AI Safety for Business: Why Human Oversight Is the Real Protection

From refund errors to rejected claims, automated decisions affect real people. Here's how companies can keep a named person responsible.

For centuries, the law has had a name for something that acts on your behalf. It calls it an agent. When the technology industry started calling its newest software "AI agents," my mind went straight to law school, which probably put me in a small minority. I spent years trying cases in front of juries before I ever started building AI, and that one word told me more about AI safety than most doomsday headlines do.

When an agent makes a mistake, the law asks one question: who sent it?

Are the AI Doomsday Warnings Justified?

Some of the loudest warnings about artificial intelligence come from people who helped build it, and I take them seriously.

The public is listening too. A Pew Research Center survey released in August 2026 found that 52% of Americans feel more concerned than excited about AI in daily life, up from 37% in 2021. In an earlier study, about six in ten said they want more control over how AI is used in their own lives.

That second number interests me most, because a desire for control is something people can act on. My advice is to take the warnings seriously and then turn the worry into a question you can actually answer.

When the Chatbot Gets It Wrong, the Company Answers

In February 2024, a British Columbia tribunal ruled on a dispute between Air Canada and a customer, Jake Moffatt. The airline's website chatbot had told Moffatt that a bereavement discount could be claimed after booking. That was wrong, and the airline refused the refund. Before the tribunal, Air Canada argued it couldn't be held liable for information from its agents, servants or representatives, the chatbot included. In effect, the tribunal observed, the airline was treating its chatbot as a separate legal entity responsible for its own actions.

The tribunal called that a remarkable submission and rejected it. The chatbot was part of the airline's website, so the airline was responsible for what it said.

The decision came from a Canadian small-claims tribunal, whose rulings apply to the parties in front of it, and its reasoning reflects an instinct as old as agency law: a company answers for its machine. Lawyers have a Latin maxim for this: qui facit per alium facit per se, whoever acts through another acts personally. Put someone, or something, to work on your behalf, and you answer for what it does inside the job you gave it.

Is AI Development Outpacing Our Ability to Manage It?

In one specific way, yes: companies are adopting AI faster than they are building the habits that keep it accountable. A business can put an AI tool in front of customers in a matter of weeks, and basic internal questions, such as who approved the tool and who can switch it off, often never get asked.

That gap worries me. In most organizations, control over AI slips away one decision at a time, as work moves off a person's desk and nobody records who now owns it. Responsible innovation, to me, means answering "who decided?" before a system launches.

How Everyday Americans Can Read an AI Warning

Anyone can weigh an AI headline by reading it the way a good juror reads testimony. Who is speaking, and what have they seen firsthand? Does the claim describe something a system actually did, or is it a forecast about a distant future? Then ask what it means for your paycheck or your kid's homework this week.

The AI most people meet today handles ordinary jobs such as drafting emails, screening job applications, quoting prices and answering support calls. For most families, the practical risk is an automated decision that's wrong and that nobody will own. So when a company uses AI on you, ask one plain question: if this gets it wrong, who do I talk to?

Where Human Judgment Can't Be Handed Off

In agency law, a great deal turns on scope: what the agent may do, and where its authority stops. Any business running AI should be able to show you that boundary in writing. The written scope should list the routine tasks the system may handle by itself. It should also name the person who reviews any decision that could seriously affect someone's health, finances, freedom, safety or livelihood, and who has the authority to reverse it.

Persuasion deserves particular care. I've spent my career studying what makes someone commit to a decision and keep it, and the lesson that keeps holding up is that lasting agreement comes from truth and genuine service to the other person. Software now holds conversations like that with millions of people every day. For AI safety, that means any tool designed to steer people's choices should meet the highest bar for honesty, with a named person accountable for it.

Responsible AI, Reduced to Its Essentials

• ● Save a screenshot of any chatbot answer you plan to act on, such as a promised refund. In the Air Canada case, the customer submitted chat screenshots to the tribunal.
• ● If a loan or insurance claim is turned down, ask in writing whether an automated system was involved and request a review by a person.
• ● Business owners can test their chatbot monthly with the questions customers ask most, checking each answer against the official policy page.
• ● Keep a dated record of the instructions you give an AI tool, so you can show what it was told to do if a dispute arises.
• ● Before launch, have a staff member play a frustrated customer and time how long the system takes to hand the conversation to a human.
• ● Check that any AI tool speaking for your business avoids pressure tactics such as invented deadlines.

Keeping AI Progress Under Human Control

I'm optimistic about this technology, partly because I'm blind and tools that fill in what I can't see have long been part of my work. AI now handles in minutes research that once took a week. I'm comfortable with that power as long as a person stays in charge of what it's used for.

The law settled this long before computers existed: anyone may put an agent to work, provided they accept responsibility for what it does. If every company deploying AI accepted that rule, the public would have far fewer reasons to fear the technology.

—----

About the Author
Sean Callagy is a blind, self-founded billionaire entrepreneur, attorney, and AI pioneer. He is the co-founder of ACTi (Actualized Intelligence) and of Unblinded, the company behind his Unblinded Formula for influence mastery. He also founded the 100-plus-person Callagy Law and Callagy Recovery, a medical-recovery firm valued at more than $1 billion. He has broken sales records across 19 appearances on Tony Robbins' stages, has spoken for Fortune 500 companies including Disney, Salesforce, and T-Mobile, and is one of only two attorneys in America to earn two Top 100 National Jury Verdicts.

LinkedIn and social
LinkedIn: linkedin.com/in/seancallagy | Instagram: @seanrcallagy | YouTube: @SeanCallagy