It’s tempting to think of an AI agent as just another tool – but is it?
Tools are passive things. When you need it, you pick it up for action; when you’re done, it rests inactive.
But recently, I’ve had quite a few uncomfortable moments when I realized what AI had access to without my express permission. Unlike a tool, AI can act with or without my agency, or even without my awareness.
Further, these AI agents, unlike most tools, can be difficult to constrain or turn off. I’m a reasonably sophisticated AI user and it took me time and effort to regain control. What about the average Joe? Is AI really under their full control?
It’s more like an employee
Like people, AI agents have…agency. And that’s what worries me. As AI progresses, many companies automatically surrender access to their hard-won company and/or client data with little or no consideration for liability.
In fact, your agents will probably have access to data routinely collected and shared by the applications, such as Google or Microsoft, that you use every day.
That’s why I believe you should treat using an AI agent like hiring an employee. When you hire someone, you submit them to an onboarding process that includes determining what systems they have access to, what training they require, and a review of the company’s performance and behavioral expectations.
There may be seminars, NDAs and more, but whatever the pieces, the point of the whole is not only to ensure the right company fit, but to reduce liability exposure for the company and its clients.
If AI and its agents cover the role or functions of employees, doesn’t it make sense to apply similar levels of due diligence? Specifically, you need to know:
- What organizational data (emails, documents, financials, etc.) will you make available to the agent? What guardrails are in place to manage access and restrictions?
- What about your clients’ data? Is any of it accessible to your agents? For example, when you correspond through email, giving access to your inbox could expose its contents.
- Beyond access, how will this data be used? Will it be added to databases accessible outside your organization? Will it be sold to 3rd parties?
- What precautions will you enforce to ensure security and privacy? Do you have plans for security audits?
Risks and opportunities
As with any new technology, AI presents a host of risks and opportunities. Managing the former helps you seize the latter. When you treat new AI agent as you would a new employee, you reduce the odds of inadvertently exposing your company to risk.
If you have any questions or just need a sounding board for the more responsible use of AI, we are here to help.
With a real person!
