Meta Hires MongoDB CEO, Amazon Adds Claude: DesignRush AI Roundup

Our analysts track the weekly developments reshaping AI and how it reaches users. Brands building AI products can partner with vetted AI companies to bring their ideas to life.

Four moves this week turned on the same question, which is whether an AI agent says who it is.

• Meta launched an Enterprise Platform, hiring MongoDB's CEO to run it and knocking about 18% off the database company's shares the same day.
• Microsoft rebuilt Copilot on three tabs, moving its agent features onto usage-based credits that admins can watch line by line.
• Amazon opened its seller backend to Claude, three days after blocking Meta's Muse from its storefront for browsing anonymously.
• PayPal added Muse checkout across its merchant network, on the same day that six major banks published trust principles that lead with transparency.

Model makers, cloud platforms, marketplaces, and payment networks are all chasing the same transaction layer, and all four hit the disclosure question on the way.

Our Take: Does an AI Agent Have to Say Who It Is?

An agent that will not name itself is one that nobody can be held responsible for.

We think the answer to the question above is yes, and the companies fighting over agent access have already settled it in practice.

Amazon blocked Muse because the agent browses without identifying itself and appears to hold customer credentials.

The banks put transparency first among five principles, asking that every party to a purchase know that an AI agent was involved.

Meta failed its own version of the test, routing Muse calls to human contractors without telling the people on either end.

Microsoft is the outlier, though its billing split is a disclosure problem, too, since a metered agent spends money unsupervised.

Access is going to the AI agents that can be named and audited, and the ones that stay hidden will keep getting locked out.

Meta Names a Database Veteran to Sell Its AI

Meta launched Meta Enterprise Platform on September 28, the company announced.

CEO Mark Zuckerberg called it "the next major pillar of our business."

The platform bundles Muse, Meta Business Agent, Muse API, and Muse Code into a single business-facing offering.

Chirantan "CJ" Desai runs it as chief enterprise platform officer, reporting directly to Zuckerberg.

He left MongoDB after less than a year as CEO and president.

Before MongoDB, he led product and engineering at Cloudflare and spent nearly eight years at ServiceNow.

The market punished both sides of the hire.

MongoDB shares closed down about 20%, having fallen as much as 24% earlier in the day, while Meta's stock dropped roughly 4%.

The hire lands next to an awkward week for Muse itself. Meta spent $31.1 billion on capital expenditures in the second quarter alone.

It guided full-year 2026 capital expenditures to a range of $130 billion to $145 billion, driven by servers, data centers, and network infrastructure.

On September 22, Meta quietly routed some Muse phone calls to human contractors in an internal staff test, then rolled the feature back.

One employee learned only from the transcript that a contractor had spoken in his name while negotiating his cable bill.

The contractor made a racist remark on the call, in a conversation believed to be handled by an AI agent.

A vice president in Meta's Superintelligence Labs unit said that starting the test without telling users a human might take the call "was a miss."

A company guiding to $145 billion in capital spending still needed people on the phone to make its flagship agent close a call.

It shipped the feature without telling anyone the callers were human, which is close to the reason that Amazon locked Muse out of its storefront.

Meta hired a database executive to run the product line, which says the fight is over enterprise contracts and not model benchmarks.

Microsoft Splits Copilot Into Two Bills

Microsoft introduced a redesigned Copilot app on September 25, organized around three tabs called Home, Code, and Autopilot.

Autopilot, previously called Scout, is a persistent agent with its own identity, memory, and workspace inside a company's tenant.

It keeps working after a user logs off.

Home and Code roll out through Microsoft's Frontier program in the coming weeks, and Autopilot entered private preview at the end of September.

The pricing changed alongside the redesign, with a standard $30-per-user license now covering chat and Office app integration.

Cowork, Code, and Autopilot run on usage-based Copilot Credits, and admins set budgets and see what each user has spent.

Fewer than 7% of Microsoft's more than 450 million commercial Office 365 seats carry a Copilot license, CNBC reported.

Code lets non-developers build apps, dashboards, and automations by describing them, which is vibe coding with a Microsoft tenant around it.

Jacob Andreou, Microsoft's EVP for Copilot, conceded that adding coding to the app is an instance of Microsoft playing catch-up.

Microsoft built two billing systems because two different buyers sign off on them.

A $30 seat license sells to a manager, while a metered agent that works all night sells to whoever reviews the invoice afterward.

The credits dashboard is a disclosure tool as much as a finance one, since it tells an admin what an agent did while nobody was looking.

Amazon Opens Its Seller Backend to Claude

Amazon introduced a Selling Partner plugin at its Accelerate conference on September 23.

It connects sellers' inventory, pricing, and listings data to Anthropic's Claude or Amazon's Quick Assistant.

The Claude version is in beta for U.S. sellers on the Professional plan, and every action still needs seller approval.

Amazon said that Seller Assistant already runs on Claude models through Bedrock.

Anthropic Field CTO Eric Burns describes sellers as making dozens of pricing, inventory, and ads decisions a day.

Amazon is far less welcoming to agents it did not approve. Three days earlier, it blocked Meta's Muse from shopping on Amazon.com.

The company stated that Muse conceals its identity as it browses and appears to capture and store customer credentials.

The split makes a certain kind of sense.

A named agent can be rate-limited, audited, and cut off as soon as it misbehaves, which makes letting it in a reversible decision.

PayPal Wires Up Muse as Six Banks Publish Agent Rules

Six banks, including Bank of America, Capital One, and NatWest, published a joint set of principles for agentic commerce on September 22.

Transparency leads the five principles, followed by safety, privacy and data, choice, and interoperability. None of it is binding.

The banks want the merchant, the card issuer, and the shopper to know when an agent is involved and who it acts for.

On the same day, PayPal announced that its checkout will let customers shop and pay through their Muse agents at merchants worldwide.

Wall Street sorted the winners quickly. Shopify closed up 7.9% on September 22 after its own Muse deal, while PayPal gained 1.3%.

Booking Holdings fell about 5% the next day, after Expedia announced its own Muse partnership.

Investors are pricing in a world where an AI agent does the comparison shopping and the aggregator never gets the visit.

Principles can define the rules of agentic commerce.

Live integrations are already setting them, and transparency is the only principle a merchant can verify without the banks.

Three Checks Before an AI Agent Gets Access

Three things turn an agent's convenience into cost, exposure, or lost control:

• Price the integration alongside the benchmark. A cheaper model with no clean API costs more to deploy than a pricier one that plugs straight in.
• Audit every AI subscription for metered billing. An agent that runs overnight can turn a predictable flat fee into an invoice nobody approved.
• Make every vendor say who handles the work. A tool that quietly routes tasks to outside agents or contractors leaves you guessing whose privacy policy holds your data.

Once an agent can spend money, reach business data, or change how a service gets billed, it deserves the same scrutiny as any other vendor integration.

The AI agents worth letting in are the ones that tell you they're there.

These leading AI companies help brands work through vendor pricing and integration decisions before they show up in next quarter's budget.