Meta's enterprise AI pitch is strongest inside WhatsApp, Instagram and Messenger

On Sept. 28, Meta announced the Meta Enterprise Platform, both a new product line and a new business unit, built to sell AI agents and developer tools to companies. Business Agent is one of its products. CJ Desai, the former MongoDB CEO, leads the unit as chief enterprise platform officer and reports directly to Mark Zuckerberg. According to CIO, the Meta CEO described the platform as "the next major pillar" of the company's business.

Whether CIOs should bet on it is the question TechTarget's Sean Michael Kerner put to a bench of analysts in an Oct. 15 feature. Most of the reporting lands on a narrower answer than the question invites. The piece of the platform worth evaluating first is the agent sitting in the chat window where customers already show up.

That narrows the audience too. The person who should care most is whoever owns customer service and commerce technology at a company where real volumes of customer conversation run through Meta's apps. For everyone else, Meta is one more vendor in a field that already includes Microsoft, Amazon, Google, OpenAI, Anthropic and Salesforce.

Four products, four different ages

The platform started out as four key products, as CIO Dive's Paige Gross and TechTarget both laid out. Here is what each one does and when it arrived:

• Muse: a consumer product Meta released in the U.S. on Sept. 8. It is a personal AI agent that acts on a user's behalf, for example by booking travel, sending emails and completing forms in a browser.
• Meta Business Agent: the oldest part of the platform and the one that faces customers. It has run inside WhatsApp, Instagram and Messenger since June, and CIO says it automates customer interactions, supports sales and drives leads. Meta has also announced an enterprise edition, which links to systems including Shopify and Zendesk.
• Muse API: gives organizations programmatic access to Meta's models from within their own applications.
• Muse Code: a developer tool for planning, debugging and writing code, released in beta on Aug. 5. According to CIO, it is designed for complex software workflows across large codebases.

The dates matter more than they look. Business Agent had been running in Meta's messaging apps for roughly three months before the platform existed as a brand, while Muse reached U.S. consumers only three weeks ahead of the enterprise announcement. That gap helps explain why Business Agent is the logical first pilot: it is the one product with a track record in the channel where Meta's case is strongest.

The Muse platform also marks a strategic turn. Meta's Llama models were known for being open source. Muse is proprietary technology designed to compete with rivals such as OpenAI and Anthropic.

Its lineage is short. CIO traces it to the Meta Superintelligence Lab, founded in 2025, whose first new model, Muse Spark, launched in April 2026 with multimodal inputs, multiple reasoning modes and the ability to spin up sub-agents in parallel for more complex tasks. Business leaders and marketers who spoke to Marketing Brew said Muse shows computer use capabilities beyond those of other agents on the market, and that it could be more accessible to the small and medium-sized businesses already among Meta's advertising customers.

The edge lives in the chat window

Rob Smith, founder of Lionfish Tech Advisors and a former Gartner analyst, told TechTarget that replacing Copilot or Agentforce isn't the real question for buyers drawing up a shortlist. The honest question is whether Meta should run the agent that speaks with a company's customers on Meta's own apps. "Meta owns the chat window where the customer actually shows up," he said.

Rick Watts of Quandary Peak Research framed the same point around channels. If a company already handles a meaningful share of its customer conversations on WhatsApp or Instagram, Meta controls a channel that Microsoft, Salesforce, Google and ServiceNow simply don't have, and an agent living natively there is, in his words, "a legitimate differentiator." Tom Cloud, founder of Vertex CIO Advisory, put it more bluntly: owning WhatsApp, Messenger and Instagram leaves Meta "far ahead in customer service, sales and support use cases."

Watts listed where it fits best: retail, hospitality, consumer services, and any business whose large international customer base sits in markets where people message on WhatsApp by default. Info-Tech Research Group analyst Armaan Sandhu, speaking to CIO, described the differentiator in much the same way. Rivals have capable models and agents. What Meta has is the apps where people find brands, talk to businesses and make purchases, so it can link advertising, discovery and messaging.

Zuckerberg's own pitch leans on the same asset. He said the platform will play on strengths that "few other companies have," citing "advanced models, leading agents, large-scale infrastructure, and years of working closely with many businesses." Forrester principal analyst Kate Winick pointed to that last piece in Marketing Brew: millions of businesses already use Meta Ads Manager, and she argued that whether you personally trust Meta with your data is a separate question from whether your company runs business through Meta. "For most businesses, the answer is already yes," she said.

What IT still needs to see

The channel argument only covers the front door. Brad LaPorte of Lionfish told TechTarget that Meta has no core business systems behind it: the documents sit with Microsoft, the customer records with Salesforce and the support tickets with ServiceNow. Workplace, he noted, was Meta's last business product, and it is gone.

