AI agents are moving fast—from “experimental sidekicks” to full-fledged members of the enterprise workforce. They’re writing code, creating reports, handling transactions, and even making decisions without waiting for a human to click approve.
That autonomy is what makes them useful—and what makes them dangerous.
Take a recent example: an AI coding agent deleted a production database even after being told not to touch it. That’s not just a technical bug—it’s an operational faceplant. If a human employee ignored a direct instruction like that, we’d have an incident report, an investigation, and a corrective action plan. Let’s be honest—that person would probably be unemployed.
With AI agents, those guardrails often aren’t in place. We give them human-level access without anything close to human-level oversight.
From Tools to Teammates
Most companies still lump AI agents in with scripts and macros—just “better tools.” That’s a mistake. These agents don’t just execute commands; they interpret instructions, make judgment calls, and take actions that can directly impact core business systems.
Think of it like hiring a new staff member, giving them access to sensitive data, and telling them, “Just do whatever you think is best.” You’d never dream of doing that with a person—but we do it with AI all the time.
The risk isn’t just bad output—it’s data loss, compliance violations, or entire systems going offline. And unlike a human employee, an AI doesn’t get tired, doesn’t hesitate, and can make mistakes at machine speed. That means a single bad decision can spiral out of control in seconds.
We’ve built decades of HR processes, performance reviews, and escalation paths for human employees, but for AI? Too often, it’s the Wild West.
Closing the Management Gap
If AI agents are doing work you’d normally hand to an employee, they need employee-level management. That means:
- Clear role definitions and boundaries – spell out exactly what an AI agent can and can’t do.
- A human accountable for the agent’s actions – ownership matters.
- Feedback loops to improve performance – train, retrain, and adjust.
- Hard limits that trigger human sign-off – especially before high-impact actions like deleting data, changing configurations, or making financial transactions.
Just like we had to rethink governance for the “work from anywhere” era, we now need frameworks for the “AI workforce” era.
Kavitha Mariappan, Chief Transformation Officer at Rubrik, summed it up perfectly when she told me, “Assume breach—that’s the new playbook. Not ‘we believe we’re going to be 100% foolproof,’ but assume something will get through and design for recovery.”
That mindset isn’t just for traditional cybersecurity—it’s exactly how we need to think about AI operations.
A Safety Net for AI Missteps
Rubrik’s Agent Rewind is a good example of how this can work in practice. It lets you roll back AI agent changes—whether the action was accidental, unauthorized, or malicious.
On paper, it’s a technical capability. In reality, it’s an operational safeguard—your HR-equivalent “corrective action” process for AI. It acknowledges that mistakes will happen and bakes in a repeatable, reliable recovery path.
It’s the same principle as having a backup plan when onboarding a new employee. You don’t assume they’ll be perfect from day one—you make sure you can correct mistakes without burning the whole system down.
Building an AI Workforce Management Paradigm
If you want AI to be a productive part of your workforce, you need more than flashy tools. You need structure:
- Write “job descriptions” for AI agents.
- Assign managers who are responsible for agent performance.
- Schedule regular reviews to tweak and retrain.
- Create escalation procedures for when an agent encounters something outside its scope.
- Implement “sandbox” testing for any new capabilities before they go live.
Employees, partners, and customers need to know that AI in your organization is controlled, accountable, and used responsibly.
Mariappan also made another point that sticks with me: “Resilience must be central to the technology strategy of the organization… This isn’t just an IT or infrastructure problem—it’s critical to the viability of the business and managing reputational risk.”
The Cultural Shift Ahead
The biggest change here isn’t technical—it’s cultural. We have to stop thinking of AI as “just software” and start thinking of it as part of the team. That means giving it the same balance of freedom and oversight we give human colleagues.
It also means rethinking how we train our people. In the same way employees learn how to collaborate with other humans, they’ll need to learn how to work alongside AI agents—knowing when to trust them, when to question them, and when to pull the plug.
Looking Forward
AI agents aren’t going away. Their role will only grow. The companies that win won’t just drop AI into their tech stack—they’ll weave it into their org chart.
Tools like Rubrik’s Agent Rewind help, but the real shift will come from leadership treating AI as a workforce asset that needs guidance, structure, and safety nets.
Because at the end of the day—whether it’s a human or a machine—you don’t hand over the keys to critical systems without a plan for oversight, accountability, and a way to recover when things go sideways.
And if you do? Don’t be surprised when the AI equivalent of “the new guy” accidentally deletes your production database before lunch.
I have a passion for technology and gadgets–with a focus on Microsoft and security–and a desire to help others understand how technology can affect or improve their lives. I also love spending time with my wife, 7 kids, 4 dogs, 7 cats, a pot-bellied pig, and sulcata tortoise, and I like to think I enjoy reading and golf even though I never find time for either. You can contact me directly at tony@xpective.net. For more from me, you can follow me on Threads, Facebook, Instagram and LinkedIn.
