Your AI Agents Need a Chief Operating Officer As Well

By Rick Folgmann

In California, landlords face a strict 21-day deadline following a tenant’s move-out to either return the security deposit or provide an itemized list of deductions. Failing to meet this requirement in bad faith can result in penalties up to double the deposit amount. The deadline makes no distinction about whether the notification was generated by a human or software; someone remains accountable for the outcome.

This reality is a frequent point of consideration at Get Covered, where 16 AI agents currently manage operations across more than 3 million rental units. Technical hurdles proved relatively straightforward during development. The greater challenge for every agent involved determining its reporting line within the organization.

A similar challenge awaits most businesses, and preparedness levels remain low. According to Deloitte’s 2026 State of AI survey, 23% of companies currently utilize agentic AI moderately, with 74% anticipating adoption within a two-year window.

Meanwhile, Gartner forecasts that over 40% of agentic AI initiatives will face cancellation by the end of 2027.

These figures support managing artificial intelligence agents with the structure applied to regulated personnel, which includes documented rules, clearly defined scopes, designated managers, and routine evaluations.

The following practices have driven success at our organization and serve as key focus areas for leadership teams scaling these systems.

Begin with well-defined rulebooks

The most suitable initial tasks for AI agents involve procedures where rules already exist outside human memory, such as statutes, contract stipulations, or insurance mandates within vendor agreements.

Consider vendor certificates of insurance. Property teams across the industry spend about three hours weekly per property pursuing these documents via email. Although tedious, the process is fully standardized.

Assignments of this nature provide leadership with two primary benefits: agents possess objective criteria to evaluate their outputs, and internal teams can review outcomes without debating subjective judgment.

McKinsey’s recent State of AI research revealed that the small subset of organizations achieving substantial bottom-line gains from AI—roughly 6% of participants—were nearly three times more likely to have fundamentally restructured their workflows.

Based on personal experience, this restructuring starts by documenting the rule. If a new employee cannot comprehend a decision-based description on their initial day, an agent will fail to follow it as well.

Establish autonomy according to workflow

Technology executives frequently inquire whether an artificial intelligence system possesses autonomy as though it were a permanent feature. In practice, identical agents should operate with varying degrees of freedom depending on the specific tasks they handle.

At Get Covered, agents are configured with distinct operational limits. For example, a reminder sent to a vendor regarding a certificate expiring next month may be automated. Conversely, replying to a tenant requesting an emotional support animal is never automated.

An effective exercise for leadership involves reviewing every workflow impacted by an agent and addressing a fundamental question: If an action proves erroneous, what is the timeframe for discovery, and what will be the associated cost? Rapid detection and minimal financial impact grant an agent greater operational freedom.

Assign a named owner to every agent

Earlier professional experiences demonstrated that automated credit decisions eventually required explanation to customers, regulators, or both. Claiming that an algorithm executed the action was never an acceptable justification.

This principle persists today, though current agents operate more frequently and across broader systems than past underwriting models.

Consequently, every agent at our firm has a designated individual ultimately accountable for its performance, mirroring how teams and budgets are managed. This owner evaluates error logs, determines when to adjust autonomy parameters, and answers for the agent’s actions.

Establishing clear ownership also prevents an overlooked complication: when multiple departments deploy agents independently without visibility into potential operational conflicts.

Where the genuine constraint exists

Ongoing improvements in underlying models are inevitable in the current market. However, organizational clarity regarding decision-making authority will not advance independently. Gartner’s cancellation predictions highlight this reality; the majority of failing agent initiatives will stem from unassigned value and unmitigated risk, both of which represent management challenges.

For leaders planning toward 2027, the recommendation is to approach every agent with the framework applied to a new hire in a regulated capacity. Document the operational rules, restrict the scope, and assign a manager to monitor performance and review outputs.

Organizations implementing this structure will successfully expand agent responsibilities over time backed by justifying evidence. Conversely, companies bypassing these steps will spend 2027 justifying actions executed by software without prior authorization.

---

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

Frequently Asked Questions

What percentage of companies currently use agentic AI according to recent data?

According to Deloitte’s 2026 State of AI survey, 23% of companies use agentic AI at least moderately, and 74% expect to use it within two years.

What prediction has Gartner made regarding agentic AI projects?

Gartner has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027.

What types of tasks are best suited for early AI agent adoption?

The best early candidates are tasks where rules already exist outside anyone’s head, such as statutes, contract clauses, or insurance requirements in vendor agreements.

Why should every AI agent have a designated owner?

Every agent needs a named owner to review its error logs, decide when to widen or narrow its autonomy, answer for its actions, and prevent independent teams from launching conflicting agents.

What percentage of companies are seeing significant bottom-line returns from AI?

McKinsey’s State of AI research found that about 6% of respondents were seeing significant bottom-line returns from AI.