When AI Becomes the Shopper: What Happens to Brand Loyalty?
Brand preference doesn’t disappear, but it splits into two audiences. The consumer still forms the preference. They decide which brands they trust enough to let an agent act for them, and that decision stays emotional and human, says Brian Kealey, Managing Director, APAC, Klaviyo.
Topics
A shopper may still have a favourite brand. But increasingly, they may not be the ones doing the shopping.
New Klaviyo research among more than 1,000 Singapore shoppers found that 76% have used AI to help find or decide on a purchase in the past three months, while 51% are open to letting AI agents handle routine repurchases on their behalf.
That changes the meaning of loyalty.
The consumer still decides which brands deserve a place in their life, but the next purchase may be made by something that has never seen an advertisement, felt an emotional connection to a brand, or stood in a supermarket aisle.
It may simply assess what is available, what fits the shopper’s preferences and what can be delivered reliably, then act.
For marketers, this creates an unusual new problem. Brand preference remains deeply human, but increasingly, its expression is becoming machine-mediated. A brand must now be recognisable and relevant to the consumer while also being legible, reliable and trustworthy to the AI acting on their behalf.
That tension is becoming more important across APJ, where consumers are adopting AI-assisted shopping rapidly while continuing to place value on relevance, personalisation and trust.
For Brian Kealey, Managing Director, APAC at Klaviyo, the rise of shopping agents does not make brand loyalty less important. Instead, it gives loyalty a second layer: brands must earn the consumer’s preference first, then ensure the systems behind the brand can live up to it when an agent comes to make the next decision.
In this interview, Kealey discusses what brand preference means when AI enters the buying decision, why customer data and integration are becoming the infrastructure of AI-led commerce, how APJ’s data-value exchange is evolving, and what will separate genuine loyalty from simply faster marketing.
Excerpts from the interview;
If AI starts making more purchase decisions on behalf of consumers, what does “brand preference” actually mean? Is loyalty still built by the consumer, or will brands increasingly need to earn the trust of the AI making the recommendation?
Brand preference doesn’t disappear, but it splits into two audiences.
The consumer still forms the preference. They decide which brands they trust enough to let an agent act for them, and that decision stays emotional and human.
Once that trust is granted, a second audience appears, the agent itself, and it judges brands on things a person might overlook, like whether product data is clean and fulfilment is reliable.
And increasingly, the agent is not just reading about brands, it’s dealing with their systems directly. So earning its trust means being reliable at the level of data and infrastructure, and not just messaging.
Loyalty is still built by the consumer, but it is increasingly expressed through an agent, and a brand has to satisfy both. The mistake would be to think brands can win the agent on its own by optimising for price and feeds. An agent is only ever acting on a mandate the customer gave it, and that mandate comes from preference.
At Klaviyo, we believe that the brand has to keep owning the direct relationship with the person, because that is what tells the agent to keep choosing you. Earn the human, and the agent follows. Try to game the agent without the human behind it, and you are one price comparison away from being swapped out.
Klaviyo’s research shows 76% of Singapore shoppers have already used AI to find or decide on a purchase. What does that tell you about how quickly the traditional path from discovery to consideration to purchase is changing in APJ?
It tells me the shopping funnel we all grew up with is compressing, and in a market like Singapore, it’s moving faster than a lot of brands realise. Discovery, consideration and purchase used to be separate stages that a brand could influence one at a time.
When 3 in 4 shoppers are already using AI to find or decide, much of that collapses into a single assisted moment. Now we can ask an assistant for a recommendation or a shortlist of desired items, and act based on what it says. Sometimes, this happens without visiting a brand’s site at all.
What strikes me about APJ is the pace. These are mobile-first, high-expectation consumers who adopt new behaviours quickly, so this isn’t a slow drift. It’s already the default for a large part of the market.
The implication for brands is uncomfortable but simple. You can no longer count on owning each stage of the journey through your own channels, because an assistant is mediating the whole thing.
What you can still own is the customer relationship and the data behind it, and that is what lets you show up well inside that assisted moment rather than being summarised by a machine that barely knows you.
You’ve worked across Salesforce, Microsoft, Amazon and MuleSoft. From that perspective, what is the biggest organisational shift B2C brands need to make if they want to compete in an AI-led commerce environment?
The thread running through all of those companies is that the hardest problem in enterprise technology is rarely the application on top. It’s the data and integration underneath.
A lack of integration has always been the biggest barrier to adopting new technology, and AI has only made that more visible, because an agent is only as good as the customer context it can reach.
The organisational shift I would point B2C brands to is not about buying AI but about tearing down the internal silos that keep customer data in separate systems. In many businesses, marketing, service, ecommerce and store data each live in a different team and tool, with its own definition of the customer.
That was survivable when people were stitching it together slowly. It is fatal when you are asking an agent to act in real time. The brands that compete will be the ones that treat a single, real-time view of the customer as core infrastructure and organise their teams around it.
That is a leadership and operating model decision far more than a technology purchase, and it’s usually the thing nobody wants to own because it’s unglamorous. And that’s exactly why it ends up being the differentiator.
There is a further reason this is the shift that matters, and it’s where things are heading next. Once your data is unified and able to act in real time, that same foundation is what lets your brand deal with an external agent at all.
Increasingly, the party on the other side of an exchange is another piece of software instead of a person clicking through your site, and that’s the machine-to-machine relationship your systems have to be built for.
If you are able to treat a connected, real-time data core as infrastructure, you will be able to transact agent to agent when it becomes routine.
Singaporean consumers say they are willing to exchange more personal data for more relevant recommendations. Do you think this represents a broader shift in the APJ data-value exchange, or will expectations around privacy and personalisation vary significantly across markets?
Both are true, and holding them together is the whole skill. The willingness is real, and I do think it reflects a broader shift.
Across APJ, we have digitally fluent consumers who grew up with personalised services and will trade data for relevance, so the 58% is not a Singapore quirk. But it is conditional, and the conditions are set locally.
Privacy laws differ sharply from market to market and draw hard lines around what brands can collect and how they can use it. Cultural comfort with data sharing varies, too, as does trust in platforms and institutions. So the underlying exchange, data for relevance, travels across the region, but the terms of it do not.
A brand cannot take one consent model or data-use approach and stamp it across APJ. What holds up is a consistent principle that data is borrowed in return for a benefit the customer can feel, paired with handling designed market by market to sit within local law and local expectation.
Get that principle right, and the willingness is there. Treat the 58% as blanket permission, and you break the trust that it was built on.
Looking three to five years ahead, what will separate brands that successfully build loyalty in an AI-mediated shopping world from those that simply automate more marketing?
The dividing line will be whether AI made the relationship better or just made marketing faster. It’s easy to use AI to send more of the same and personalise superficially.
That improves efficiency for a while, but if the underlying experience is not more relevant, brands are spending down customer patience, especially given the fact that what people want most is for brands to stop sending them things that do not matter.
The brands that build real loyalty will use AI to understand the customer better and then act on it, so the person feels known rather than targeted. The measure that separates them is customer lifetime value, because it reflects the whole relationship instead of the last campaign.
If lifetime value is rising, AI is deepening the relationship. If it is flat while activity climbs, brands are automating their way to diminishing returns.
Three to five years out, as agents mediate more of what people discover and buy, the durable advantage will be being a brand the customer wants to return to, and being legible and trusted enough that their agent keeps choosing you. That is earned through relevance and consistency, not volume.
Everything else is just faster marketing.
ALSO READ: Has RCS Solved Mobile Marketing’s Biggest Problem: Trust?
