79% of enterprises run AI agents in production, but only ~2% have identity governance for them
usehive · reddit · 2026-10-11
Two enterprise vendors shipped the same message within days: agents need guardrails, and almost nobody has them.
- RSA launched Agent ID: every agent call goes through a policy gateway with authorization down to the tool and argument level; wire transfers, restricted data, and payments require approval from a named human via a separate phishing-resistant channel, and each approval creates a record tied to the approver.
- SailPoint expanded its Agentic Fabric to hunt down shadow agents, applying runtime authorization and killing rogue agents on the spot while legit ones keep running.
- The key stat: SailPoint's own research finds 79% of enterprises already run AI agents in production, but only 2% have identity security tools designed to govern them.
The author argues the demand side is settled — identity vendors are building the contract layer for agents — and the open question is all the software with no identity team behind it. The poster's company HIVE builds permissions into the access layer from day one.
More from coding & agent
- One prompt to audit your docs: have an agent catch contradictions in pricing and API pages — gethackteam · 2026-10-11
- Card store owner builds MCP-powered Bloomberg terminal for trading cards with AI agents — VolumeIcy915 · 2026-10-11
- Two prompts + Claude made a Monet-style MV for Jay Chou's 'Qi Li Xiang' — AlchainHust · 2026-10-11
- Musk amplifies claim: AI firms that don't field agentic-AI deployment armies will lose — elonmusk · 2026-10-11
- Veteran engineer: today's coding agents outperform many developers I've worked with — gethackteam · 2026-10-11
- Dev testing Gemini Spark finds the agent can code via MCP but can't run the bots it writes — tristanbob · 2026-10-11