What happens to brand loyalty when AI agents like Meta's Muse do the shopping?

Loyalty that rests on habit weakens, because an AI agent has no habits, only instructions. When an agent such as Meta's Muse books, shops and cancels on someone's behalf, it matches the person's stated need to the clearest offer, so inertia stops protecting revenue and loyalty has to be re-earned on every request. The brands that get picked are the ones whose public footprint plainly answers a specific need.

Key takeaways

• Much of what brands book as loyalty is inertia: people rebook, renew and stay with what they already know.
• Meta's Muse agent launched on 8 September 2026 and can book tickets, shop, send messages and cancel things for its users; Meta says it carries no advertising today.
• Agents do not browse, so banners have nowhere to land; any future ads may be limited to a marked sponsored slot below the answer, or replaced by a fee on the transaction.
• In Southeast Asia, Muse inside WhatsApp joins an existing habit of delegating through chat, and broad regional taglines will not match customers' specific requests.
• Plan for three shifts: selection outranks attention, clarity outranks reach, and loyalty has to be re-earned every time.

Why has habit been doing marketing's job?

For years, a quiet share of marketing growth has not come from marketing at all. It has come from inertia. People rebook the hotel chain they already know, renew the subscription they forgot they had and stay with the insurer their parents chose. Brands have been renting that habit and booking it as loyalty.

AI agents do not rent anything. An agent has no habits, only instructions. Investors seem to have noticed: shares in subscription businesses that depend on customer inertia have slumped since Meta's Muse agent launched on 8 September.

What did Meta actually show?

At Meta Connect this week, Mark Zuckerberg held up the Muse Charm, a small keychain gadget with an animated avatar on its screen. You press it, speak, and your Muse agent gets on with the task. Meta says it will ship by December. The Muse agent itself went straight to the top of the app charts and can already book tickets, shop, send messages and cancel things on your behalf.

The gadget may well flop, as plenty of AI hardware has. The behaviour behind it is the part to take seriously: people are getting comfortable saying what they want and letting something else go and find it.

Why does a banner ad have nowhere to land?

Digiday's briefing this week made the point that an agent does not browse. It pulls a stripped down, machine readable version of what is out there, completes the job and moves on. There is no feed to interrupt and no page to sit beside.

Meta insists Muse carries no advertising today. If ads do arrive, the briefing argues they cannot simply buy the agent's answer, because users would stop trusting it the moment they suspected the top pick had gone to the highest bidder. The most money could buy is a marked, sponsored position below the answer, which you could call the alternative slot. Zuckerberg has floated a different model altogether: a small fee charged to the merchant whenever Muse helps someone buy or save. That is a toll on the transaction, not an auction for attention.

Either way, the old lever of buying more reach and frequency loses most of its grip.

Why will Southeast Asia feel this first?

Muse is also offered inside WhatsApp in the markets where it is available. In a region where WhatsApp is already how people contact their clinic, their property agent and their favourite bakery, a chat agent is not a new behaviour. It is a new participant in a very old one. A decade of super apps has taught people here to delegate the ride, the bill and the phone top up. Delegating the choice is a short step.

The region also exposes what agents punish. Many brands run one campaign across six markets and several languages, adapted just enough to clear local review. But agents work from the customer's own words, and those words are specific: halal options, near the MRT, open on Hari Raya, quiet enough for a baby. A brand that has only ever spoken in broad regional taglines will not match any of them.

Agencies face the same problem. Much of what they sell is still media weight. When the first reader is an agent, the valuable deliverable becomes knowing, in the audience's own language, which needs a brand should be the obvious answer to, and whether a competitor already owns that answer. Independents building their own insight stacks are closest to it.

What does losing a sale to an agent look like?

Picture a Friday evening in Singapore. A marketing manager on the MRT squeezes the charm on her tote and says, "Find us somewhere quiet for Saturday night, near the river, under three hundred dollars."

By Clementi she has her answer: one boutique hotel that fits, with two sponsored options beneath it. She taps yes. She never opened Instagram or a search page, and she never met the chain that spent the most on her attention this quarter.

That chain did not lose on price. Nothing the agent could read tied it clearly to "quiet, near the river, a Saturday for two". Its content was built to catch an eye, not to answer a need. One agency chief told Digiday the world "could suddenly become very small". Brands used to get many chances in front of a customer. Soon they may get one.

Which three shifts should brands plan for?

Ads will not disappear this quarter. Muse still has to earn enough trust to act for people, and Amazon has already shut it out. But the direction is clear enough to plan around.

1. Selection outranks attention. Before a human sees your brand, a machine may have to decide whether you fit what that human asked for. That judgment starts from the person's words, not your campaign keywords.

2. Clarity outranks reach. Agents match stated needs to offers. The brands that get picked have a public footprint that answers a specific need plainly, across content, reviews, product pages and positioning. Generic messaging that tries to be everything to everyone reads to a machine as belonging to no one in particular. Some brands already track how often AI answers cite them.

3. Loyalty has to be re earned every time. When cancelling takes one sentence, inertia stops protecting revenue. The relationship has to survive a sceptical piece of software reviewing it on the customer's behalf.

How do you find the need your brand can own?

All of this comes down to reading what audiences really say before an agent does, and that is the problem SOMIN was built around. The platform reads what people express across social content, groups it into the needs and tensions behind it and shows where competitors already sit. A brand can then see which specific needs it can credibly answer and aim its work at being the clear answer rather than the loudest option. Our own tracking shows the market shifting: over the past 30 days, one ad tech competitor put more than a quarter of its posts into how AI is changing creativity and discovery, while we had mostly been discussing AI's effect on content quality. The payoff from reading real audience signal is already visible. Mothercare Singapore, whose local team had been guessing what would land under global direction, cut research and ideation time by 80 per cent and won back more than 25 hours a week. Singapore performance agency KPI Media reports cutting pitch preparation from a week to two days and from US$4,000 to US$1,000 a pitch, and it won a new retainer worth US$5,000 a month. Neither result came from shouting louder, and both point to the question to take into your next planning meeting: if a customer's agent had to explain in a single sentence why it chose your brand, what would it say?

How does SOMIN fit, and how is it used?

In this new ecosystem, SOMIN acts as an audience-intelligence bridge that translates raw consumer signals into machine readable brand clarity before an AI agent makes a selection. Rather than relying on guesswork or broad global taglines, the platform systematically ingests organic social expressions, identifies underlying consumer needs and tensions, and maps out where competitors currently sit. Marketers handle this intelligence by using SOMIN to pinpoint unowned market white spaces, tailoring their messaging, product pages, and positioning to explicitly answer specific, localised long tail queries. By shifting the focus from buying media weight to defining a brand's exact value proposition, SOMIN drastically cuts research and pitch times allowing teams to establish the precise, single sentence rationale that both AI agents and human buyers need to select their brand over a competitor.

FAQ

Related reading

• What is agentic commerce, and how do brands get chosen by AI shopping agents?
• Does a faster AI agent make better marketing decisions?
• What is agentic AI in marketing, and how is it different from automation?
• How is AI changing the role of the marketing scientist?
• What is a consumer tension in marketing, and how do you find one?