By Rick Folgmann
In California, property managers operate under a strict and unforgiving legal timeline. A landlord has precisely 21 days after a resident vacates a property to either return the security deposit in full or provide a comprehensive, itemized list of deductions. Failing to meet this deadline due to negligence or bad faith can result in severe financial penalties, sometimes reaching up to twice the original deposit amount.
The state statutory deadline does not account for administrative oversight, nor does it care whether the required notification was supposed to be generated by a human employee or a piece of software. Ultimately, accountability remains with the enterprise. Either way, someone must answer for the outcome when things go wrong.
This high-stakes regulatory reality is a constant consideration at Get Covered, where sixteen specialized artificial intelligence agents now manage operations across more than 3 million rental units. When the engineering teams initially designed and deployed these digital agents, the complex technical architecture turned out to be the straightforward part of the equation. The far more difficult challenge for each individual agent was determining where it fit within the organizational chart and, crucially, whom it ultimately reports to.
Most modern enterprises are rapidly approaching that exact organizational question, and very few are genuinely prepared to answer it. According to Deloitte’s 2026 State of AI survey, roughly 23% of companies currently utilize agentic artificial intelligence at least moderately in their daily operations, while an overwhelming 74% expect to integrate these autonomous systems within the next two years.
Simultaneously, prominent market research firms are sounding notes of caution regarding rushed deployments. Gartner has published predictions forecasting that more than 40% of all agentic AI enterprise projects will be officially canceled by the end of 2027.
These contrasting industry metrics underscore a compelling case for operational strategy. Organizations must begin running artificial intelligence agents with the same rigorous oversight, clearly written operational rules, defined scopes of responsibility, designated managers, and regular performance reviews that they would apply to human staff operating in heavily regulated professional environments.
For decision-makers navigating the rapid expansion of digital automation, establishing clear management frameworks is no longer optional—it is a vital prerequisite for survival and sustainable scale.
Start with work that already has a rulebook
The most effective early candidates for artificial intelligence agent deployment are tasks and workflows where the operating rules already exist explicitly somewhere outside of an employee’s head. This includes formal statutes, established contract clauses, or precise insurance compliance requirements outlined in vendor agreements.
Consider the routine administrative task of managing vendor certificates of insurance. Across the property management industry, operational teams typically spend roughly three hours per property every single week manually chasing these vital documents through email correspondence and phone calls. While the work is exceptionally tedious and repetitive, it is simultaneously completely specified by predetermined compliance standards.
Deploying agents to handle assignments like these provides decision-makers with two distinct operational advantages. First, the digital agent has something entirely objective and measurable to check its generated output against. Second, the internal human team can easily audit the resulting work without needing to debate subjective judgment calls or interpret ambiguous guidelines.
Recent research from McKinsey’s State of AI study revealed that the small group of companies realizing significant, measurable bottom-line financial returns from artificial intelligence—representing approximately 6% of survey respondents—were nearly three times as likely to have fundamentally redesigned their underlying business workflows.
Based on executive experience, that necessary workflow redesign invariably begins with writing the operational rule down in plain language. If a newly hired employee could not successfully follow a written description of a decision-making process on their very first day on the job, an artificial intelligence agent will not be able to follow it reliably either.
Set autonomy by workflow
Technology executives frequently ask whether a specific artificial intelligence system possesses autonomy, speaking about the capability as if it were a fixed, permanent property of the software product. In practical, day-to-day enterprise operations, however, the exact same agent should run with vastly different levels of freedom depending entirely on the sensitivity of the workflows it touches.
At Get Covered, engineering and operations teams configure individual agents across varying tiers of operational autonomy. For example, a routine automated reminder sent to a vendor whose certificate of insurance is set to expire next month can safely go out automatically without human intervention. Conversely, a formal response to a resident who has submitted a legal request for an emotional support animal is never handled autonomously.
A highly useful exercise for any corporate leadership team is to list every single workflow that an intelligence agent will touch and ask one critical foundational question about each process: If this specific autonomous action turns out to be wrong, how long will it take the organization to discover the error, and what will the financial and reputational cost have been by then? Quick detection capabilities and low potential costs earn an agent more operational room and freedom, while high-risk scenarios demand strict human oversight.
Give every agent an owner with a name
Earlier in professional leadership careers, executives learned that every automated credit decision or algorithmic underwriting choice eventually had to be clearly explained to an individual customer, an external industry regulator, or both. Simply stating that the predictive model made the decision was never an acceptable or legally compliant answer.
That fundamental principle still holds true today, except that modern autonomous agents act far more frequently, make decisions faster, and integrate across vastly more enterprise systems than any traditional underwriting model ever did.
Consequently, every active agent within an organization must have a designated human owner with a real name attached to it, functioning in the exact same manner as a human team or a departmental financial budget. That responsible person is tasked with regularly reviewing the agent’s error logs, deciding when to safely widen or narrow its operational autonomy, and ultimately answering for its behavior to leadership and regulators.
Establishing clear individual ownership also heads off a quieter, more insidious structural problem within growing companies. When multiple internal departments independently launch artificial intelligence agents without central coordination, leadership quickly loses visibility into where those various digital systems might contradict one another or create operational friction.
Where the constraint actually sits
In the modern marketplace, it is an absolute inevitability that underlying machine learning models will continue to improve in capability, speed, and accuracy. However, a company’s internal organizational clarity regarding who decides what will not improve on its own.
Gartner’s forecast predicting widespread project cancellations directly reflects this management gap. Most agentic artificial intelligence projects that ultimately fail will do so because leadership never clearly defined the business value or established proper risk controls—and both of those fundamental challenges are management problems rather than technical ones.
For decision-makers looking ahead, the wisest strategic approach is to treat each artificial intelligence agent the exact same way they would treat a newly hired employee stepping into a regulated corporate role. Organizations must write down its operating rules clearly, limit its functional scope appropriately, and assign it a dedicated human manager who regularly checks its work and evaluates its overall performance.
Companies that implement this disciplined approach will successfully gain the ability to grant agents greater responsibility over time, securely backed by the operational evidence required to justify it. Meanwhile, businesses that skip these foundational management steps will spend their time explaining to stakeholders why their software executed actions that nobody in authority ever approved.
Rick Folgmann is the chief operating officer at Get Covered, an insurance technology company providing cost-effective property coverage solutions. He brings more than 25 years of executive leadership experience spanning the financial and insurance industries.
