Imagine asking an AI agent to find a hotel for a weekend in Tokyo, compare the best options, account for your budget and preferences, and make the booking. You never open Google, click a travel ad, visit five hotel websites or compare prices yourself.
For marketers, that scenario raises a basic question: if the customer delegates more of the journey to an AI agent, where does marketing influence happen?
Meta made that question more immediate this week with Muse, its new personal AI agent. Meta says Muse can take actions on a user’s behalf, work across connected services, use a browser, remember information from earlier interactions and proactively suggest what to do next.
The important marketing signal is not simply that Meta has another AI product. It is that a major consumer platform is building an intermediary that could increasingly sit between a person’s intent and the brands, products and services that satisfy it.
Table of contents
Jump to each section:
- Meta wants Muse to do things, not just answer questions
- AI agents could become another gatekeeper between brands and customers
- Marketing to an agent may look different from marketing to a person
- Brand still matters even if the agent handles the transaction
- Meta could end up controlling another layer of consumer intent
- The next funnel may have two audiences
Key Takeaways
- AI agents could add a new intermediary between consumer intent and brand choice.
- Machine-readable product data, reputation, availability and integrations may become more important alongside traditional brand building.
- Marketers may increasingly need to influence human preference while also making products easy for AI systems to evaluate and transact with.
Meta wants Muse to do things, not just answer questions
Meta describes Muse as a personal AI agent designed to act, not simply generate responses. In its launch announcement, the company says people can give Muse a goal and let it work through connected services and a dedicated browser.
Muse also remembers information from previous interactions and can use that context to make suggestions without waiting for a new prompt. Meta says the agent runs inside a dedicated Muse Secure VM and is powered by its Muse Spark model family.
That matters because the commercial implications are different from a chatbot that mostly answers questions. The more an agent can compare, book, fill forms, communicate and complete transactions, the more of the customer journey can potentially happen through the agent rather than through a sequence of websites and apps controlled by brands.
Introducing Muse, your personal AI agent from Meta that gets things done across every part of life.
Download the Muse app and get started: https://t.co/KBjYWfshGo pic.twitter.com/1exp56xj93
— Muse (@Muse) September 8, 2026
AI agents could become another gatekeeper between brands and customers
Digital marketing has largely been built around the assumption that a person performs the discovery process. The consumer searches, scrolls, clicks, compares and eventually makes a decision.
An agent-mediated journey could compress some of those steps. A traveler might ask for the best hotel near a conference venue. A shopper might ask for running shoes below a certain budget that suit a previous injury. A restaurant customer might ask for a dinner reservation that fits dietary requirements, travel time and past preferences.
In each case, the agent may narrow the market before the consumer sees most of the available choices. That does not remove marketing influence, but it can move part of the contest upstream. Brands may increasingly compete to be included in the set of options an agent considers credible, available and relevant enough to recommend.
This is still an emerging behavior, not a replacement for search, social or ecommerce. But the architecture matters. Once an agent can act across services, the customer journey is no longer limited to a person moving manually from one interface to another.
Marketing to an agent may look different from marketing to a person
People respond to creative, emotion, familiarity, social proof and brand meaning. An AI agent evaluating options may depend more heavily on whether it can reliably access and interpret information about the product or service.
That puts practical infrastructure into the marketing conversation. Accurate product feeds, structured information, current pricing, availability, review signals, trustworthy third-party references, APIs and reliable websites could all shape whether an agent can understand what a brand offers.
The point is not to invent a new discipline called “AI agent SEO” before the market exists. It is to recognize that digital discoverability may gain another layer. A brand can be memorable to people and still be difficult for software to evaluate if its information is incomplete, inconsistent or inaccessible.
The reverse is also true. A technically accessible product with weak reputation or little consumer preference is not automatically going to win. Agent-mediated discovery would likely reward a combination of machine-readable facts and the same trust signals marketers already work to build.
Brand still matters even if the agent handles the transaction
The strongest counterargument to an agent-controlled funnel is simple: people often know what they want before they ask software to help them get it.
“Book me a Marriott” is different from “find me a hotel.” “Buy me another pair of Nike running shoes” is different from “compare running shoes.” “Find me another Patagonia jacket” already contains a brand decision before the agent starts evaluating anything.
That is why brand building remains important. Advertising, social media, creator campaigns, sponsorships, customer experience and word of mouth can all influence the instruction a consumer gives an agent in the first place.
The agent may handle evaluation and execution, but human preference can still define the boundaries of the task. For marketers, that suggests a split rather than a replacement: brand creates preference, while data and digital infrastructure help an agent act on that preference.
Meta could end up controlling another layer of consumer intent
Muse is particularly relevant because Meta already sits across major discovery and communication surfaces through Instagram, Facebook and WhatsApp. A personal agent adds the possibility of moving from discovery and conversation into action.
That creates an unresolved strategic tension for Meta’s advertising business. If agents complete more tasks directly, some journeys could involve fewer traditional clicks and page visits. At the same time, richer agent context could make consumer intent much more detailed than a single search query or social interaction.
A search for “hotels Tokyo” reveals a narrow slice of intent. A persistent agent could potentially know the user’s dates, budget, previous trips, loyalty memberships, dietary needs, preferred neighborhoods and who is traveling with them, provided the user has shared that information and granted permission for it to be used.
That context could be commercially valuable, but it also intensifies privacy questions. CNBC reported that Meta is exploring possible commerce-related monetization around agent transactions, although the company has not settled on a concrete model. Meta executive Alexandr Wang told CNBC, “We think the commerce business model is potentially really interesting for this product.”
What Meta has not confirmed is an advertising model inside Muse. Marketers should treat that as an open question rather than assume ads are coming to the product.
The next funnel may have two audiences
The most useful way to think about personal AI agents is not that they eliminate the marketing funnel. They could split part of it into two audiences.
The human layer is still about awareness, affinity, trust, aspiration and preference. People form opinions about brands through culture, creative work, experience and recommendation.
The agent layer is more operational. It may need to understand what a product is, whether it is available, whether the price fits the request, whether reviews are credible, whether the service can be accessed, and whether a transaction can actually be completed.
That makes the future marketing question broader than “Can customers find us?” Marketers may also need to ask whether the systems acting for those customers can understand, evaluate and transact with the brand.
Meta’s Muse does not prove that this model will become mainstream. It does, however, make the possibility more concrete. The next battleground may not only be convincing the customer. It may also be making sure the customer’s AI can confidently put your brand into consideration.
