Imagine asking an AI assistant to find a pair of Nike running shoes for a 51-year-old man attempting his first London Marathon. They need to cost less than £150 and provide enough support for a long-standing injury to his left meniscus.
To be clear, you are more likely to see me run the marathon than wear a Tottenham shirt… but neither is likely.
That said, a man can dream.
You do not need much imagination for the rest. An AI assistant can already identify the features that count, compare products and reviews, check what is in stock and help complete the purchase.
Nobody really knows how quickly AI will become part of everyday life. But the direction is clear. AI is moving beyond answering questions and starting to manage what happens next.
For brands, the risk goes beyond losing website traffic. They could also lose influence over why someone chooses them and what happens after the sale. It’s one reason the conversation needs to stretch beyond marketing and into how brand, commerce, data and customer experience work together.
AI Is Shaping Choices Before It Completes Purchases
We are in a slightly awkward middle stage of a transformation cycle.
A McKinsey survey of 749 people in France, Germany, and the UK found that 38% had used AI to research products or decide what to buy. Comparing brands, models, prices, and reviews was the most common use. Far fewer people used AI to build a basket or complete a purchase.
AI is already helping people decide what to buy. Most, however, are not yet ready to hand over their credit cards and let it take care of the rest.
Yet, the commercial signals of what this technology is capable of are getting harder to ignore. Adobe Digital Insights found that, in March 2026, people arriving at US retail websites through AI were 42% more likely to buy than visitors from other sources. Amazon says more than 250 million customers had used its Rufus shopping assistant at the time of its update, and that those who used Rufus were 60% more likely to make a purchase.
The numbers are striking, although there is reason to be cautious about what they tell us. They do not prove that AI caused the sale. People asking an assistant for a specific recommendation may simply be more ready to buy.
A study of 31 million Ctrip users adds some balance. Only 6% had adopted its AI assistant at the start, and those who did still moved between AI and traditional search rather than replacing one with the other.
To be clear, the old customer journey has not disappeared. AI is finding a place within it, and how large that place becomes will probably depend on the purchase and the risk involved. Someone might happily let an assistant reorder washing powder. A family holiday, a car, or a financial product is another matter.
Brands Must Appeal to People and Make Sense to AI
A lot of the current conversation is about “agentic discoverability.” Put simply, that means whether an AI assistant can find your brand, understand what it offers, and include it in its recommendations.
Visibility gives a brand a chance of making the shortlist and, from there, the sale.
Of course, making the shortlist is only the beginning.
McKinsey’s US research suggests that a brand’s own website may account for only 5% to 10% of the sources used by AI search. AI assistants also look to publishers, reviews, online communities and product information supplied to retailers and platforms.
So brands need accurate, consistent information wherever an AI assistant might look. Their claims must also be supported by sources that people can trust and AI can verify. The complication is that an AI assistant can understand a brand perfectly well without making anyone particularly want it.
If brands focus only on product information, availability and price, they risk becoming rows in an efficient comparison table. Brand building is still required to create the memories, trust and preferences people bring to the conversation or type into their prompt.
A customer might tell an assistant they prefer Nike, trust a particular retailer, or want to use an existing loyalty scheme. Equally, the assistant might introduce a lesser-known brand because its reviews are better or its product is more suitable.
The complication is that an AI assistant can understand a brand perfectly well without making anyone particularly want it.
Brands now have to work on both sides, giving people a reason to choose them while making sure AI has reliable information to work with.
AI Is Reshaping the Whole Customer Journey
This is why, at PMG, we work shoulder to shoulder with our customers across the whole customer journey and across platforms. In practice, we help customers build brands in a world where AI increasingly shapes choice. Then we connect that work to the sale and what follows, from retention and loyalty to advocacy.
There’s risk in treating being found by AI as a separate search problem, because it preserves the very gaps that the large technology platforms are moving in to fill. I find it useful to think about four connected jobs.
Four Connected Jobs Across the Customer JourneyJob | What the brand needs to achieve | What that means in practice |
|---|---|---|
Preference | Give people a reason to choose and trust it | Strong brand building, good creative, a good product and a positive reputation |
Understanding | Give AI a credible reason to recommend it | Accurate product information, supported claims, reviews and trusted coverage |
Access | Make the product easy to check and buy | Live prices and stock, simple payment, dependable delivery and clear returns |
Relationship | Turn the sale into an ongoing customer relationship | Permission to communicate, linked accounts, loyalty, service and retention |
In practice, these jobs tend to run into one another. The recommendation needs to lead to an available product, the sale needs to leave room for a customer relationship, and all of it rests on giving people a reason to want the brand in the first place.
The Customer Relationship Is Starting to Shift
Retailers and AI platforms are already becoming more joined up, with Walmart’s planned integration with Gemini giving us a useful picture of how these four jobs might come together. Gemini can already help someone discover Walmart products, and if they link their Walmart account, it can also use previous purchases, add products to an existing basket, and recognise Walmart+ benefits. The result is an AI experience that can draw on Walmart’s customer knowledge and loyalty programme.
But that creates a new power dynamic. Walmart may process the sale, while Gemini becomes the place where the customer explains what they need and the shortlist takes shape.
OpenAI’s Instant Checkout exposes the same tension. The retailer may still handle payment, delivery, returns and customer service, even when the recommendation and checkout happen inside ChatGPT.
That means the sale and the customer’s attention may end up in different places.
This is where agencies need to help customers navigate the complexities of our ecosystem and look beyond the immediate transaction. Alongside the familiar question of whether we made the sale, we need to ask who shaped the choice, who understands the customer and who is most likely to influence what they do next.
Brands Should Prepare Without Chasing Hype
Some people believe this is simply another wave of AI hype. It may take years for customers to become comfortable with AI buying things for them, and some purchases may never become highly automated.
That caution feels reasonable. Even so, brands may find it difficult to wait until everything is settled before they start preparing. Product information, measurement, customer consent and loyalty systems cannot be fixed in a week if behaviour suddenly changes.
A sensible approach is to build steadily, test with real customers and invest more when and where the evidence supports it.
I don’t think the answer is to drain money from brand building to fund a new technical specialism simply because AI is fashionable. Perfect product information can only take a brand so far if people don’t want it, while fame offers little protection if an AI assistant misunderstands or misrepresents it.
For now, we need to keep working on both sides, for humans and AI assistants, at once.
Leaders Must Connect the System Now
A useful place to start is with the journeys that offer the greatest value. Take the real questions customers ask and follow what happens from initial research through to purchase and repeat purchase.
That means continuing to invest in the brand and creative work that makes people want you, while making sure your products, prices, claims, reviews and availability are understandable and verifiable by AI.
It also means thinking carefully about the customer relationship, including how accounts, permissions, loyalty, service and returns will work when the journey begins on somebody else’s platform.
Bringing those pieces together will require one shared plan. Brand, media, search, commerce, data and customer teams need common goals, shared measures, and someone responsible for seeing the whole picture.
Measurement will need to look beyond the sale. Brands should track whether AI recommends them, whether its description is accurate and which sources it uses, then connect that to sales, recognised customers, loyalty membership, repeat purchases, returns and long-term value.
In your next leadership meeting, I would ask three simple questions.
If an AI assistant brought us tomorrow’s customer, would it understand why it should recommend us? Could the customer complete the journey easily? And would we be able to build a relationship with them afterwards?
For me, the work now is about connecting brand, commerce, data and customer experience before AI platforms connect them on the brand’s behalf. If those questions sit with several different teams and nobody can see the whole picture, that is probably where your work needs to start.
