Meta’s Muse Agent Sparks Cautious Optimism Across Industries

Meta’s Muse, an agentic AI shopping assistant, is now being discussed on earnings calls at dozens of public companies. The tool lets users issue natural‑language commands to browse, compare and purchase items, and Meta says more than 10 million North‑American users have activated it, with over a quarter using it for shopping – a figure that exceeds the company’s own expectations.

---

What is Meta’s Muse and why is it in the news?

Muse is Meta’s answer to the growing class of AI agents that can act on behalf of a user online. Unlike a simple chatbot, Muse can navigate e‑commerce sites, apply discounts, and complete checkout flows without further human input. According to the WIRED story, the assistant has become a recurring topic on corporate earnings calls because investors worry it could reshape brand‑loyalty dynamics and squeeze margins if it drives aggressive price comparison.

---

How are businesses currently experiencing Muse?

Executives from roughly 50 firms – ranging from food delivery (DoorDash) to telecom (T‑Mobile) and hospitality (Hilton) – all reported that Muse‑driven transactions are still "insignificant" or "very, very small". DoorDash said orders from agents are in the "hundreds," while Ally Financial called the activity "quite low". Even companies that have partnered with Meta’s retail network – Walmart, Best Buy, Gap – have not yet released post‑launch data. The consensus is that adoption is in an early‑stage, with most revenue still coming from traditional channels.

---

What are the potential risks and opportunities for companies?

Wall Street analysts fear that proactive agents could erode brand loyalty, trigger price wars, and cut profit margins. At the same time, many executives argue that product quality, service, and trust will continue to matter, even when a bot places the order. Security is another concern: Bank of America recently warned customers not to share login credentials with AI assistants, stressing that users remain liable for any activity performed with those credentials.

Meta plans to charge businesses a transaction fee for purchases completed through Muse, and CEO Mark Zuckerberg suggested that this revenue stream could eventually outpace Meta’s advertising business. For now, however, the fee structure has not been disclosed, and several large retailers – Amazon, for example – have taken a defensive stance, temporarily blocking Muse while they develop their own agent solutions.

---

How are firms adapting their digital presence for agents?

Even with low current volumes, companies are already tweaking their websites to be more agent‑friendly. Chipotle’s chief digital officer noted that agents stumble over loyalty‑discount logic and after‑hours ordering, prompting the chain to clean up its checkout code. Costco and Extra Space Storage reported rising sales from chat‑based agents like Gemini and ChatGPT, and they intend to apply similar optimizations for Muse.

The effort typically involves simplifying page structure, providing clear API‑style data endpoints, and ensuring that discount rules are machine‑readable. These changes benefit both human users and any future AI agents that can parse the site more reliably.

---

What does this mean for developers and creators?

For developers, Muse is a reminder that AI agents will soon be a first‑class client of web services. Building robust, well‑documented APIs and exposing structured product data (prices, SKUs, inventory) will make it easier for agents to complete transactions without errors. Creators of AI‑driven experiences can also experiment with agentic workflows in a sandbox environment to see how their content performs when a bot, rather than a human, is the end‑user.

---

Exploring agentic AI with Grafy

Grafy offers several tools that let you prototype and test agent‑like interactions without writing a full backend. The Grafy Bots app lets you create embeddable support bots that can be trained on your product catalog, giving you a low‑risk way to see how an autonomous assistant would handle browsing and checkout flows. If you prefer a more conversational sandbox, the Grafy Chat interface lets you talk to multiple models side‑by‑side, letting you experiment with prompt‑driven shopping tasks and observe how different models interpret product queries.

For teams that need to audit existing documentation or research the technical details of agentic commerce, Grafy Reader can ingest PDFs, earnings call transcripts, and white‑papers, then let you ask questions across the whole corpus. Using these apps, developers can simulate Muse‑style interactions, identify friction points in their own checkout pipelines, and iterate quickly before a real‑world agent arrives.

---

Frequently asked questions

How many users does Meta say have activated Muse?

Meta disclosed on its earnings call that more than 10 million people in the U.S. and Canada have activated Muse, with over 25 percent using it primarily for shopping.

Are companies seeing large revenue shifts because of Muse?

So far, the data shared by executives indicates that Muse‑driven sales are "insignificant" or in the low hundreds for most firms. The impact on overall revenue remains minimal at this stage.

Can I test a Muse‑like agent on Grafy today?

Yes. Using Grafy Bots you can build a simple shopping assistant that pulls product data from a spreadsheet or API, and with Grafy Chat you can experiment with natural‑language prompts that mimic Muse’s behavior.

---

Source: WIRED – Meta’s Muse Could Upend All Kinds of Businesses. It Hasn’t Yet

---

Source: WIRED — AI — Meta’s Muse Could Upend All Kinds of Businesses. It Hasn’t Yet