact in seconds.
Agentic commerce is not just a mild evolution of the shopping channel – it is a structural shift because shopping is becoming conversational. Customers ask a question and receive a handful of recommendations. No scrolling. No ‘second page’ on Google. And no second chance for retailers and brands that are otherwise invisible to the AI Agent. Earning the loyalty of the AI agent, therefore, is becoming as important as earning the loyalty of customers themselves.
At dunnhumby, we see this shift clearly because we sit at the intersection of retailer data, brand investment and customer behaviour. The signals that are shaping the new agent decisions (price, availability, substitution patterns, loyalty behaviours) are the same signals that have always driven commercial performance. The difference now is that they need to be structured and communicated for machines, not just humans[4].
Loyalty is being disintermediated
Traditional loyalty was built on a direct relationship between retailer, brand and customer. Agentic commerce breaks that link. AI agents are much more rational, being influenced far less by brand storytelling, impulse merchandising and theatre. They optimise for value, clarity, relevance and convenience[5]. They compare prices, apply constraints and execute without emotion.
The consequences are significant:
Customer relationships become indirect, mediated through platforms
Price and value transparency becomes near perfect
Traditional drivers of loyalty lose influence
New gatekeepers - LLMs and agent ecosystems - gain disproportionate influence over consumer choice
If nothing changes, loyalty will migrate towards the AI agents.
Discoverability is the new competitive battleground
This is what we see from working with brands and retailers in this new domain of agentic commerce optimisation. Discoverability is now the primary growth driver in agentic commerce. Most purchases influenced by agents will still close on retailer sites (declining from c.95% today to ~80% over time)[6] but the decision - what gets considered, shortlisted, and selected - is increasingly happening upstream in agent environments. Winning is no longer about winning the shelf or the search result. It is about winning the shortlist. AI agents typically recommend 5-8 products, and there is no page two of google anymore. The competition for one of those prized spaces will only get more intense.
And that is where many brands are already exposed:
Products not appearing in agent recommendations (“agent blindness”)
Incomplete or inconsistent product representation (inaccurate “AI shadows”)
“Loyalty leakage” as AI agents redirect demand to competitors
In a winner takes most dynamic, even small gaps in visibility have the potential to translate into disproportionate losses.
The second layer of agentic loyalty: memory and personalisation
There is a second, less visible shift happening in parallel and it raises the stakes further. AI agents are not just making decisions in isolation. They are starting to remember, learn and personalise over time[7].
This means:
Agents build a persistent understanding of preferences, behaviours and past choices
Future recommendations are influenced by what the agent has previously seen, selected and trusted
“Default” brands and products become embedded in personalised recommendation patterns
In effect, agents are developing their own version of loyalty. This changes the competitive dynamic again. It is no longer just about being discoverable once. It is about becoming:
Repeatedly selected
Consistently preferred
Embedded in personalised decision models
And here is the critical implication. If you are not influencing the agent today, you may not be considered tomorrow. Because once preferences are learned and reinforced, switching becomes harder, not easier.
Why most organisations are not ready
Despite the pace of change, most retailers and brands are structurally unprepared. Product data is not designed to be interpreted by machines. Content is inconsistent across channels. Value is often implicit, not explicit. Processes for developing and syndicating content as slow and cumbersome. Access to derived behavioural attributes that can influence LLMs are scarce and hard to come by. And critically, most organisations are still optimising for human journeys not machine decision models.
As a result, when agents interpret the market, they rely on whatever signals are most structured, consistent, and easy to understand. That is not always the best product. It is the best-described product.
What winning looks like
Winning in this new environment requires a shift in mindset:
1. Design for agents, not just customers
Structure product data so it can be parsed, compared and ranked with confidence.
2. Make value explicit
AI Agents reward clarity - on price, benefits, availability, and use cases.
3. Engineer discoverability upstream
Influence how agents understand, shortlist and recommend your products or brand through content, context and behavioural attributes which add ‘brand authority’[8].
4. Compete on both sides of the journey
Win discoverability in external agents and deliver differentiated experiences in your own channels.
5. Build for memory, not just moments
Ensure your brand is consistently selected so it becomes part of the agent’s learned preferences and future recommendations.
Start now or lose the race before it begins
Agentic commerce will not replace traditional retail overnight. But it will reshape how demand is created and allocated, faster than most expect. Early movers will build structural advantage:
Influencing agents to become ‘loyal’ to their brand and products
Embedding into recommendation ecosystems
Becoming the “default” choices in personalised agent memory
What to do now
1. Audit your “AI visibility”: test how your top products appear in ChatGPT, Gemini and retail agents. Identify gaps in ranking, representation and consistency.
2. Fix your product data first: prioritise your revenue-driving products. Ensure attributes, benefits, use cases and price signals are structured, explicit and consistent across channels.
3. Define your “agent signals”: discover what proof points matter for agents (price advantage, ratings, availability, need states, substitution logic, nutritional or sustainability data) and make them machine-readable.
4. Run controlled experiments: treat agent environments like a new channel. Test changes to content and measure impact on your products share of voice.
5. Align ownership: assign clear accountability across data, media and loyalty teams as the new agentic commerce world cuts across the silos and it won’t happen by accident.
Late movers will be forced to buy back visibility in a market where control has already shifted. The implication is clear:
In the age of agentic commerce, you are no longer just competing for the customer’s attention. You are competing for the agent’s recommendation and its memory. And those that move first will not just win loyalty. They will define it.