Should We Stop Laying People Off and Crediting/Blaming AI?

The market is voting on whether AI did the layoffs, and the votes are coming back surprisingly readable (which doesn’t happen often).

CNBC tracked 23 S&P 500 firms that blamed AI for their layoffs. Fifty-six percent are trading in the red. Salesforce, which fired four thousand customer support roles and credited AI, is down 32 percent since the announcement. Nike is down 35. Fiverr is down 54.

Cisco cut almost the same number of employees in the same week, announced nine billion dollars in AI hyperscaler orders alongside the cut, and watched its stock jump 15 percent. The headcount move was nearly identical to Salesforce’s, but the market reacted in opposite directions based on how each company explained itself. Cisco’s story sounded a lot more like revenue than apology, and that did the work.

Three honest reasons companies do layoffs

Companies fire people for three reasons that have nothing to do with AI (there are probably more, but these are the three that get attached to every press release). The most common is that growth slowed, demand softened, the pipeline shrank, and the headcount no longer matched the work. Sometimes a bet went wrong, like a product line that missed or an acquisition that didn’t integrate, and the cut is the restructuring around the mistake. Occasionally the org chart simply aged out, because the company that worked at 500 people doesn’t work at 5,000, and the cut is the second draft of the company.

None of these is “AI did the work.” All three of them currently get AI dressed onto them in press releases, because AI sounds like the future and “growth slowed” sounds like a quarterly miss.

What AI is actually replacing right now

Some jobs really are disappearing into AI in 2026. They are real, they are measurable, and they are not the jobs in most of the layoff announcements.

HR: At the Level 1 recruiting layer, agents now handle resume screening, initial candidate outreach, interview scheduling, and most of the pipeline triage. The human recruiter picks up at the second round and the offer stage.

Customer Service: Agents handle most Tier 1 support now (password resets, order status, return intake, the FAQ-shaped tickets), and at companies that have actually deployed them they cover about half of customer interactions. The de-escalations and the retention saves still belong to people.

Finance and Accounting: Accounts payable is where the AI is doing the most visible work. Invoice ingestion, three-way match, exception routing, and vendor inquiries are largely automated, and the touchless invoice rate at well-deployed shops has moved from single digits to north of 60 percent in two years. The controller and the FP&A team are not on that list (I checked).

Sales Development: The first two hundred outbound touches a day that an SDR used to send manually are now mostly generated, with a human stepping in once a prospect actually replies. The closers are not affected.

Marketing: Generic copy production now lives with the agents (product descriptions, A/B test variants, social posts, and most first drafts of newsletters). The brand voice, the campaign concept, and the launch strategy still come from humans (at the companies you’ve actually heard of).

IT Help Desk: Password resets, software provisioning, and the first pass on common tickets are agent-handled now. The senior systems engineer holding the production database together at 3 AM is fine.

Legal: Standard contract review, NDA generation, and clause-by-clause compliance scans are AI-assisted at most firms with any volume at all. The lawyer in the actual deal negotiation hasn’t gone anywhere.

Procurement: Purchase order processing, vendor enrichment, RFP intake, and basic spend analytics are increasingly automated. The strategic sourcing leader who picks the contract is still the one picking it.

Engineering: Code completion, test scaffolding, documentation drafts, and most of the boilerplate now have an AI editor finishing the developer’s thoughts faster. The architects designing the system are still doing the design.

The pattern across all of those is the same: the jobs AI is actually doing are first-pass, repetitive, and high-volume. The job titles in the layoff press releases (senior engineer, regional sales manager, head of strategy, VP of customer success) are almost never on the list. The titles don’t match the technology, which means somebody, somewhere, is dressing.

Cisco and Salesforce told different stories

Cisco and Salesforce both cut around four thousand employees in the same week of May, and the market sent the stocks in roughly opposite directions afterward.

Cisco told a revenue story: AI hyperscaler orders had just doubled, and they were rotating headcount from the slowing categories into the growing ones. The cut paid for the growth in the parts of the business that were actually growing.

Salesforce told a cover story: AI agents now handle half of customer interactions, so they could cut four thousand customer support roles. There was no revenue line under that one.

Both announcements had AI in them, but only the Cisco announcement had cash showing through, and investors priced them accordingly. Cisco got the pop because there was a real revenue line behind the framing, and Salesforce got marked down because the AI framing was carrying weight the numbers couldn’t back up.

The cost to your remaining people

When you announce that AI took someone’s job, you are also announcing it to everyone whose job you didn’t take.

Your survivors hear: the AI is coming for you next (whether it actually is or not). Adoption of the new tools slows because nobody wants to help the thing that’s hunting them. Retention craters in the high-skill roles you actually need to keep, and the number you saved by laying off four thousand people gets eaten by the number you spend re-recruiting eighteen months from now.

I have watched a finance team’s AI rollout die because the rollout immediately followed an AI-themed layoff in the same building. The technology was fine, but the framing turned every new tool into a perceived threat, and the rollout never really recovered.

The cost to your former people

“We laid you off because the AI got better” is a worse goodbye than “we laid you off because the business slowed.”

When you tell someone they were replaced by a machine, you’ve ended the conversation before it started. The honest version keeps it human, with a real reason the person can actually respond to. Former employees become former customers, former referrers, and former boomerang hires, and the framing decides which list they end up on.

The honest version is five minutes of discomfort for the CEO and a graceful exit for the person leaving. The dishonest version saves those five minutes and costs the company a chunk of its ambassador network for the next decade. Nobody would take that trade if they wrote the math out (but plenty of CEOs do, because the math doesn’t get written).

The honest CEO playbook

What an honest version sounds like: “Our touchless invoice rate is 62 percent, which is why the AP team is going from 40 people to 20.” Or, when the business slowed: “Our growth is now 11 percent instead of the 22 percent the headcount plan assumed, and we are rightsizing.” Or, when a bet went wrong: “We invested in the wrong region last year and we are restructuring around what worked.” All three of those have arithmetic in them, which is what the market is in the business of pricing (arithmetic, plus occasional panic, but mostly arithmetic).

What the market does not reward is the dressing-up. “We are leveraging AI to drive operational efficiency” is the giveaway phrase. If you are using the word “leverage” as a verb, you are not having a real conversation, and the market is not having one with you either.

What I actually believe about all of this

I’m a libertarian. I think the law should enforce exactly two things: do everything you said you were going to do, and do not encroach on other people or their property (a remarkably low bar most institutions still fail to clear). Everything else, the market and the courts can sort out between themselves.

So companies should be free to fire whenever they need to fire (and people should be able to quit whenever they want to quit). A boss has no obligation to keep a job that the business doesn’t need, and that has always been the deal. I don’t want the government writing rules about layoff justifications, AI-related or otherwise.

What I do want is the second half of that contract.

If you are firing because AI did the work, prove it. Publish the touchless invoice rate, the tier-one deflection rate, the SDR pipeline coverage per rep. Let the market and your remaining employees look at the same number and agree it explains the cut. If the number is real, you are going to enjoy the stock pop Cisco got, and congratulations.

If you cannot prove it, you should not say it. Plenty of companies are lying right now, and the market is not gullible. The market is in the business of catching the dressing-up, and 32 percent down is what catching the dressing-up looks like.

The free market does the accountability work here without needing a single new regulation. We just need executives to stop helping themselves into the wrong column. Fire when you need to fire, say what you actually did, let the people who left walk out the door with their dignity, and let the people who stayed believe what you tell them. And if the AI actually did the work, prove it on the way out, the same way you would prove anything else.

None of that requires a regulation, just executives keeping the first half of the contract they signed when they took the corner office.