AI agents are coming between you and your customers

Agents are everywhere, and not just on the enterprise side of transactions anymore. Meta’s Muse isn’t the first consumer agent to launch, and it bears a striking resemblance to its open-source predecessor, OpenClaw. It’s causing quite a stir. Its fast growth is forcing brands to make some quick decisions or reap the consequences of making none.

Just as brands are making sense of their own AI agents, consumers are getting agents of their own. Those agents are changing who controls the customer relationship. They can compare products, evaluate policies, and make purchases without customers ever visiting a brand’s website, leaving brands with less visibility into customer preferences and purchasing decisions.

Brands now face a choice: Figure out how to compete when consumer agents increasingly mediate the shopping experience or risk losing customers to competitors that do.

A less-than-amusing start

Retailers’ signals are mixed. On Sept. 20, 2026, Amazon started showing a pop-up to people using Meta’s new Muse agent to shop its store. The message warned that continued access by an “unauthorized AI agent” violates Amazon’s Conditions of Use. Yet two days later, PayPal announced that its customers could shop and check out through Muse at PayPal merchants worldwide. Similarly, Shopify enabled Shop Pay checkout for Muse across all of its stores the day before. Within days, three major players made two different decisions.

Most of the agentic AI conversation in marketing this year centers on the agents that brands deploy, including virtual teammates, orchestration layers, and pilots trying to move beyond the pilot phase and roll out across the enterprise. Meanwhile, consumers forged ahead and downloaded agents of their own. Muse passed 2.5 million downloads in its first two weeks, according to Sensor Tower data reported by CBS News, and briefly moved past ChatGPT at the top of Apple’s App Store.

Muse has company. OpenAI launched Dots in late September, an always-on assistant inside ChatGPT that runs workflows in the background across connected apps. Dots requires a paid ChatGPT Pro or Business Premium plan and isn’t available in the EEA, U.K., or Switzerland. Google has Gemini 4 Argon in limited preview through its Fairwind Program, with no consumer release date yet.

Apple’s agentic Siri works across on-device apps and personal data, and Anthropic offers desktop control through Claude’s Cowork app on its Pro and Max plans. Meta and OpenAI both run each user’s agent in a dedicated cloud environment, so the browsing, comparison, and checkout steps happen on infrastructure the brand never touches. Each company sets its own rules for how an agent identifies itself and when it asks for the user’s permission. A brand’s access policy must account for all of them.

Every brand now faces the same access decision Amazon, Shopify, and PayPal just made. But there’s a harder question: When a customer sends an agent to shop, who holds the relationship, and how can your brand retain as much of it as possible?

Recommended Articles

The state of AI in marketing: Q4 2026

Oct 12, 2026

AI budgets are growing faster than proof of ROI

Oct 9, 2026

Not everyone in marketing should be an AI builder

Oct 9, 2026

What marketing teams should do now

Your brand needs to decide how to handle consumer AI agents and how much access to give them. The right approach depends on your products, services, and customer experience, but the decision shouldn’t happen by default.

Be intentional about the decision

Amazon’s objections to access from Muse are specific, at least based on public statements. The company told GeekWire that Meta never notified it that Muse would access the store. It also says the agent doesn’t identify itself while browsing and appears to capture and store customer credentials.

Amazon has the gravitas to take that position. It generated more than $68 billion in ad revenue last year from people browsing its pages and runs its own shopping agent, Buy for Me, which identifies itself and lets brands opt out. By contrast, most brands have neither an ad business even close to that size nor an agent of their own to protect.

The more common scenario looks like Target’s. A CNN reporter testing Muse could browse Target’s site with the agent but had to enter personal details by hand because the retailer blocked autonomous clicks at checkout. I don’t know whether anyone in Target’s marketing organization made that call. Still, at many companies I’ve worked with, a decision like that starts as a bot-management rule set by the security team, and marketing learns about it when the numbers come in as conversion drops.

The legal ground also shifted in favor of the companies behind the first wave of consumer agents. In August, the Ninth Circuit ruled in Amazon’s case against Perplexity that the “user” is the one accessing the site under federal anti-hacking law, and the AI company acts on that user’s behalf. The ruling limits Amazon’s ability to use federal anti-hacking law to block agents acting on customers’ behalf. Terms of service remain another way for retailers to restrict agent access, as Amazon’s warning to Muse users demonstrates.

There’s a potential “middle” path for many brands. Visa’s Trusted Agent Protocol, developed with Cloudflare, lets merchants verify signed agents and separate them from malicious bots. Mastercard requires agents to register and be verified before they transact on its network, and every Mastercard issuer in Europe is now enabled for Agent Pay. A brand can use those signals to admit verified agents at checkout while still blocking scrapers.

Amazon argued that third-party agents buying from other businesses “should operate openly.” Most brands would agree with that standard, and the networks now give them a way to enforce it without shutting the door.

A block keeps the browsing session on your site for customers who give up and shop manually. The customers who keep using their agent take it, along with everything it knows about them, to a retailer that lets it in.

