Treat Your AI Software Like a New Hire With a Manager

Companies adding AI agents, software that carries out tasks on its own, should manage each one like a new hire in a regulated job. That is the advice from Rick Folgmann, chief operating officer at Get Covered, an insurance technology company. His company runs 16 AI agents across more than 3 million rental units. In his view, the technical questions were the easier ones. The harder one was who each agent reports to.

Why the stakes are real

A California landlord has 21 days after a resident moves out to return the security deposit or send an itemized list of deductions. Miss it in bad faith, and the penalty can reach twice the deposit. The deadline does not care whether a person or software sent the letter. Someone still answers for it.

Many companies are heading into this. A Deloitte survey found 23% of companies use agentic AI at least moderately today, and 74% expect to within two years. Gartner has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027.

Where to start

The best first jobs are ones where the rules already exist in writing, such as a law, a contract clause or an insurance requirement. One example is checking vendors' certificates of insurance. Property teams spend roughly three hours per property every week chasing those documents by email. The work is dull but fully specified.

• The agent has something objective to check its work against.
• Your team can audit results without arguing over judgment calls.
• If a new hire could not follow your written rule on day one, an agent will not either.

McKinsey research points the same way. About 6% of companies saw significant bottom-line returns from AI, and they were nearly 3x as likely to have fundamentally redesigned their workflows.

Give freedom in steps

Folgmann argues that autonomy is not a fixed trait of a product. The same agent can get different levels of freedom depending on what it touches. At Get Covered, a reminder to a vendor whose certificate expires next month can go out automatically. A reply to a resident asking for an emotional support animal never does.

His test for each workflow is simple. If this action is wrong, how long until we find out, and what will it have cost by then? Quick detection and low cost earn an agent more room.

Every agent needs an owner

Each agent at Get Covered has a person ultimately responsible, the way a team or a budget does. That person reviews its error log, decides when to widen or narrow its freedom, and answers for it. Ownership also helps when several teams launch agents on their own, because someone can spot where those agents contradict each other.

Folgmann says most failed agent projects will fail because nobody defined the value or controlled the risk. Those are management problems, not software problems. Companies that keep records of how agents perform will have the evidence to give them more responsibility later.