The person-or-AI agent call is about more than cost and speed.
A company announces layoffs, names AI as the reason, and months later starts hiring some of those roles back. Forrester predicted in late 2025 that half of cuts attributed to AI will be reversed as organizations realize that replacing humans with machines isn’t always the right call, with many of those jobs returning offshore or at lower wages.
Harvard Business Review’s survey of more than 1,000 executives found the same thing from another angle. Most AI-attributed layoffs were made in anticipation of what AI might do, not in response to what it had done. Only about 2 percent tied large cuts to AI that was already in place.
These are big, visible companies making these calls ahead of reality. The same decision is now arriving on far more desks, in smaller and less visible forms. Most leaders haven’t slowed down to consider what it means.
The AI decision facing executives
The daily digest for entrepreneurs and business leaders
An Inc.com Featured Presentation
Some AI coverage frames the choice as build
Some AI coverage frames the choice as “build versus buy” your AI stack. That procurement question belongs to CIOs. The decision reaching founders and executives is deeper: For a growing share of the work in front of you, you can now assign it to a person or stand up an AI agent to do it.
Depending on the work, an AI agent can be faster, steadier, and cheaper. About 88 percent of organizations already use AI in some form, as of mid-2025, with AI agents moving from pilot into deployment
Standing up an agent takes a day and no recruiting, so the call gets made quickly. This means the decision often gets made without anyone examining the weight it carries.
Weighing a person against an AI agent brings
Weighing a person against an AI agent brings several factors into play: cost and speed, consistency and quality, and whether the work is defined clearly enough to hand off at all.
That last one looks simple but isn’t, because you usually can’t tell. When a capable person owns messy, half-defined work, they absorb the mess. They fill the gaps with judgment, ask the right person when something’s unclear, and cover for the fact that nobody wrote down how any of it works. It comes out fine, so you never find out the work was undefined. The person was holding it together, invisibly.
An AI agent holds nothing together. It does exactly what you specify and nothing more. Point it at that same half-defined work and the gaps surface immediately because there’s no judgment underneath patching them. Consider a client-onboarding process that has “run smoothly” for years. Handing it to an AI agent, a leader discovers it never ran on the documented steps. It ran on one client success manager knowing which clients needed a call before the contract, which paperwork to chase, and when to bend the sequence. None of that was the process. It was the manager.
Handing work to AI is less
Handing work to AI is less of a technology decision than it looks. The tools are good at working out how to reach an outcome, but you need to supply the standard for what a “good” output looks like, the judgment for edge cases, and an owner when the result is wrong. Point automation at work you haven’t defined clearly and it won’t resolve the ambiguity. It reproduces it at scale, quickly and confidently.
Then there’s the factor that’s easiest to skip because it’s the hardest to measure. A person carries things an AI agent doesn’t: judgment that transfers to the next unfamiliar problem, relationships and trust that compound over years, institutional knowledge, and accountability. None of that argues against automation. Plenty of work runs better as an agent, and treating every task as sacred would be its own mistake. The point is that this factor belongs on the table with the others.
Each of these calls gets made one task at a time. Individually they feel small. Made enough times without stepping back, they add up to an unintentional operating model, a shape your organization takes without anyone choosing it. For a lot of leaders, that accumulation is already underway.
Dividing work between people and AI agents
Dividing work between people and AI agents is harder than sorting tasks into two piles. For any given workflow, someone must decide what stays with a person, what an AI agent takes over completely, and what becomes human-led and AI-executed. Then they need to rebuild the process and the accountability around that choice, so it holds. That’s senior operational work: judgment about where the line falls, redesign of how the work flows afterward, and ownership when it doesn’t go to plan.
It’s also work most organizations have no spare capacity or expertise to take on. The people who could rethink the operation are the same ones currently running it, which is why the reflex is to buy another tool and leave the structure untouched. The harder move is handing the decision to someone with the strategic operational judgment to define the work and stand behind the outcome. That could be a leader already on the team or a COO brought in on a fractional basis to get the foundation right before the stakes climb higher. These choices reshape hiring and workforce planning too, so the people side has to be built in from the start, not bolted on after.
For most leaders, the person-or-AI agent call isn’t front and center yet. It’s happening at the edges, one task at a time, in decisions small enough to make without much thought. But those small decisions are what an organization eventually becomes. In the years ahead, leaders will hand more and more of the work to AI. Whether they do it deliberately is the part still in their control.
Get 1 Smart Business Storydelivered straight to your inbox when you subscribe to Inc.’s free daily newsletter.
Source: www.inc.com
One Comment