- Valon has implemented a new policy requiring most new hires to learn their jobs without AI before they can use it.
- CEO Andrew Wang introduced this policy last month after reviewing how employees were using AI tools, which he found to be costly.
- Wang discovered that employees were using expensive AI models for simple tasks, raising concerns about costs and understanding.
- Under the new policy, new hires can only use AI after their managers confirm they can identify incorrect output.
- Wang projects that the company's annual token spending will decrease to $4-5 million from $15-20 million.
- Wang's policy aims to prevent employees from becoming overly reliant on AI, which could weaken their judgment and understanding of their roles.
- The policy applies to new hires across almost all business areas, with engineers exempt due to peer review requirements.
- Wang noted that experienced employees have welcomed the change, as they had been correcting poor-quality AI-generated work from new hires.
Valon CEO Andrew Wang has mandated that new hires learn their jobs without AI, a decision made after observing excessive reliance on AI tools for simple tasks. This policy aims to enhance employees' understanding and reduce costs associated with AI usage.12
Wang noted that employees were frequently using expensive AI models for basic tasks, which raised concerns about both soaring computing costs and the potential weakening of their judgment. He stated, “By doing the basic work, rather than relying on AI, you start to form an understanding,” emphasizing the importance of foundational knowledge.36
The new policy applies to nearly all new hires, including senior employees, who can only use AI after their managers confirm they can identify incorrect outputs. Engineers are exempt due to peer review requirements for code.4
Wang anticipates that Valon’s annual AI token spending will decrease significantly, from an estimated $15 million to $20 million to approximately $4 million to $5 million. This reduction is coupled with a shift back to traditional workplace training, where new employees engage with experienced colleagues rather than relying solely on AI.
The initiative reflects a growing concern in the tech industry about the over-reliance on AI, as highlighted by a September 2025 survey indicating that 40% of U.S. desk workers received AI-generated work from colleagues, often requiring significant time to address.
Wang's approach, while seemingly counterintuitive for an AI-focused company, aims to cultivate a more knowledgeable workforce capable of critical thinking and problem-solving.
“Wang, a former Goldman Sachs analyst, said employees were using the most expensive models for simple tasks, and that reliance made them less likely to question AI output. A September 2025 BetterUp-Stanford survey found 40% of 1,150 U.S. desk workers received AI-generated work from colleagues, costing nearly two hours per instance.”
Tekedia
