What happened
According to Business Insider, Mercor CEO Brendan Foody said on the "20VC" podcast, "Right now we're spending more on tokens for our internal agents than we are on employee head count." Business Insider reports that when host Harry Stebbings asked whether Mercor's token spend on AI agents exceeded salaries, Foody replied, "That's correct." Business Insider reports Mercor is a $10 billion startup, which supplies networks of human experts to AI labs, and that PitchBook listed about 300 employees as of October 2025. Business Insider reports Foody said Mercor has conducted more than 5 million AI-assisted interviews and uses AI agents for project management, recruiting, accounting, fraud detection, and candidate evaluation. Business Insider reports the company did not respond to a request for comment.
Editorial analysis - technical context
Companies that use externally metered generative models often face a direct, token-based cost model where runtime compute and API calls appear as line-item spend. As a general pattern, that shifts part of unit economics from fixed labor cost to variable compute cost, which raises the importance of monitoring, rate-limiting, and cost-aware prompt engineering. Teams that deploy large numbers of autonomous or semi-autonomous agents typically need more granular telemetry, quota controls, and model-version cost comparisons to avoid runaway spend.
Context and significance
Business Insider frames Foody's comment amid a wider executive debate over whether higher AI spend yields proportional productivity gains. Foody attributed his five-year prediction to falling model costs and improving capability driving a Jevons-paradox effect, in which cheaper AI leads to far more consumption rather than less, per Business Insider. For data teams and ML engineers, the anecdote underscores an emerging budgeting trade-off between human labor and API or cloud compute as AI usage moves from experimentation to operating scale.
What to watch
- •Adoption metrics tied to business KPIs, such as time saved or conversion lift, versus raw token and compute spend.
- •Unit-cost tracking per use case and per model or API endpoint.
- •Controls around agent proliferation, including quotas, circuit breakers, and cost alerts.
Key Points
- 1Mercor CEO Brendan Foody told the 20VC podcast the firm now spends more on internal-agent tokens than payroll, per Business Insider, confirming it directly to the host.
- 2The anecdote signals a shift from fixed labor cost toward variable, metered compute, pushing teams to adopt cost-aware design, telemetry, and quota controls.
- 3Practitioners should track ROI per use case and per model endpoint, since agent proliferation can drive token bills past headcount as workflows scale.
Scoring Rationale
The story signals a notable operational shift where API/token costs can outstrip payroll at scale, which matters for ML engineers and finance teams. It is important but not structural industry-changing on its own.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
