Meta spent years building AI into Facebook, Instagram and WhatsApp for the billions of people who use those apps daily. Now it wants companies to pay for that same technology to run parts of their own business. On Monday, Meta launched Meta Enterprise Platform, bringing together its AI models, agents and developer tools for companies and outside developers to build on.
To run it, Meta hired away a sitting tech CEO. Chirantan "CJ" Desai left MongoDB, the database software company he had led for less than a year, to become Meta's Chief Enterprise Platform Officer, reporting directly to Mark Zuckerberg. The market reaction to losing him was immediate: MongoDB shares fell sharply on the news, dropping as much as 20% intraday and closing down more than 17%, while Meta's stock slipped about 4%.
Desai is not a random hire. Before MongoDB he spent nearly eight years at ServiceNow, helping grow that company from 1.5 billion dollars in revenue to more than 10 billion. Zuckerberg called the new division the next major pillar of Meta's business, putting it alongside its consumer apps and advertising business for the first time.
Meta is not alone in this push. One day after Meta's announcement, OpenAI launched Dots at its developer conference: always-on AI agents that run on dedicated cloud computers and keep working toward a goal around the clock instead of waiting for someone to type a question. Both moves point to the same pattern. The companies that built today's AI chatbots are racing to turn them into tools that operate inside real businesses, using agents as the entry point.
Meta has a case to make. Its consumer AI agent, Muse, has already topped 5 million downloads faster than ChatGPT or Claude did at the same stage, and briefly overtook ChatGPT as the top free app on Apple's store. That kind of consumer traction is exactly what Meta wants to translate into enterprise sales.
But enterprise software is sold differently than a free app. Companies that run their operations on Microsoft, Salesforce or ServiceNow already have AI features bundled into tools they have spent years integrating and securing. More than 60% of Fortune 500 companies now run at least 10,000 seats of Microsoft's Copilot alone, which means a new entrant is not pitching into empty space. It is pitching against tools already paid for and already trusted.
That makes Meta's own track record a liability it cannot talk its way around. The company shut down Workplace, its previous enterprise product, earlier this year after a ten-year run, closing a tool that had around 7 million paying users but had stopped growing since 2021. A new AI platform is a different product, but a company that walked away from enterprise customers before is asking those same customers to trust it again.
The stakes of that trust question are rising because of what these platforms are being asked to do. Giving an AI agent access to a company's systems is not like turning on a better spellchecker. Research from the Cloud Security Alliance found that roughly two out of three organizations have already suffered a security incident tied to an AI agent, and Gartner has forecast that by 2028, a quarter of enterprise security breaches will trace back to AI agents being misused or over-permissioned.
That is the real test facing Meta, OpenAI and every other company chasing this market. Winning a business customer now means proving not just that an agent works, but that the company behind it will still be around and accountable once that agent has the keys to the business.