Start to assume the customer’s agent holds the context

Meta designed Muse to remember what matters to the consumer, including details the person mentioned only once. Meta’s own launch example has the agent recalling friends’ dietary restrictions before it sends dinner party invitations. Muse keeps a full audit trail for the user, and Meta says it doesn’t share conversations or data from the user’s secure VM with its ad systems.

The preference data brands spent a decade collecting through on-site behavior now accumulates in a private environment the brand can’t see.

Marketing teams will first notice the impact of consumer AI agents in their measurement data. According to The Keyword’s reporting on Shopify’s Meta channel terms, Google Analytics and custom pixels don’t fire in Meta’s direct checkout. Shopify passes only server-to-server events for checkout started and completed. The order is recorded as the merchant’s, and the behavioral trail that led to it never reaches the merchant’s analytics. Some teams spend years building their CDPs around browsing sessions, yet none of that infrastructure captures an agent checkout.

In Checkout.com’s 2026 research, 72% of merchants in the U.K. and U.S. agreed that consumers will adopt agent-led shopping faster than most merchants are prepared to accommodate.

Consumers still have limits on fully delegating tasks. Deloitte’s survey of 13,500 European consumers found that 31% are open to leaving the final purchase to AI and 47% are unwilling. Brands that treat today’s reluctance as a given for the coming years will repeat the mobile cycle, when customers carried smartphones for years before most brands had a checkout that worked on one.

Expect consumers’ agents to hold your policies to their own standard

Muse checks out with Stripe’s Link, and Meta says Muse is the first agent covered by Link’s purchase protections, which include coverage for damaged or lost items, price drops, and no-fee returns on eligible purchases. The consumer’s wallet now sets a baseline that the agent compares a brand’s terms against on every purchase, and the brand never sees the comparison happen.

For example, picture an apparel brand whose policies don’t line up across its channels. It offers 30-day returns on its website, 14 days on orders placed in its app, and final sale on anything bought through a marketplace.

For years, human shoppers either missed the differences or sorted them out with a customer service call. Now a customer asks her agent to find a wool coat under $300 from a brand with straightforward returns. The agent pulls policies from four brands in a few seconds, notes that the apparel brand’s terms change by channel, and puts a competitor with one consistent policy at the top of its results. The apparel brand’s team never sees a bounce or an abandoned cart. The sale simply goes elsewhere.

I call this a “coherence at velocity” problem. Teams are moving quickly, but the challenge is maintaining consistency across siloed teams. Policy inconsistencies that human shoppers tolerated for a decade will now cost sales because an agent needs specific, consistent, quantifiable information on every transaction.

Consumers also have opinions on who answers for agent purchases that go wrong. In a Riskified survey of U.S. and U.K. shoppers, 50.8% said the AI platform should be responsible for unauthorized purchases, 23.2% said the retailer or brand, and 18.7% said the consumer. Retailers carry less of the expected liability in that split. They also have less influence over the transaction because the agent platform owns the interface and memory.

Make your brand compelling for consumer agents

Brand preference can survive this new type of delegation, even if it works differently from the traditional mechanisms brands have relied on for decades. Accenture’s “Talk to My AI Agent” research found that 56% of consumers would tell an agent which brands to consider, while 37% of behaviorally loyal consumers would accept an agent switching brands if the alternative better fit their needs. Brand preference holds when the agent can verify it against something concrete, such as consistent terms, accurate inventory, and loyalty benefits that apply without a human logging in.

Starting with the loyalty ID and the first-party account, the assets that carry through an agent transaction are the ones a brand owns outright. And while much of the focus is on the purchase step, the post-purchase relationship matters just as much because returns, exchanges, and service requests will increasingly start with the customer’s agent as well.

The cost of getting started has dropped, following a familiar pattern with emerging technologies. PayPal supports WebMCP, an emerging standard that enables merchants to participate in agent commerce through their existing storefronts. “We’re not ready” no longer justifies waiting another planning cycle.

Here are some practical first steps you and your team can take in the months ahead:

• Audit your default settings: Shopify shares merchant products with Meta via Shopify Catalog by default, and direct checkout is enabled by default for eligible stores. Even if you aren’t ready to make new investments or do additional implementation, at least know what you’ve already agreed to.
• Determine what steps and how far in the purchase funnel you want consumer agents to go: Decide separately what verified agents can do during browsing, cart creation, and checkout, using Trusted Agent Protocol or Agent Pay signals in place of blanket blocking.
• Create your policies as machine-readable structured data: Keep return, shipping, and warranty terms identical across every channel, and make them machine-readable.
• Make your loyalty program agent-ready by making it portable: Let an agent pass a loyalty ID and have benefits apply at checkout.
• Rebuild your measurement for server-side events: Set up agent-channel reporting before you’re asked to explain an unexpected Q4 reporting gap.

A new field for competition

When customers use agents to shop, brands whose information agents can verify and whose benefits are tangible enough for machines to understand will increasingly win. Amazon can defend its relationships with a wall and its own agent, but everyone else has to compete on what the agent can read and use.

Take this opportunity to document siloed policies and make sure the teams and platforms involved in your customers’ transactions are aligned. These inconsistencies likely affected your customer experience even before consumer agents entered the picture. While some standardization may be coming, your brand, regardless of size, has an opportunity to capture customers through this growing channel.