Meta is moving its internal workplace communications from Google Chat to Slack, according to an internal memo from Alexandr Wang, Meta's AI chief. Wang said in the memo that Slack is the strongest platform available today for agents, citing its conversational interface, developer tooling and third-party integrations. The move is not a decision about chat in the traditional sense. It is a decision about AI agents and where they can operate most effectively inside a large technology company. Meta and Salesforce declined to comment on the report, while Google did not respond.
What makes Slack attractive to AI agents
AI agents are software systems that can carry out tasks on behalf of a user. They can summarise conversations, retrieve internal documents, file issue reports, update customer records, schedule meetings and interact with multiple external services. Unlike simple commands, agents often need to understand context, decide what to do, call an API and remember the result. That requires the messaging platform to be more than a chat window. It needs a predictable event model, flexible permissions, a searchable message history, support for long-running workflows and a broad library of connectors.
Slack has spent more than a decade turning itself into an ecosystem. Thousands of third-party applications are available in its directory, and developers can write custom integrations with its APIs. Slack's interface is persistent, organised around channels and threads, which gives agents a natural place to observe conversations and contribute without seeing every message in a company. Its workflow builder allows recurring tasks to be automated without heavy custom code. These features matter more when a company wants to embed AI into daily operations rather than have it live in a separate consumer app.
Meta's memo reportedly highlighted all three dimensions: the conversational interface that makes agents easy for employees to use, the developer toolkit that makes them easy to build, and the third-party integrations that connect them with existing business software. It is a reminder that AI agents are not standalone products. They are only as powerful as the environment in which they operate.
Google Chat and the shift in enterprise messaging
Google Chat is part of Google Workspace, one of the most widely used office platforms in the world. It is deeply tied to Gmail, Calendar, Drive, Meet and Google's broader cloud portfolio. Google has also introduced generative AI capabilities across Workspace, including writing assistance, email summaries and note-taking tools. Still, Meta's decision suggests that when an organisation must support thousands of developers and complex internal workflows, the ecosystem around a communication tool can outweigh the convenience of being already bundled with office software.
There are also practical reasons why Slack may feel more mature for agent developers. Slack has had a public API and an application directory since long before Google Chat became a standalone enterprise offering. Its application model separates channels from conversations, making it easier for an automated tool to post in the right place at the right time. It also has support for Slack Connect, which lets different companies collaborate in shared channels, a useful feature when teams inside a company work with outside vendors, contractors and partners. Google Chat supports external collaboration too, but its product has not historically been the first place developers build integrations.
The migration also matters for Google. Losing Meta as an internal user does not mean Google Chat loses a paying customer, but it is a symbolic loss. To have Meta, known for building its own infrastructure, choose another vendor's product signals to other companies. It could accelerate a broader trend in which enterprise chat procurement is led by AI strategy rather than by office suite convenience.
A major win for Salesforce and Slack
The switch is a substantial victory for Salesforce. Salesforce bought Slack in 2021 for 27.7 billion dollars, a deal that was questioned by investors who wondered whether Slack could survive in a world of Microsoft Teams and Google Chat. Since then, Salesforce has integrated Slack into its customer relationship platform and made it the surface for its own agent tools. Salesforce's Agentforce is designed to let companies build and deploy autonomous agents that work across sales, service, marketing and commerce. Slack is a natural delivery channel for those agents because it is where employees already communicate.
Meta and Salesforce did not comment. Google did not respond. The report did not disclose exact migration timelines or whether Meta is moving entirely off Google Chat. But a decision of this kind from Meta, one of the world's most valuable companies, will be studied closely. If Slack can handle Meta's size, the case for using it as a hub for AI agents becomes easier for other enterprises to accept. If Google responds by expanding Google Chat's own agent ecosystem, enterprise communication tools may begin to look even more alike.
