Your Employees Are Already Using AI. Here Are the 6 Questions Every Business Owner Should Be Asking Now.

Your employees are probably already using AI. They may be drafting emails with ChatGPT, analyzing spreadsheets with Copilot, researching competitors with Claude or experimenting with tools you have never even heard of. Some of that usage may be officially sanctioned. Some almost certainly isn’t. For business owners, that changes the question.

We spent the first few years of generative AI asking whether employees should be allowed to use it. Increasingly, that’s the wrong question. Employees are using it, and the much more important issue is whether anybody is actually managing how.

KPMG’s 2025 American Worker Survey found that 87% of U.S. workers were using AI at least weekly and 51% daily. Meanwhile, a separate global study from KPMG and the University of Melbourne, covering more than 48,000 people across 47 countries, found that 44% of employees had used AI in ways that contravened organizational policies. It also found that 66% had relied on AI output without evaluating its accuracy.

AI adoption is already happening inside our businesses. Now we have to learn how to manage it.

AI isn’t just a tool anymore

Think about how AI entered most businesses. An employee discovered ChatGPT, tried it on a few tasks and realized it could save time. Maybe they used it to summarize a document, improve an email or brainstorm ideas. That’s relatively easy to manage.

But imagine that same employee six months from now. AI researches potential customers, qualifies prospects, drafts personalized outreach and updates the CRM. The employee may still be responsible for the outcome, but they are no longer doing all of the work themselves.

This progression is what I call the AI Management Ladder.

At Level 1, there is no AI and the employee does the work. At Level 2, AI assists them, while Level 3 sees AI handling defined parts of a workflow. By Level 4, an AI agent can execute multiple steps toward an objective. And at Level 5, you have a human-and-agent team, where employees increasingly manage outcomes while AI executes more of the work.

Moving up that ladder can create enormous productivity gains. But every step also increases the importance of management, because you are giving technology greater responsibility for work that ultimately affects your customers, employees and business.

Every manager needs answers to 6 questions

You don’t need to understand every AI model your employees use or create a complicated governance structure to get started. But every manager should be able to answer six questions.

1. Who is using AI?
2. What are they using it for?
3. How are they using it?
4. What company, customer or employee data are they putting into it?
5. Who checks the output?
6. Who owns the result?

That last question matters more as AI becomes capable of doing more.

I saw this recently while working with the owner of a growing service business. We were reviewing employee roles and realized that the traditional job descriptions no longer told us enough. They explained what each employee was responsible for, but not what parts of the job AI could perform, where AI required human supervision or where it shouldn’t be used at all.

So we started looking at the roles differently. Which responsibilities were Green, where employees could use AI freely? Which were Amber, where AI could help but a human still needed to review the work? And which were Red, where the task or information involved meant AI should not be used without approval?

That changes a job description from simply defining what an employee does to defining what the employee is accountable for — even when AI does part of the work.

Suppose an employee uses AI to help prepare a customer proposal. The AI includes an outdated price or invents a product capability, the employee doesn’t catch it, and the proposal goes to the customer.

Who owns the mistake? The answer cannot be ChatGPT.

AI may increasingly perform the work, but accountability still has to sit somewhere inside your organization. As business owners, we need to make that ownership explicit before something goes wrong, not afterward.

Give employees an operating agreement

One response to this problem is to write a long list of restrictions. But if the rules make AI difficult to use, employees may simply continue using it without telling you.

A better starting point is what I call an AI Operating Agreement. Divide AI use into three categories: Green, Amber and Red.

Green means use freely: This could include brainstorming, formatting, summarizing non-sensitive information, first drafts and ideation. These are low-risk activities where employees should be encouraged to experiment with approved AI tools.

Amber means use AI, but a human remains in the loop: Client-facing materials, financial analysis, important research, hiring-related work, recommendations or anything where an error could materially affect the business might sit here. AI can help produce the work, but an accountable person checks it before anything happens.

Red means don’t delegate it to AI: This might include sharing prohibited confidential information, uploading sensitive customer or employee data into unauthorized systems or allowing AI to make consequential decisions it has not been authorized to make.

The exact boundaries will be different for every company. A five-person marketing agency should not have the same rules as a financial services company with 5,000 employees.

What matters is that everyone understands where those boundaries are.

A policy isn’t the same as management

This is where business owners can easily develop a false sense of security.

Write an AI policy, email it to everyone, ask employees to acknowledge it, and the problem appears solved. Except it isn’t.

The KPMG and University of Melbourne research found widespread use of AI that contravened organizational policies. Having rules about AI and managing how people actually use AI are not the same thing.

Managers need to start making AI part of normal conversations about work. How did you produce this? Where did AI help? What did you check? What information did you give it? Could we improve this workflow? Could AI safely take on more of it?

Those shouldn’t be “gotcha” questions. They should become as ordinary as asking who is responsible for a project or whether a customer has approved a proposal.

The goal isn’t less AI

There is an understandable temptation for business owners to respond to AI risk by restricting it. But banning tools doesn’t necessarily stop people using them. It may simply make their usage harder to see.

The better objective is to create an organization where employees are comfortable saying when they use AI, understand the boundaries around that use and remain accountable for the results.

We’ve spent a lot of time teaching employees how to use AI. The next challenge is teaching managers how to manage people who use it. As AI moves from answering questions to completing workflows and executing tasks, that will become a fundamental management skill. Most businesses haven’t developed it yet.