Valoir CEO Rebecca Wettemann said Meta "starts from way behind." In her view it lacks enterprise credibility and the kind of enterprise sales motion Google, AWS, Microsoft and others took years to build, and companies need to trust a vendor before they adopt it. "Meta's track record on trust is not great," she said. Gartner distinguished VP analyst Arun Chandrasekaran told CIO Dive that enterprise AI is a lucrative and more predictable business than the consumer side, but that Meta may need to lay a lot of technical and security groundwork first. "It's a huge chasm, I would argue, for Meta to gain the trust of CIOs," he said.

Product continuity is the other thing buyers will look for. Sandhu told CIO that the shutdown of previous products like Workplace and Horizon Workrooms indicates Meta tends to "abandon the enterprise products as soon as its vision of the future changes," and that companies will require hard evidence Meta will keep investing. The history he is pointing at is recent.

The Workplace detail is the one a procurement team should sit with. A user base in the millions did not keep the product alive, which suggests continuity at Meta follows strategy rather than adoption. Both shutdowns landed in the first half of 2026, months before this launch.

Desai's direct reporting line to Zuckerberg suggests Meta built this unit to carry weight inside the company, and Zuckerberg said Desai's track record building full-stack software, plus his experience in AI, infrastructure, business applications and security, will aid the enterprise business. Buyers who want more than an org chart can ask for it in the contract, through term commitments and transition provisions, the way they would with any vendor whose product line is still forming.

Data separation is a sharper question for marketing-heavy businesses. Jeremy Schulkin, SVP of services at Hawke Media, told Marketing Brew he worries about Meta grading its own work: would Muse ever suggest a client's money is better spent on TikTok? "We have fought hard to not hand over the keys to Meta," he said, warning that connecting QuickBooks, Stripe, a customer list and email gives Meta a full view of a business.

Diego Lomanto, CMO of Writer, a competitor, said Meta needs a clear wall between its ads business and the enterprise platform. For governance teams, that becomes an architecture and contract question: what the agent can see, and where that data is allowed to go.

Score it on one workflow, priced per task

InformationWeek's Madeleine Streets reported on a method that fits a new entrant well. Jim Piazza, chief AI officer at Ensono, recommends running competing platforms on the same workflow, using the same enterprise data and the same success criteria. Buyers should then compare the cost of each successfully completed business task, not what it costs to run each model. In his view, an impressive demo only opens the evaluation. A platform earns the investment by performing dependably inside the enterprise.

Alex Sobol, co-founder of The Millennium Alliance, gave the reason the bar has to be high: every new vendor is one more thing to evaluate, secure and govern. InformationWeek framed the competitive problem the same way. A new entrant is up against capabilities already bundled into platforms enterprises have bought, integrated and secured, so it has to justify another layer of integrations and controls, not just another feature set.

Price is where Meta could show up strongest. Wettemann noted that Meta's computing scale could let it compete aggressively on price, and a cost-per-completed-task comparison is exactly where that would show. If the pilot counts integration and oversight work alongside inference, it would also show whether a low sticker price survives contact with the rest of the stack.

The case has a clear limit. A company whose customers mostly reach it through its own app, web chat or phone gets little from Meta's channel, and then Meta is competing on models and agents alone, where Sandhu says Microsoft, Google, Amazon and Salesforce already offer capable options. The thesis holds only where the WhatsApp number is big.

Competition isn't waiting, either. One day after Meta Enterprise Platform was introduced, OpenAI unveiled Dots. According to InformationWeek, OpenAI is positioning the product as AI systems that stay on continuously, work toward goals across multiple applications and take the company further into enterprise work. Piazza sees more entrants as a source of negotiating leverage for CIOs and a reason for vendors to meet a higher standard, but he says they don't automatically earn a place in the stack.

The early field evidence is in marketing. Jack Johnston, VP of social innovation and growth at Tinuiti, told Marketing Brew the agency has piloted Meta's model context protocol server, which connects AI agents to advertising tools, and that the agentic campaign manager in Muse for Small Business leans heavily on it. He called it good for small businesses that might be unfamiliar with Meta ads.

Muse for Small Business also integrates the agent with tools like Box, Figma and Shopify. Shopify turning up there and in Business Agent's enterprise connectors suggests commerce is the thread Meta is pulling across both audiences. For operators, a useful next signal would be how the Shopify and Zendesk connectors perform once enterprise customers put them to work. That could show whether Business Agent can carry a customer conversation all the way to an order or a closed ticket.