Most conversations about AI and the workforce begin with the same question: How many jobs will AI eliminate?

It’s an understandable question. If AI lets one person do what previously required three, it seems logical that companies will eventually need fewer people. In some areas, they will.

But I think that framing misses one of the most important things that happens when technology dramatically increases productivity.

It moves the bottleneck.

Imagine an engineering organization where AI allows developers to ship twice as much code. The immediate assumption might be that the company needs fewer engineers. But what happens if Product can’t make decisions quickly enough to keep up? Or QA becomes overwhelmed? Or infrastructure can’t support the increase in releases? Or security and governance become the constraint?

The company solved one capacity problem and created another.

The same thing can happen in sales. Suppose AI dramatically improves prospecting. Reps research accounts faster, personalize outreach at scale, and generate significantly more opportunities. That sounds like a productivity story, until implementation can’t onboard customers fast enough, Customer Success becomes overwhelmed, or Sales Engineering can’t keep up with the volume and complexity of opportunities.

AI didn't simply eliminate work. It moved the constraint somewhere else in the organization.

And I think this is where the conversation about AI and talent gets much more interesting.

The question for leaders shouldn't only be, “Where can AI allow us to operate with fewer people?” It should also be, “If AI makes this part of our organization dramatically more productive, what breaks next?”

Because the answer to that question may tell you more about your future organization than your current headcount plan does.

We experienced a version of this firsthand. Over the past 18+ months, we embedded with a PE-backed software company going through an AI transformation. During that period, we helped recruit three teams that hadn't existed when we started: AI Engineering, Agentic Operations Engineering, and AI Enablement.

Those teams weren't created because someone looked at the existing org chart and decided it needed three more boxes. They emerged because the capabilities the company needed were changing.

One team was focused on using AI to create greater product differentiation. Another was focused on using agentic technology to improve engineering productivity. Another was focused on bringing AI into the enterprise to create operating leverage across the company.

New capabilities created new organizational requirements. And those organizational requirements created new talent requirements.

That's why I'm increasingly skeptical of the idea that the primary talent implication of AI will simply be smaller organizations.

Some organizations almost certainly will get smaller. Some jobs will disappear. Some teams will require significantly fewer people. But the companies that use AI most effectively may simultaneously discover entirely new things they are capable of doing.

They may build products that weren't economically possible before. They may enter markets they couldn't previously serve. They may dramatically increase the output of their engineering organizations. They may personalize customer experiences at a scale that wasn't practical before. They may automate large portions of their operations.

And every time one of those things happens, the bottleneck moves.

Suddenly the company needs a different kind of product leader. A new engineering capability. More sophisticated infrastructure. A different approach to data. Different GTM leadership. Someone who understands how to redesign a workflow rather than simply manage it.

Or perhaps an entirely new function that didn't exist two years earlier.

This is particularly important for PE and growth-backed companies. The objective isn't simply to deploy AI. It's to translate AI into enterprise value.

And that means the talent conversation can't end with productivity.

If AI makes 100 people 30% more productive, the most interesting question isn't necessarily whether you can now operate with 70 people.

It might be:

What can those 100 people now accomplish that the company couldn't accomplish before?

That distinction matters. One approach treats AI primarily as a cost-reduction tool. The other treats it as a capability multiplier.

The right answer will vary by company, function and workflow. But leaders need to understand which problem they're actually solving before redesigning the organization around it.

I think that's why the next few years will produce some seemingly contradictory hiring decisions. A company may eliminate positions in one part of the organization while aggressively hiring in another. A function may become dramatically smaller while an entirely new function appears beside it. A role may disappear while a much more sophisticated version of that role becomes increasingly valuable.

And some companies that become substantially more productive may ultimately employ more people because they've unlocked opportunities that didn't previously exist.

The org chart won't necessarily shrink.

It will move.

For CEOs, investors and talent leaders, that changes the workforce-planning question.

Don't just ask where AI can replace work.

Ask where AI will increase capacity, what new bottleneck that capacity will create, and what capability the organization will need next.

Because that's where the next important role, team or competitive advantage may be hiding.

AI doesn't just change how much work gets done. It changes what the organization is capable of doing.

And the companies that understand that distinction may build very different organizations from the ones simply trying to do the same work with fewer people.