A landlord in California 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. That deadline doesn’t care whether the letter was supposed to come from a person or a piece of software. Either way, someone will answer for it.
I think about this a lot. At my company, Get Covered, we now run 16 AI agents across more than 3 million rental units. When we built them, the technical questions turned out to be the easier ones. The harder question, for each agent, was actually who it reports to.
Most companies are about to face that same question, and few are ready for it. Deloitte’s 2026 State of AI survey found that 23% of companies use agentic AI at least moderately today, and 74% expect to within two years.
Gartner, meanwhile, has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027.
Those numbers make a case for running agents the way you would run staff in a regulated job, with written rules, a defined scope, a manager and a regular review.
Here is what has worked for us, and what I would advise any decision-maker to focus on when it comes to scaling.
Start with work that already has a rulebook
The best early candidates for agents are tasks where the rules already exist somewhere outside anyone’s head: a statute, a contract clause, an insurance requirement in a vendor agreement.
Take vendor certificates of insurance. Across the industry, property teams spend roughly three hours per property every week chasing those documents by email. The work is tedious, however at the same time, it is completely specified.
Assignments like these give decision-makers two advantages. First, the agent has something objective to check its output against, and second, your team can audit the result without debating judgment calls.
McKinsey‘s most recent State of AI research found that the small group of companies seeing significant bottom-line returns from AI, about 6% of respondents, were nearly 3x as likely to have fundamentally redesigned their workflows.
In my experience, that redesign starts with writing the rule down. If a new hire couldn’t follow your description of a decision on the person’s first day, an agent won’t follow it either.
Set autonomy by workflow
Tech execs I speak with often ask whether an AI system is autonomous, as if that were a fixed property of the product. In practice, the same agent should run with different levels of freedom depending on what it touches.
At Get Covered, we configure each agent on different levels. For instance, a reminder to a vendor whose certificate expires next month can go out automatically. A response to a resident who has requested an emotional support animal never does.
A useful exercise for any leadership team is to list every workflow an agent will touch and ask one question about each: 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.
Give every agent an owner with a name
Earlier in my career I learned that every automated credit decision eventually had to be explained to a customer, a regulator, or both. The model did it was never an acceptable answer.
That still holds, except agents now act more often and across more systems than any underwriting model did.
So every agent at our company has a person ultimately responsible, the same way a team or a budget does. That person reviews its error log, decides when to widen or narrow its autonomy, and answers for it.
Ownership also heads off a quieter problem. When several teams launch agents independently, nobody sees where those agents contradict one other.
Where the constraint actually sits
In today’s marketplace, it’s inevitable that the models will keep improving. A company’s clarity about who decides what will not improve on its own. Gartner’s cancellation forecast reflects that: Most agent projects that fail will do so because nobody defined the value or controlled the risk, and both of those are management problems.
For decision-makers heading into 2027, my advice is to treat each agent the way you would treat a new employee in a regulated role. Write down its rules, limit its scope, and assign it a manager who checks its work and discusses its performance.
Companies that do this will be able to give agents more responsibility over time, because they will have the evidence to justify it. Businesses that skip it will spend 2027 explaining why their software did something nobody approved.
Rick Folgmann is chief operating officer at Get Covered, an insurance technology company that provides cost-effective property coverage with just a few clicks. He has more than 25 years of executive leadership experience in the financial and insurance industries.
Illustration: Dom Guzman
