Key Takeaways

  • Town spends at least $75,000 per year per engineer on AI development software, splitting the budget primarily across tools like Cursor and Codex.
  • Higher developer efficiency does not shrink engineering teams; it converts unprofitable potential hires into profitable additions to payroll.
  • Grèze skips technical interview loops entirely when hiring candidates who have direct, close working history with existing team members.
  • Engineering headcount decisions should follow Grèze's AI Engineer ROI Framework.

The Grèze's AI Engineer ROI Framework

Baseline Unit Economics

Establish baseline revenue, fully loaded operating costs, and net profit per hire. For example, a company with $1,000,000 of revenue and $800,000 of cost produces $200,000 in profit available to hire one engineer.

Marginal Productivity Threshold

Evaluate whether the engineer's marginal revenue production exceeds their cost of employment. In a non-AI environment, if an engineer generates $150,000 against a $200,000 cost, the hire is unprofitable.

AI Tooling Multiplier

Equip the developer with AI tooling at a run rate of $75,000 per year per engineer across Codex, Cursor, and Devin to increase marginal output to $250,000, generating an incremental $50,000 net profit.

Headcount Decision Rule

As long as aggregate tooling and compute spend yield output exceeding the loaded cost of Silicon Valley engineering talent, aggressively scale hiring and AI tooling spend simultaneously.

When This Works (and When It Doesn't)

This framework applies directly when your startup has abundant, uncaptured customer demand and integrations where engineering capacity is the primary growth bottleneck. If your sales pipeline is overflowing and deals stall because product cannot ship integrations fast enough, subsidizing high tooling costs pays for itself immediately. As Grèze observed: “AI means that suddenly that engineer can generate more than they could have before. So maybe before they could only generate 150k of revenue. Maybe now they can generate 250k of revenue.” That shift changes the hiring math completely.

It fails when product distribution or market demand is the real bottleneck. If customers are not buying your software, doubling an engineer's commit velocity with $75,000 in compute tools only burns cash faster. Tooling spend creates return only when shipped code converts directly into recognized revenue. It also breaks down if your codebase has strict regulatory constraints that block third-party agentic tools from touching source files.

What to Do With This

Take your top engineering candidate in your pipeline this week. Run the math on their output before scheduling another technical screen.

First, calculate their loaded cost: if base salary, equity, and benefits equal $200,000, write that down. Second, estimate the tooling stack they need. If giving them unlimited access to Cursor and autonomous agents adds $75,000 annually, your total hurdle is $275,000. Third, look at your backlog. If shipping those enterprise integrations allows your sales team to unlock $350,000 in contracted annual recurring revenue, approve the hire and buy the licenses immediately. Finally, check if anyone on your core team has worked directly with the candidate before. If they have, skip the multi-stage coding test and extend the offer today.