Valon has temporarily restricted AI access for most new hires, asking them to learn their roles without the tools until a manager decides they can recognize and verify incorrect output, according to the company's explanation of the policy.
- Policy scope
- Valon temporarily restricts AI access for most new hires
- Company size
- Roughly 320 employees
- Engineering exception
- Engineers can use AI because code receives peer review before release
- Manager gate
- Access returns after a manager judges that a worker can spot incorrect AI output
- Projected token spend
- A range of $15 million to $20 million falling to a range of $4 million to $5 million annually
AI access now depends on job knowledge
Valon introduced the rule after giving employees broad access to AI tools. In an August 6, 2026 post, the company said it had temporarily removed access from most new hires while it rebuilds an apprenticeship. The stated aim is to let people learn the underlying work before automation becomes part of their routine.
Andrew Wang, Valon's chief executive and cofounder, told Business Insider that the rule covers new employees in almost every part of the business, including senior recruits. A manager must decide that a worker can recognize when an AI answer is wrong before access returns.
Engineers are the exception because their code is peer reviewed before release, Wang said. Finance and human resources do not have the same review system, so new hires in those functions remain under the restriction.
The onboarding experiment is about judgment
Wang said the policy followed a review of Valon's AI use, which showed employees reaching for the most expensive models even on simple tasks. His argument is that doing basic work first creates context. A new employee may produce an answer that sounds correct while lacking the experience needed to spot the small part that is wrong.
The company is therefore treating AI access as a stage of onboarding rather than a default benefit. Valon says workers can use the tools after they have learned enough to question the output and check it against the job. New hires are asking more experienced colleagues for help, while tenured workers have welcomed the change because they had been correcting poor AI work, according to Business Insider.
That is Valon's account of an internal policy, not proof that employers across the technology labor market are limiting AI. The company is testing whether a slower start produces employees who can use automation with more independence later.
Valon expects a large drop in token spending
Wang also links the onboarding rule to cost. He projects annual token spending to fall from a range of $15 million to $20 million to a range of $4 million to $5 million this year, Business Insider reported. Those figures are company projections, not audited financial results.
The policy is temporary and gives managers unusual control over when new employees can use a central workplace tool. It may reduce early speed, but Valon believes the tradeoff is deeper role knowledge and fewer unverified answers. For a company whose own product is built around AI agents, the experiment turns a basic hiring question into a practical test: should workers learn the job first, then learn how to make the machine faster?
For now, Valon has set a clear threshold rather than a permanent ban. Access begins when a manager can trust a new hire to challenge the model, verify its work and take responsibility for the result.
Primary source
Check the original source
Valon blog is the source to consult for the underlying data, statement, ruling or live context.