Should HR professionals be using ChatGPT? Here's what I actually think.
Andy Maren — March 10, 2026
I get some version of this question every week now.
"Andy, is it safe to use AI tools for HR work?" "Which one should I be using — ChatGPT or Claude?" "Our IT team sent a policy saying we can't use any AI tools at work. Is that right?"
So I'm going to answer it as honestly as I can, from someone who has been actually using these tools for HR work — not just writing about them.
First, a distinction that matters
"AI tools" is too broad a category to give one answer to.
There's a meaningful difference between:
Consumer AI chat interfaces — ChatGPT.com, Claude.ai — free or paid accounts that you access through your browser, where the data you enter may be used by the vendor in various ways
Enterprise or API versions — the same underlying models, but accessed through agreements with specific data handling terms, often with data retention turned off
On-premise or private deployments — AI systems running inside your organization's infrastructure, where data never leaves
When your IT team says "don't use AI tools," they almost certainly mean category 1. And they're not wrong to be cautious.
The risk isn't that Claude or ChatGPT is going to maliciously misuse your employee data. The risk is that data entered into a consumer interface may be:
Stored and potentially used for model training (depending on the vendor's settings and your account type)
Subject to the vendor's data breach exposure, not your organization's security controls
Visible to vendor staff in some support scenarios
Not covered by your organization's data processing agreements
For HR work — where you are routinely handling compensation data, performance information, employee health accommodation requests, termination details — that matters.
What I actually use, and why
I'll be direct: I use Claude, and I prefer it for this kind of work.
That's not a knock on ChatGPT — it's an excellent product and I've used it too. For me, the preference comes down to a few things.
Anthropic's approach to safety and data handling aligns better with the way I think about HR data. Claude tends to handle ambiguous requests more carefully, which I find reassuring when I'm working with sensitive employee information. The reasoning quality on complex HR policy questions — multi-jurisdiction leave analysis, compensation band modeling, HRIS logic — has been consistently strong in my experience.
I also just find the way Claude reasons through problems clearer and easier to follow. When I'm building something, I want to understand the reasoning, not just get an answer.
That said, I don't think either tool is inherently safe or unsafe for HR work. What determines safety is how you use them.
Rules I actually follow
Here's my personal operating policy for using AI tools in HR work. It's not an official framework — it's what I've landed on through trial and error.
Never enter personally identifiable information. Full names, employee IDs, Social Security numbers, dates of birth, specific salary figures tied to named individuals — none of this goes into a consumer AI interface. Ever. If I need to work through a scenario involving a specific employee situation, I describe it in general terms: "a salaried exempt employee in California with 14 months of service" rather than "Sarah Johnson in our San Francisco office."
Use enterprise versions when available. If your organization has a Microsoft Copilot agreement, a Claude Teams account, or an OpenAI enterprise contract, use those versions. They have different data handling terms. If you don't know whether your organization has these agreements, ask IT. The answer might surprise you.
Don't paste policy documents you haven't checked. Employee handbooks, compensation frameworks, benefits plans — these documents may contain information that your organization considers confidential. Understand what you're pasting before you paste it.
Don't ask AI to make decisions. Ask it to help you think. This is the most important one. AI tools are genuinely excellent for drafting, analyzing, stress-testing, and exploring. They are not a substitute for human judgment in HR decisions, and using them as one creates both legal exposure and real risk of harm to employees.
Screenshot or log what you used it for. If you're using AI assistance in an HR process — drafting a performance improvement plan, analyzing accommodation options, reviewing a job description for bias — keep a record of where AI was involved. You'll want that documentation.
What companies are actually doing
The honest answer is: all over the map.
Some organizations have blanket bans on consumer AI tools and haven't yet built out enterprise alternatives — leaving employees either ignoring the policy or doing without tools their peers at other companies use daily.
Some organizations have adopted AI tools organization-wide without adequate HR-specific guidance on how to use them responsibly.
The ones doing it well have: a clear policy that distinguishes between consumer and enterprise tools, specific guidance for HR and people functions on what can and can't be entered, an enterprise agreement with at least one AI vendor, and a training program that treats AI literacy as a core HR competency.
If your organization doesn't have that yet, the gap is worth raising.
The question underneath the question
When HR professionals ask me if they should be using AI tools, what I often hear underneath is: "Am I going to get left behind if I don't figure this out?"
The honest answer is yes — but probably not in the way you're imagining.
The HR professionals who will struggle are not the ones who used the wrong tool. They're the ones who never developed a working mental model for what these tools can and can't do, who never got their hands on them, who waited for the official guidance that sometimes never comes.
You don't need to become a technical expert. You need to be curious, careful, and willing to experiment with appropriate guardrails.
That's very much what I'm still doing myself.
Update your curiosity.
— Andy
Andy Maren is the founder of Trailhouse Solutions, an advisory firm for HR and payroll leaders navigating the shift to agentic AI. She writes about what's actually happening in the market — without the keynote optimism.
