NoteTools & Agents

AI agents need employee badges too

Behind cybersecurity companies CrowdStrike and Okta jumping 20% and 28%-plus respectively after their earnings reports and hitting record highs is a new narrative put forward by Okta. The logic is that where companies once had to manage just one account per employee, a single employee now creates dozens to hundreds of agents, so those agents too become things to manage, each with its own identity and access rights. In the same vein, Jensen Huang reportedly said on NVIDIA's earnings call that the company has 40,000 employees but the AI agents it runs could number 400,000, and eventually 4 million; since agents run 24 hours a day with no vacations or working-hour limits, the number of things to manage grows exponentially. If the mood at the start of the year was that AI would cut headcount, the view has now shifted to demand for identity management exploding. This frame — the more agents there are, the more accounts, permissions and audit logs need managing — is useful as an opening for governance of in-house automation agents, and the 4 million figure is a secondhand quote, so it needs a caveat.

Key points

  • When one employee creates dozens to hundreds of agents, IAM is needed at the agent level
  • Jensen Huang: 40,000 employees vs 400,000–4 million agents (secondhand quote, needs checking)
  • Agents run 24 hours a day, 365 days a year, so what needs managing grows exponentially
  • May be an early sign of AI gains spreading from hardware to software

This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.

Same week's notes

Three concepts the term 'graph engineering' lumps togetherHermes, a personal multi-agent system split up by roleSutton's diagnosis: an LLM whose weights never change is not a mindKorea's independent AI model project: second-round results and an agent for every citizenFrom teleoperation to onboard vision, the evolution of robots that see the ball