Enterprise AI agents, the software that plans and executes multi-step work across business applications, are a new and growing feature of business tech. But it’s already clear that we can’t let old tools and practices dictate how we think about and use this revolutionary new technology.
Especially when it comes to the risk and opportunity management of agents, the kind of agent governance we’re focused on at Barndoor.
The identity trap
Start with security. Lots of incumbent software companies talk about “agentic identity,” as if agents were no different from humans, and everything simply needs Identity Management. Besides making the error of anthropomorphizing AI software, this misses the point that identity is just one of many components of a secure and productive human-agent workplace.
What governance requires
There are identifiers for agents, but there are also things like the context the agent has at any given moment, what they are trying to do. There’s how the agent will complete the task, what the whole thing will cost, and how it matters in relationship with human workers, teams, and businesses. There is how this is observed, monitored, and reviewed. Different contexts have different rules, and they all matter.
Agents need their own form of governance, not something that was built for humans. And by the way, humans need more tools and capabilities to get maximum value from agents. That’s another aspect of agent governance.
Credentials are just the starting point
Agent governance is holistic; it covers the contexts of credentials for access and interaction with both LLMs and MCP servers; it covers credentialing for interaction with other agents, as well as humans. It’s a key element in determining the cost and benefits of the system (something else you’ll find at Barndoor.)
Compare that to humans. Human governance is established by everything from employment contracts and background checks, to job titles, reputation, and years of experience.
Agents don’t have that kind of governance, and that is your opportunity. Companies need agent governance systems that lean into operational activities, blending security with human and agent workflows, specific purpose with larger mission, rules and permissions with costs and performance. We’re focused on the creation of agents formed to the right purpose, with the narrowest possible tools for the best outcome, that people and systems can monitor and govern to ensure the right performance and cost.
The legacy IAM pitch is dated
This opportunity should be clear, but it’s often confused by old language and standing customs. Legacy security companies, in particular, don’t want to admit how much times have changed, and how needs have changed with them. So they try to sell “comprehensive” or “complete” identity management across humans, machines, and AI agents, as if in all three cases identity meant the same thing.
Both humans and cars need fuel, too, but that doesn’t mean people should drink gasoline, or you should expect 30 miles to a burrito in your Hyundai.If someone tells you that agents just need another version of IAM, ask yourself if they’re the ones who sold you that IAM system.
Then consider a few of the ways that agents need governance, because they’re different from humans. People are hired not simply to work, but to learn a corporate culture, find their best role for themselves and the company, innovate and grow; agents carry out tasks. Humans have careers, and agents have lifecycles – those stories about the CEO who started in the mailroom won’t be happening with agents. Humans are restricted through denying permissions; agent software is often adjusted when a violation takes place, so the system can work better. Humans have trust based on accountability, while agents have monitoring and verification.
Built for the agent world
Barndoor was created for the agent world, so we’re not trying to protect legacy ideas about identity, observability, governance and accountability that don’t make sense in this new context. We appreciate and we support other company’s human and machine IAM, but we don’t confuse it with how you should treat agents.
Introducing AgentProfile
This is why Barndoor recently released AgentProfile, a way to connect and bind everything agent governance requires – its unique credentials, access scope, cost, spend, all in one place and separate from a human’s identity. My co-founder and Chief Product Officer Chris wrote about it here.
Humans have AI. Barndoor has your back, by giving you true visibility, control, and governance, across your agents.
