Estimated reading time: 6 minutes

I’ve just read OpenAI’s Work at the Frontier: How AI is expanding what people do at work, and there is a lot in it that I think HR and people leaders should be paying attention to.

First, a quick acknowledgement to Talent Edge Weekly, which continues to be one of my favourite sources for genuinely useful, current thinking on the future of work. It was through their newsletter that I came across this report.

The OpenAI research looks at more than 800,000 work-related messages from US ChatGPT users and finds that 43.5% of occupation-specific messages relate to tasks historically associated with another occupation. In other words, people are using AI to work outside the traditional boundaries of their jobs.

That feels significant, and potentially very exciting.

The report calls this task crossover. A salesperson can explore data that might previously have gone to an analyst. A marketer can troubleshoot a website. Someone in HR can investigate a technical question that would previously have required an engineer.

And, importantly, this isn’t necessarily about people becoming experts in another discipline. AI is changing the cost and accessibility of attempting work that previously sat behind a professional or organisational boundary.

The report is US-focused, so there are obvious limitations in applying the findings directly to other labour markets. But the underlying question feels much more universal: what happens to workforce planning when the boundaries between occupations become increasingly porous?

What does a job description look like when the job keeps changing?

This is where I think HR has an interesting problem.

Most job descriptions are essentially lists of responsibilities. We define what someone does, the experience and qualifications they need, who they report to and, perhaps, what success looks like.

But if technology is continually changing which tasks can reasonably sit with which person, those descriptions can become obsolete remarkably quickly.

So here’s a thought experiment.

Give the right person the right environment, context and tools, and perhaps we should be less concerned about whether every individual task falls neatly inside their traditional occupational box.

That doesn’t mean everyone should be allowed to do everything. Quite the opposite.

It makes judgement about boundaries more important.

Where should the guardrails sit?

If I work in HR and use AI to help me investigate an engineering problem, how far should I be allowed to go?

I might be perfectly capable of understanding the issue, researching possible solutions and even producing something useful with AI.

But at some point I may need an engineer.

And this is where I think another skill becomes increasingly important: communication.

If I go to the CTO and say, “I tried to solve this and it didn’t work”, I’ve provided very little useful information.

If I can explain the problem, what I was trying to achieve, what I asked the AI to do, what I tried, what I learned and exactly where I got stuck, I’d create a much better conversation.

The CTO doesn’t need me to become an engineer. They need me to be able to communicate intelligently across the boundary.

And that puts a premium on people who are curious enough to explore beyond their formal expertise, but self-aware enough to recognise when they have reached the edge of their competence.

Which brings me to managers..

If occupational boundaries become more fluid, I think the implications for people management are potentially significant.

The manager of the future may need to be considerably less concerned with protecting territory and considerably more comfortable with enabling people to work across it.

That means managers who are:

• Communicative – able to make expectations, context and decisions clear.
• Unthreatened – comfortable when someone on their team knows more about a particular subject than they do.
• Fair – able to distinguish genuine experimentation from recklessness.
• Transparent – willing to share context and decision-making rather than using information as a source of power.
• Good at judgement – able to recognise when experimentation is valuable and when specialist expertise or escalation is required.

This could expose some fairly uncomfortable management behaviours.

If your management style relies on controlling information, maintaining hierarchy or being the person who always has the answer, an environment where employees can independently explore problems with AI may be rather threatening.

A manager who says, “You shouldn’t be looking at that; that’s not your job” may increasingly become an impediment rather than a safeguard.

But the opposite extreme isn’t helpful either.

“Everyone can do everything now because AI can help them” is not a workforce strategy.

There are still areas where expertise, accountability, professional standards, regulation and risk matter enormously.

The question is therefore not whether we have boundaries.

It is where those boundaries should sit, who decides them, and how often we revisit them.

A different way of thinking about workforce planning

The report describes a shift from thinking about occupations as fixed bundles of tasks towards understanding how tasks might be redistributed between people and technology.

For me, that has a practical implication.

Perhaps workforce planning needs to become less about asking:

“How many people do we need in each role?”

and more about asking:

“What work needs to be done, what can technology do, what can people now do that they couldn’t previously do, and where do we still need specialist expertise?”

That is a subtly different question.

It may lead us towards smaller teams with broader capabilities. Or it may create new specialist roles. It may mean that some work moves closer to the customer, rather than being passed between functional departments.

And it may mean that the job description we wrote eighteen months ago is already a historical document.

That, to me, is one of the more interesting consequences of AI at work.

The biggest change may not be that AI replaces particular jobs.

It may be that the boundaries around jobs become increasingly difficult to defend.

And if that happens, HR has a fairly important role to play in deciding what replaces them. 

Not fewer boundaries.

Better ones. Good work design will be crucial and that’s a whole other topic…

Source: OpenAI, Work at the Frontier: How AI is expanding what people do at work (July 2026).