
By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR: Chuck Gallagher, AI ethics speaker and author, argues that AI can help employees do work outside their expertise but cannot give them the judgment to evaluate it, so owners must decide which tasks need review and which need a qualified expert.
Picture a fourteen-person property management company. The owner wants the employee handbook updated before the new year, and nobody on staff handles HR full-time. So he asks his office manager, who is sharp, organized, and reliable, to take care of it with AI. She does. In one afternoon she produces a clean, professional handbook. New policies. Updated language. Even a section on final paychecks.
It looks right. That’s the problem. Months later, a departing employee’s attorney points out that the final-paycheck policy doesn’t match state law. The office manager did exactly what she was asked. The AI did exactly what it was prompted to do. Nobody with the expertise to catch the mistake ever looked at it.
As an AI ethics speaker and author, I think that story captures a question every small business is about to face. Atlassian’s State of Product 2027 report, built on a Wakefield Research survey of 1,000 product professionals in the United States, Germany, and France, found that 89% of teams now regularly use AI to complete tasks previously handled by another function. And 67% agree it has become hard to clearly define their role because AI has blurred the boundaries between disciplines.
Does AI Give Employees New Skills or Just New Capabilities?
There’s a difference between being able to do the work and being able to judge it. AI hands people the first. It doesn’t hand them the second. A tool can help a marketing coordinator write a contract clause, a salesperson build a financial projection, or a receptionist draft a safety policy. What it can’t do is give any of them the experience to know whether the result is right.
The Atlassian report shows both sides of this. In its survey, 52% said prototyping is no longer limited to specific roles, 50% said salespeople can now create collateral without help from product or design, and 43% said designers are building and shipping functional work. That’s real progress. But the report also notes that the industry’s early fear, that specialized roles would merge into one, hasn’t happened. It describes a perfect example of the Dunning-Kruger effect, where everyone assumed the other craft was in trouble because they didn’t understand how complex it really was.
Atlassian applies that lesson to its own teams. The company says it doesn’t push product managers to write code just because they can. Ship it when the project calls for it. Otherwise, let people do the work their judgment is built for.
Why Is Role Stretching Riskier for Small Businesses?
Small businesses already run on people wearing three hats. The bookkeeper handles benefits. The office manager does marketing. The owner is the HR department. AI makes it tempting to add a fourth hat, then a fifth, because the work suddenly looks easy. Write the contract. Answer the tax question. Draft the safety procedure. Done by lunch.
But a small business rarely has a legal department or a compliance officer standing behind those employees. When a large company gets something wrong, there are usually layers that might catch it. In a small business, the first person to catch it may be a customer, a regulator, or an attorney. I spent the early part of my career as a CPA. Trust me, a lot of tax work looks simple right up until it isn’t.
Who Decides When AI-Assisted Work Needs Expert Review?
Let me be clear. The person who assigns the work is responsible for deciding whether it needs review. That’s the owner or the manager, not the employee who’s handed an unfamiliar task and a chatbot. It isn’t fair to stretch someone’s role, give them no training and no backup, and then blame them when the result falls short. That’s not delegation. It’s setting someone up.
Employees carry a responsibility too. If you’re asked to do work outside what you know, the honest move is to say so, and to say plainly what you’re unsure of. “AI helped me draft this, and I’m not qualified to say whether it’s legally correct” is not a confession of weakness. It’s exactly the kind of sentence that keeps a business out of trouble.
Now, the way I see it, fair expectations come down to three commitments from the owner. Train people for the new hats you’re handing them. Tell them clearly where their authority ends. And make it safe for them to say, “This is beyond me.”
How Can Owners Sort Tasks by the Expertise They Require?
A simple sorting habit helps. Start with tasks employees can complete independently. These are low-stakes, easy to check, and easy to undo, such as drafting social posts, summarizing meeting notes, or organizing a schedule. If the AI gets it a little wrong, the cost is small and the person doing it can usually spot the problem.
Next come tasks that require review. These reach customers or carry moderate risk, such as proposals, price quotes, marketing claims, or a basic spreadsheet model. The employee can do the work with AI, but someone with knowledge of the subject checks it before it goes out. Name that reviewer ahead of time so the step doesn’t get skipped when things get busy.
Last are decisions that require qualified expertise. Anything involving legal obligations, taxes, employment law, safety, health, financial commitments, or security belongs to someone trained and accountable for that field. AI can help prepare the questions. A qualified professional makes the call.
As an AI ethics speaker and author, I’ll leave owners with this. AI can make almost anyone capable of producing almost anything. It can’t make them qualified to stand behind it. Knowing the difference is your job.
Frequently Asked Questions
Should employees use AI to do work outside their job role?
They can, but with limits. AI lets employees produce work outside their training, yet it doesn’t give them the expertise to judge whether that work is correct. In Atlassian’s State of Product 2027 survey, 89% of teams regularly use AI for tasks previously handled by another function, which makes clear review rules more important than ever.
Who is responsible when an employee uses AI to do unfamiliar work and gets it wrong?
The owner or manager who assigned the work is responsible for deciding whether it needed review, training, or an expert. Chuck Gallagher, an AI ethics speaker and author, argues that it isn’t fair to stretch an employee’s role without training and then blame them for the result. Employees share responsibility for being honest about what they don’t know.
What tasks should always be reviewed by a qualified expert?
Decisions involving legal obligations, taxes, employment law, safety, health, financial commitments, and security should be made or reviewed by someone qualified in that field. AI can help gather information and draft questions, but it can’t replace professional judgment. Errors in these areas often carry legal or financial consequences.
How is AI changing employee roles?
AI is blurring the lines between jobs by letting people do work once handled by other departments. Atlassian’s report found that 67% of product professionals say it’s hard to clearly define their role because AI has blurred the boundaries between functions. Roles are expanding rather than disappearing, which raises new questions about training and oversight.
How can a small business decide which AI-assisted tasks need review?
Sort tasks into three groups: low-stakes work employees can complete independently, customer-facing or moderate-risk work that needs a named reviewer, and decisions that require qualified expertise. Base the sorting on how costly a mistake would be and how easily it could be caught. Revisit the list as roles and tools change.
Bring This Conversation to Your Organization
If AI is stretching the roles on your team faster than anyone has stopped to ask who should review what, now is the time to draw those lines. I help leadership teams, associations, and conference audiences build fair, practical expectations for AI-assisted work, so employees aren’t set up to fail and businesses aren’t exposed by work nobody qualified ever checked. To bring a keynote or workshop on AI ethics and accountability to your organization, connect with me, Chuck Gallagher, at ChuckGallagher.com.
Research source: Atlassian, The State of Product 2027, based on a survey of 1,000 product professionals conducted by Wakefield Research, August 13–26, 2026.
Five Questions for Reflection
- What tasks has someone on your team taken on recently simply because AI made it possible?
- Who in your organization decides whether AI-assisted work needs an expert’s review, and is that written down anywhere?
- Have you ever been handed work outside your expertise? What would have made it fair?
- How comfortable are your employees saying, “This is beyond what I know”?
- If an AI-assisted mistake reached a customer or regulator tomorrow, would the blame land on the person who did the work or the person who assigned it? Should it?