Europe's opposite direction
Meta's move points in one direction: toward platforms with the largest number of integrations and the most flexible developer environment. But Europe's largest institutional technology buyers have spent the past two years moving in a different direction. They are not choosing tools with the strongest AI ecosystems. They are choosing tools that run on servers they control.
France has ordered government ministries off Windows and set a 2027 deadline to replace Microsoft Teams and Zoom for more than 2.5 million civil servants. The replacement is not a commercial collaboration platform but a sovereign stack built around the state's own infrastructure. Tchap, the French government's Matrix-based messenger, has more than 600,000 users and runs on certified sovereign infrastructure. The state also operates its own video service and webmail on the same foundations. This is a deliberate attempt to make digital administration independent of American cloud providers.
Germany has moved further in some regions. The northern state of Schleswig-Holstein last year moved 44,000 mailboxes and roughly 110 million emails off proprietary software. It reported savings of about 15 million euros in licence costs. Other German institutions are using openDesk, a sovereign suite developed by Germany's Centre for Digital Sovereignty, known as ZenDiS. Element, a company built around the open-source Matrix protocol, supplies the chat component for some of these installations. The German armed forces, the Bundeswehr, and the Swiss Post also run Matrix-based systems.
The sovereign stack dilemma
The pattern in Europe is not limited to France and Germany. The International Criminal Court runs a version of the same sovereign stack. Matrix, the open-source communication protocol used by Tchap, openDesk, the Bundeswehr and others, is reported to be in conversation with around 35 governments about open-source messaging systems. These are not low-level trials. They are production deployments in ministries, courts, defence organisations and public postal services.
The motivation behind those choices is different from Meta's. Public sector buyers in Europe are concerned about data protection, digital sovereignty, procurement control and dependence on non-European infrastructure. The U.S. CLOUD Act gives American authorities a legal route to demand data from U.S. companies, even when that data is stored in Europe. European governments do not want citizens' administrative records, court documents and security-related communications to fall under those rules. They also want to avoid being locked into licensing models that become more expensive over time.
The trade-off is that these sovereign platforms were designed to answer a political question, not an AI question. They are not built to match Slack or Microsoft Teams in the number of integrations, the maturity of the developer ecosystem or the readiness for autonomous agents. The EU's public sector may get more control over its data, but it may not get the same automatic benefit of the AI agent explosion. That gap is what makes the comparison between Meta's decision and Europe's trajectory so striking.
EU regulation makes exit easier
European policymakers are also changing the legal and economic conditions for switching. The EU Data Act applies to cloud services and to software-as-a-service providers, not just to physical infrastructure. It is designed to remove the obstacles that have made migration between platforms difficult, time-consuming and expensive.
The regulation says a provider may require no more than two months' notice when a customer asks to switch. The provider must then complete the transfer within 30 days. It must make it possible to export data in a structured, machine-readable format, including metadata and security settings. Until 12 January 2027, switching charges are capped at cost. After that date, providers cannot charge customers for switching at all.
This has an important consequence: the cost of leaving a major platform falls dramatically. If a government or company wants to move from Teams or Slack to an open-source sovereign suite, it will no longer have to pay an exorbitant exit fee at the end of a procurement contract. That makes open standards and self-hosted systems easier to adopt. It also means providers cannot simply keep customers by making migration too expensive.
However, low switching costs do not solve the agent gap. A government can move from Microsoft Teams to Matrix tomorrow and be free from commercial licences, but it will not automatically have tools like AI agents that can write code, organise government records and connect to thousands of SaaS products. European institutions might have to invest in open-source agent frameworks and train their own developers to build what Slack has accumulated over many years.
Control versus ecosystems
Meta's migration to Slack is a bet that internal workplace productivity will be shaped heavily by AI agents. Europe's sovereign stack is a bet that public data should stay close to the state. The two bets are not completely incompatible: open-source systems can add AI agents later. But right now they represent competing procurement philosophies. The platform with the best agents may not be the platform with the strongest guarantees of control, and the platform with the strongest control may not be the one where the next generation of agents is easiest to build.