The AI Layoff Reversal Has Started. Here's What Operations Leaders Should Take From It.

Written By Valorie Robles

A few weeks ago I wrote that the companies cutting hardest in the name of AI would look smart on a quarterly call and average-to-bad on a decade timeline.

I didn't expect to have receipts this fast.

The story that's been ricocheting around LinkedIn and Hacker News this month is Salesforce. Reports surfaced that the company had eliminated around 4,000 roles — primarily in customer support — on the bet that Agentforce, its AI agent platform, could absorb the work. By late last year, internal sentiment had shifted. The headline went viral: Salesforce regrets the cuts. The AI wasn't carrying the load they had projected.

Salesforce has pushed back on the framing. The official line is that it didn't lay off 4,000 people — it "rebalanced," redeploying support headcount into sales. Senior leaders have walked back the earlier confidence about AI as a blanket replacement and now talk about "targeted automation in well-defined use cases."

Both things can be true. Roles were eliminated. The company is, by its own admission, recalibrating what it thought AI could do. And whichever version of the Salesforce story is closer to the full truth, the bigger pattern around it is impossible to wave off.

The 55% number

Forrester estimates that 55% of employers who laid off workers for AI-related reasons now regret the decision.

More than half.

That's not a fringe data point. It's a majority of companies who pulled the trigger on AI-led headcount reductions saying, with hindsight, that they shouldn't have.

Meanwhile, IBM, Salesforce, Google, and Meta have all been quietly adding workers back into redefined roles — many of them specifically to run, train, oversee, and clean up after the AI systems that were supposed to replace humans in the first place.

You won't see this on the front page. Rehiring doesn't get a press release the way layoffs do. It happens role by role, requisition by requisition, in functions that get renamed just enough to obscure what's actually going on. But the trend is unmistakable.

What actually happened

The companies that cut hardest made a category error. They treated AI as a 1:1 replacement for human labor, when AI is actually a task-level technology operating inside human-level roles.

A customer support representative isn't doing one job. They're doing fifteen overlapping jobs at once: answering questions, de-escalating frustration, recognizing the customer who shouldn't be churned, escalating the issue that's actually a product bug, noticing the pattern across calls that nobody else is seeing. AI can do some of those. It cannot do all of them, and it cannot do the ones that require judgment and relationship.

What companies discovered is that when you remove the human and leave the AI, you don't end up with a leaner version of the original function. You end up with a degraded version of it — one that handles the easy cases fine and quietly fails on everything else.

Customers notice. NPS slides. Churn ticks up. The metrics that actually move the business start to drift in the wrong direction, and by the time leadership connects the dots, the institutional knowledge that would have caught the problem is already out the door.

That's the moment companies start rehiring.

What operations leaders should take from this

This isn't a told-you-so moment. It's a data moment. And the data should be informing what we recommend, build, and push back on inside our own companies.

A few things I'd take from it:

Stop treating AI as a headcount strategy. AI is a productivity strategy. Those are different conversations, made by different teams, with different time horizons. If the AI conversation in your company is happening in the same room as the workforce reduction conversation, that's a leadership tell — and it usually goes badly.

Map tasks before you map roles. Before anyone makes a decision about a person, do the actual work of identifying which tasks in their role are mechanical, which are pattern-based, and which are pure judgment. Most roles look very different once you decompose them. The "obvious" automation targets often aren't.

Track what happens after. If your company is already running AI-augmentation experiments, instrument the outcomes. Quality of work. Customer satisfaction. Time-to-resolution. Errors caught vs. errors missed. The companies that are quietly rehiring are doing so because something in their data forced them to. Don't wait that long.

Make the case for redeployment loudly. Most operations leaders are sitting on the exact data that would make the case for building over cutting — they just haven't packaged it. The institutional knowledge that walks out the door during an AI-driven layoff is enormously expensive. The cost of replacing it twelve months later, plus the opportunity cost of the time in between, is not a line item most CFOs are tracking until it's too late. That's your case.

Be the voice in the room. When a leadership team is considering AI-driven cuts, the absence of someone making the long-term case is usually how the decision tips toward the short-term answer. Operations is in the best position to be that voice — and the recent data finally gives us the receipts to back it up.

The takeaway

A year ago, the argument for using AI to grow your people instead of replacing them was a values argument. Now it's a values argument and a competitive one.

The data is starting to come in. The companies that cut hardest are quietly rehiring. The leaders who pushed AI as a labor replacement are softening their language. Salesforce — the loudest voice on agentic AI in the enterprise — is publicly recalibrating what it thinks the technology can actually do.

This is the part of the cycle where the early movers got it wrong and the smart movers learn from it.

Don't be the company that has to learn this the expensive way.

Build the workforce that grows alongside the technology. The receipts are already coming in for the people who didn't.

Sources

Valorie Robles
Previous

Why Operations Is the One Role AI Can't Replace — and Why Founders Keep Trying
Next

AI Isn't Replacing Workers. Leadership Is.