Salesforce Fired 4000 for AI Now Admits It Was Wrong

Salesforce fired 4,000 people this year to fund AI automation. Today, the company's Senior Vice President of Product Marketing admitted something that could reshape the entire tech industry: they were wrong. "All of us were more confident about large language models a year ago," Sanjna Parulekar said in a recent interview. This isn't some junior engineer second-guessing a decision. This is a C-suite executive at one of the world's largest software companies publicly admitting the AI revolution they bet the farm on might not deliver what they promised.

It's the kind of moment that makes you realize the AI hype cycle might be hitting a wall.

The Great AI Gamble That's Not Paying Off

2024 was the year tech companies went all-in on AI. Salesforce, like so many others, decided the path forward was clear: eliminate headcount, automate with AI, and watch the savings roll in. So they handed out pink slips to around 4,000 employees - roughly 10 percent of their workforce. The message was simple. We don't need these people anymore. AI can do their jobs cheaper and faster.

But here's the problem. It's been a year, and the results aren't showing up.

Salesforce is now in the uncomfortable position of admitting that maybe, just maybe, they moved too fast. Maybe they believed in the hype a little too much. The company's own leadership is questioning whether they should have fired all those people in the first place. And if Salesforce is having doubts, you have to ask: what about everyone else?

The AI Confidence Collapse

Parulekar's statement isn't just an offhand comment. It's a public acknowledgment that the expectations around large language models (LLMs) have dramatically shifted in 12 months. A year ago, the narrative was unstoppable. AI would replace knowledge workers. It would automate customer service, finance, HR, sales operations. All those middle-management roles? Gone. The consultants, the analysts, the specialized support staff? Unnecessary.

Except it didn't work like that.

The reality of deploying AI in real business operations turned out to be messier than the press releases suggested. LLMs hallucinate. They require constant human oversight. They're not nearly as reliable as a human expert for complex decision-making. And here's the kicker: training and maintaining AI systems requires skilled people. So companies that fired their specialists are now scrambling to hire consultants to clean up the mess.

Salesforce is experiencing what researchers have quietly been confirming for months. AI-generated code contains more bugs and errors than human output. AI systems make mistakes that trained humans would catch. And using AI without human review is just asking for disaster.

Other Tech Giants Are Having the Same Doubts

Salesforce isn't alone in this existential crisis. Across the tech industry, cracks are appearing in the AI-everything narrative. Companies that spent billions on AI infrastructure are asking uncomfortable questions. Are we getting the ROI we expected? Are our employees actually happier? Is automation really cutting costs, or are we just shifting them around?

The most telling sign? Tech CEOs are still spending on AI, but they're doing it with noticeably less confidence. Companies that said they'd cut 30% of their workforce through automation are now quietly hiring back for specialized roles. The narrative that AI could replace 80% of knowledge work has shifted to "AI can augment human workers on repetitive tasks." That's not a pivot. That's a retreat.

Salesforce's admission is significant because it comes from someone who's been inside the machine. Parulekar isn't some external critic. She's part of the team that made the decision to lay off 4,000 people. And she's basically saying: we trusted the hype machine too much.

What Actually Happened in Those 4000 Jobs

When you fire 4,000 people betting on AI, you're assuming those people were replaceable by technology. But most of them weren't doing repetitive, algorithmic work. They were doing the messy, human-centered stuff that keeps customers happy. They were solving edge cases. They were handling exceptions. They understood customer context in ways that no LLM can replicate without constant babysitting.

Now Salesforce is learning what dozens of other companies have discovered: you can't just swap humans for AI and expect the same output. The customer experience takes a hit. The quality declines. And then you're hiring consultants to fix the damage, which costs way more than keeping the original people would have.

The real kicker is that Salesforce still has to pay the severance and termination costs. Those employees don't disappear free of charge. And many of the best ones found new jobs or started competing companies. So Salesforce paid to lose institutional knowledge while betting on AI that didn't deliver.

The Ripple Effect Across Big Tech

This matters way beyond Salesforce because it signals something the tech industry has been reluctant to admit: the current wave of AI hype might be overselling what the technology can actually do. Companies like Microsoft have been pushing AI everywhere - from Windows to Office to Azure. But how many of those AI features are actually being used? How many are solving real problems versus just generating PR?

Meanwhile, smaller companies and startups are watching what the giants do. If Salesforce is doubting their AI investments after firing thousands, why would a mid-market company confidently bet their entire strategy on LLMs?

This is the beginning of the AI reckoning. Not the end of AI - the technology still works and still has real applications. But the end of the idea that AI can replace entire categories of human workers without consequences.

What Comes Next

Salesforce is reportedly looking to hire back in areas where they over-cut. That's not a coincidence. It's damage control. They're realizing that specialist roles - the people who understand your specific industry, your specific customers, your specific problems - can't be replaced by a chatbot trained on the entire internet.

Expect to see similar patterns across tech in 2026. Companies that went aggressive on automation will shift to "AI augmentation" strategies. HR departments that got slashed will quietly rehire. The rhetoric around AI will change from "this will replace humans" to "this will help humans be more productive."

And the employees? Many of them won't come back. They're already at competitors or working for startups that understand that human expertise still matters.

Bottom line: Salesforce's admission that they were wrong about AI might be the most important tech story of the week because it shows the AI hype cycle starting to crack at the foundation. When executives at one of the world's biggest software companies admit they moved too fast on automation, every other company betting their entire strategy on AI should be paying attention. The narrative hasn't changed yet in press releases and earnings calls, but behind closed doors, doubts are spreading. And that's where real change starts.

---

AI Generated Image | AI Generated Image