Most companies think they have an AI adoption problem. They don’t. They have an unlearning problem, according to Deloitte vice chair China Widener.

“If I said to you, ‘Everything you know about making spaghetti, I need you to let go of,’ you would look at me like I was stupid,” she said. “We aren’t talking about things that didn’t work. We are asking people to let go of things they have done every day for five or 10 or 20 years that worked, and we’re asking them to stop.”

For AI adoption to succeed, employees must unlearn, relearn, and learn simultaneously, Widener told me at the Fortune Brainstorm Tech conference last week. That’s easier said than done, especially when organizations are devoting far more resources to AI tools than to helping employees adapt to them.

Widener cited a Deloitte study showing that executives spend 93% of their AI budget on technology and only 7% on workforce adoption.

Eric Vaughan, CEO of enterprise-software company IgniteTech, sees the problem differently. He argued that the issue isn’t the tech spend or a skills gap—it’s employee resistance.