A new survey finds that 99% of executives expect AI to cause layoffs in their organizations within the next two years. That number is striking. But what's more concerning is the gap between executive confidence in AI replacement and the actual capabilities of current AI systems.

Having built and sold a company, I know how executives think about headcount. Employees are a line item. If you can reduce that line item while maintaining or increasing output, that's attractive from a balance sheet perspective. AI promises exactly that: automation that doesn't require salary, benefits, or time off.

The problem is many executives are making layoff decisions based on AI's potential rather than its current reality. Yes, AI can write code, generate marketing copy, and handle customer service inquiries. But it can't navigate office politics, understand institutional context, or make judgment calls in ambiguous situations.

When companies bet wrong on AI replacing workers, they lose something they can't easily get back: institutional knowledge. That's the accumulated understanding of why things work the way they do, what's been tried before, and where the bodies are buried. AI doesn't capture that. And when you lay off the people who have it, it's gone.

We've already seen this pattern with previous automation waves. Companies aggressively cut staff based on what technology should be able to do, then scramble to rehire when they discover the technology can't actually handle edge cases, complex workflows, or the messy reality of their specific business.

The 99% number suggests executives aren't thinking 'maybe AI will enable some workforce reduction.' They're thinking 'AI will replace workers, and we need to be ready.' That certainty is worrying when the technology is still hallucinating, struggling with reasoning tasks, and requiring significant human oversight.

Here's what I'd ask these executives: have you actually tested whether AI can do the jobs you're planning to eliminate? Not in a demo. Not in a pilot. In production, at scale, with real customers and real consequences for failures. Because if the answer is no, you're making layoff decisions based on vendor promises and hype.

The ironic part is that AI can genuinely augment human workers. It's good at automating rote tasks, generating first drafts, and handling structured workflows. Used correctly, it makes employees more productive - which should mean you need fewer people for the same output. But that's different from replacing employees entirely.

The danger is that by the time these companies realize AI can't fully replace human judgment, they'll have already laid off experienced workers who won't come back. That's institutional knowledge loss you can't AI-generate your way out of. And 99% of executives apparently aren't worried about it. They should be.