Take Ownership of AI’s Mistakes

One foundational idea in AI adoption is that AI is a tool, and the people who use it need to take ownership of what it produces. That applies when the tool gets it right, and it applies when the tool makes a mistake.

This matters because AI does not answer for its own work. A proposal that goes out with a wrong number is still your company’s proposal. A customer who receives a poor answer does not care whether a person or a piece of software wrote it.

Companies often feel that accountability for AI belongs to the business as a whole, or to a department, and not to an individual. Sometimes the reasoning is that deploying AI is a business-led initiative. Sometimes it is that AI is not a person, so it cannot be reprimanded the way a human employee would be. That view misses the key point, which is that AI is a tool. Even when an AI agent is designed to run autonomously in a workflow, one person needs to be responsible for that agent’s output. This creates the right behaviour, because the person responsible will make sure the AI is monitored and adjusted as needed so it performs to expectations, or removed if the AI is unable to do its job.

A published account of a European investment bank shows what can happen when no one person is responsible. An AI system that monitored trading for compliance had run for nine months under what appeared to be strong oversight. When it missed a pattern that a human expert would have caught, the regulator asked who was accountable and what the system had evaluated. The bank could not reconstruct what the system had done that day. The owner of the AI was listed as the compliance function, which had four hundred people, and none of them had personally signed off on anything. The head of compliance resigned, and fixing the problem took more than a year.¹

Your business is probably not an investment bank, but the lesson applies to any business. If AI drafts your quotes, answers customer questions or prepares reports, one person should be responsible for each of those uses. It should be the person who owns the business result, not IT and not a committee.¹

Taking ownership is accountability, and accountability depends on more than an assignment. It requires a corporate culture where people are held accountable for their results. If your culture does not have this attribute, there will be challenges with AI adoption. AI produces more work, faster, and someone has to answer for all of it.

Here is a short exercise. List the three places where AI does the most important work in your business. Beside each one, write the name of the person who is accountable for its output. If you write a department, or leave a blank, you have found the place to start.

Sources

¹ Pascal Bornet, “Who Owns It When the Agent Gets It Wrong?,” newsletter article, August 21, 2026, adapted from his book The Human-Agent Orchestrator.