AI Startup Bans New Hires From Using AI to Learn Their Jobs

AI company restricts AI use for new employees

Valon, a New York-based AI startup valued at $1.75 billion, has implemented an unusual policy: most new hires cannot use AI tools until their managers confirm they've mastered fundamental job skills without assistance.

CEO and cofounder Andrew Wang instituted the ban last month after discovering employees were defaulting to the most expensive AI models for routine tasks. The practice was driving up costs and, more critically, preventing newer workers from developing the judgment needed to identify when AI produces incorrect outputs.

"By doing the basic work, rather than relying on AI, you start to form an understanding," Wang told Business Insider, which first reported the policy details.

The cost of over-reliance

Valon, which employs roughly 320 people and builds mortgage-servicing software powered by AI agents, gave employees broad access to AI tools before implementing the restrictions. When Wang examined usage patterns, he found a troubling trend: workers consistently chose premium models even for straightforward assignments.

When questioned, employees told him the AI was "almost always right" — an answer that revealed they were no longer questioning outputs or developing the expertise to spot errors. The financial impact was substantial: Valon had been spending $15 million to $20 million annually on AI tokens. Under the new policy, Wang projects that figure will drop to $4 million to $5 million.

The restriction applies across most business functions, including finance and human resources, and extends to senior-level recruits. Engineers are exempt because their code undergoes peer review before release, providing a built-in quality check that other departments lack.

Why it matters

Valon's experience reflects a broader workplace challenge as AI tools become ubiquitous. A September 2025 survey by BetterUp and Stanford's Social Media Lab found that 40% of 1,150 full-time U.S. desk workers received AI-generated work from colleagues in the previous month. Each instance required nearly two hours to review and correct on average. The tech community has even coined a term for those who share unvetted AI outputs: "meat proxies," popularized by German software developer Niklas Gruhn.

Early results and pushback

Wang acknowledged the irony of an AI-focused company limiting AI access, calling it "a weird decision for a company that is at the frontier of AI usage" in a recent blog post. However, he reports no internal resistance so far. Tenured employees have welcomed the change, having grown frustrated with cleaning up subpar AI-generated work from newer colleagues.

The policy is already changing behavior. New recruits now seek guidance from experienced team members rather than defaulting to AI for problem-solving, which Wang says is helping them develop deeper understanding of their work.

Some AI advocates have criticized the approach on social media. Wang's response: "If you have a much better idea here of how to make sure people learn, please tell me."

Valon has raised $275 million from investors including Andreessen Horowitz, WestCap, and 166 2nd. The company was last valued at $1.75 billion in 2024.

Business Insider first reported these details about Valon's new AI policy.