Before You Scale AI Agents, You Need to Answer One Question: Who Is Responsible?

Before You Scale AI Agents, You Need to Answer One Question: Who Is Responsible?

Most business owners thinking about AI agents spend their time asking, "What can it do?" The smarter question is: "When it does something wrong, who answers for it?" That gap between excitement and accountability is exactly where most agentic AI deployments collapse and it is already happening at scale.

Rick Folgmann, COO at Get Covered, an insurance technology company serving more than 3 million rental units, published a detailed operational framework for managing AI agents based on his company's direct experience running 16 of them simultaneously. His core finding: the technical build was the easy part. The harder challenge was governance. Folgmann's piece, published on Crunchbase News, cites Deloitte's 2026 State of AI survey, which found that 23% of companies already use agentic AI at least moderately, and 74% expect to within two years. Gartner, however, has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027. The reason for that predicted failure rate is not the technology. It is management.

Folgmann outlines three specific principles Get Covered uses to run agents responsibly at scale. First, start with work that already has a written rulebook outside anyone's head, such as a statute, a contract clause, or a compliance requirement. His example: property management teams across the industry spend roughly three hours per property per week chasing vendor certificates of insurance by email. That task is completely specified and therefore a strong early candidate for an agent. Second, calibrate autonomy by workflow risk. At Get Covered, a reminder about an expiring vendor certificate goes out automatically, but a response to a resident requesting an emotional support animal never does. The test Folgmann recommends: ask how long it would take to detect an error and what it will have cost by then. Low detection time and low cost earn more autonomy. Third, every agent must have a named owner, a specific person who reviews its error log, decides when to widen or narrow its scope, and is accountable when something goes wrong. McKinsey's most recent State of AI research supports this discipline: the roughly 6% of companies seeing significant bottom-line returns from AI were nearly three times as likely to have fundamentally redesigned their workflows, not just bolted tools onto existing ones.

For small-to-mid-size business owners, this framework matters even if you are not running 16 agents across millions of properties. If you are using any AI tool that takes an action on your behalf, including sending emails, responding to inquiries, processing applications, or generating client communications, you already have an accountability problem waiting to surface. The Gartner cancellation forecast is not primarily about technology failure. It is about organizations that deployed agents without defining the value they expected, without limiting the scope of what the agent could touch, and without naming a real person to check its work.

The operational insight here is particularly sharp for businesses in regulated industries: real estate, insurance, healthcare, finance, or any field where a deadline or a disclosure requirement has legal teeth. Folgmann opens with a vivid example about California landlords who have 21 days to return a security deposit. The penalty for missing that deadline in bad faith can reach twice the deposit amount. That rule does not care whether the delay came from a person or a piece of software. Regulators and courts assign accountability to organizations, not algorithms. The same principle applies to any AI agent sending communications, making pricing decisions, or managing customer relationships on your behalf.

For your marketing and customer-facing operations specifically, this means your AI agents, whether they are scheduling social posts, responding to leads, or personalizing email sequences, need the same governance structure Folgmann describes. Who owns each agent? What decisions can it make independently and which ones require a team review before going out? What does your error log look like and when did someone last read it? If you cannot answer those questions today, your AI deployment is running on trust rather than process.

This week, take one AI tool or workflow currently running in your business and write down its rules the way you would explain it to someone on their first day: what it can send or do on its own, what it must flag for review, and whose name is on it when something goes wrong. If you cannot write that description clearly enough for a new team member to follow, the agent is not ready to act without oversight.

The businesses that will still be using AI agents successfully in 2027 will not be the ones who adopted fastest. They will be the ones who built the management layer that lets the technology earn more responsibility over time.

Originally inspired by: Your AI Agents Need A Chief Operating Officer, Too (https://news.crunchbase.com/real-estate-property-tech/agentic-ai-staffing-coo-folgmann-get-covered/)

Your turn

What is your traffic actually doing?

Send us your details and we will come back with a short, specific read on what your traffic, your pages and your pipeline are doing today — and the first three things we would change. A real person reads every submission, and you get the read whether or not we ever work together.