Hey everyone! Our latest report, Governing the Agentic Workplace, is out today. We surveyed an independent panel of 300 IT, engineering, security, and AI leaders about how their organizations build, run, and govern AI agents, and how their teams are changing along the way.
The short version is that agents are already in production almost everywhere, and the governance and staffing needed to run them safely are still catching up. A few findings that stood out to us:
- Agents are already in production. 89% of leaders have at least one agent running, and 23% run more than 20.
- Readiness is lagging. Only 8% are very confident their systems are ready for AI at scale.
- Workarounds are common. 20% have shared credentials insecurely to get an agent working.
- Builders are moving faster than policy. 16% know non-technical employees are building agents without approval.
When you build agents, how do they usually get access to data?
Through a central, governed data layer
Per-agent credentials or API keys
It depends on the project
Our agents don't touch production data (yet)
0
voters
If you've had to work around governance to get an agent running, or you're the one setting up the guardrails, tell us how it's going in the replies.
