Major enterprise platforms are drawing a line against the flood of AI agents trying to access their systems. Slack, Workday, and LinkedIn are resisting efforts by customers to connect autonomous AI agents to their services, according to The Information.
The pushback signals a growing tension in enterprise software: companies want to deploy AI agents that can act on their behalf across tools, but the platforms hosting critical business data aren’t eager to open the gates.
Why This Matters
AI agents are autonomous software that can browse, click, fill forms, and make decisions without human input. They’re the hottest trend in enterprise AI right now. Every major AI lab (OpenAI, Anthropic, Google) is building agent capabilities. Startups are raising hundreds of millions to create agents that navigate existing software on behalf of workers.
But here’s the friction: these agents need access to the platforms where work actually happens. And those platforms have good reasons to be cautious:
- Security risks: AI agents acting autonomously could expose sensitive employee data, messages, or HR records
- API abuse: Agents hammering APIs at scale create load and scraping concerns that go beyond normal user behavior
- Business model threats: If an AI agent can extract value from a platform without users ever logging in, the platform loses engagement and leverage
- Liability questions: When an agent makes an error inside Workday’s HR system or sends a wrong message in Slack, who’s responsible?
The Bigger Picture
This resistance creates a real bottleneck for the “agentic AI” wave that companies like Microsoft, Google, and Salesforce are betting billions on. If the most popular enterprise tools won’t cooperate, agents can’t deliver on their promise of automating cross-platform workflows.
LinkedIn’s resistance is particularly notable given that its parent company Microsoft is simultaneously pushing Copilot agents across its own ecosystem. Slack, now owned by Salesforce (which has its own Agentforce platform), faces a similar internal contradiction.
Workday’s caution makes sense from a compliance angle: the platform holds some of the most sensitive employee data in any organization, from compensation to performance reviews.
What Comes Next
Expect this to become one of the defining battles of 2026 in enterprise AI. Platform companies will likely try to control agent access through their own sanctioned integrations and partnerships, rather than letting third-party agents roam freely.
For AI practitioners building agent-based workflows, the takeaway is practical: don’t assume your agents will get unfettered access to major enterprise platforms. Build with fallbacks, and watch for official agent APIs and partner programs that these companies will inevitably launch on their own terms.
The full story is available at The Information.