Let’s talk about companies firing people, replacing them with AI, and then quietly hiring people back.
Not because the press release was wrong, exactly. Because the spreadsheet was too clean.
This is the pattern we are going to see more of. A company looks at support, operations, admin, sales development, finance, customer success, QA, even software delivery. Someone says the work is repetitive. Someone else says the model can do eighty percent of it. The CFO hears headcount reduction. The board hears margin expansion. The vendor hears enterprise contract.
And a few months later the company discovers the missing twenty percent was not twenty percent. It was the business.
The obvious version is customer support. A company replaces the support team with a bot. The bot answers faster. Ticket volume looks controlled. The demo is impressive. Then refunds climb. Escalations get ugly. Customers stop trying to explain the problem because the system keeps asking them to restate the thing they already said. The team saves money on salaries and starts paying in churn, reputation, chargebacks, and executive attention.
That is not automation. That is moving cost from payroll to damage.
The less obvious version happens inside the company. You remove the person who knew which supplier always needs a follow-up call. You remove the person who understood that one client never means the first thing they say. You remove the person who knew the workaround, the exception, the buried agreement, the old promise, the political landmine, the spreadsheet nobody admits is still important.
AI can summarize the process. It does not automatically inherit the institutional scar tissue.
That scar tissue is ugly. It is also useful. Companies pretend they run on processes. Many of them run on people remembering all the places where the process is lying.
This is the part executives underprice. A job title is not the work. A workflow diagram is not the work. A dashboard is not the work. The work is the living friction between customers, tools, edge cases, deadlines, incentives, half-broken systems, and humans who know how to get the thing over the line without setting the room on fire.
Can AI help there? Of course. It can draft replies. It can summarize calls. It can find patterns. It can pull documents together. It can check for missing fields. It can triage. It can write first-pass code. It can monitor exceptions. It can make a competent person much faster.
But if the plan is fire the competent person and let the system figure it out, you are not doing transformation. You are gambling that the visible task was the whole job. It usually is not.
There is also a moral laziness in the way some companies talk about this. They say AI replaced the role when what they mean is: we cut people before we understood the work. Then, when the mess becomes measurable, they quietly rehire contractors, consultants, fractional operators, senior people, or the same humans under a worse commercial arrangement.
So the employee becomes expensive again, only now with less trust.
The most interesting version is the boomerang hire. The company lets people go, automates the surface layer, then realizes it needs them back to supervise, repair, interpret, or calm down the system. The person returns, but the relationship has changed. They now know the company saw them as removable. They also know the company was wrong.
That is leverage.
If you are one of those people, do not return as if nothing happened. Return with a different frame. You are not coming back to your old job. You are operating the hybrid layer the company did not understand.
That means scope, authority, documentation, pay, and boundaries must change. If you are now responsible for the AI output, you need control over the workflow. If your judgment is the safety net, it should be priced as judgment, not nostalgia.
For leaders, the lesson is simple and inconvenient. Do not ask, can AI do this task? Ask what breaks if the human who understands this context disappears.
Ask what the person catches before it becomes visible. Ask which customers they protect. Ask which exceptions they handle. Ask which bad decisions they quietly prevent. Ask which part of the company is running on their memory because nobody bothered to build a real system.
Then automate around that person before you remove that person.
There is a humane way to do this. It is not soft. It is better business. Map the work. Capture the exceptions. Use AI to reduce the garbage. Give the best people better tools. Move them closer to judgment, escalation, relationships, and accountability. Let automation eat the parts that should never have required a human in the first place.
But do not confuse repetitive with simple. Do not confuse fast output with responsible work. Do not confuse a model’s answer with a company’s promise.
The future is not humans or AI. That is a childish frame.
The future is humans who can carry consequences using AI, against companies that tried to remove the humans and then had to hire them back to carry the consequences anyway.
And if you are building a company right now, remember this: the most expensive human is not always the one on payroll. Sometimes it is the one you fired, only to discover they were holding the messy parts of the business together.