Self-assessment
AI spend is up nearly 28x. There's a good chance you're overpaying by 30%.
28x figure: median quarterly AI spend in tech, over one year. DX, State of AI Impact in Engineering, Q2 2026
Answer sixteen questions about how your group is using AI, and get an estimate of what costs are structurally recoverable and which levers would recover them.
Opportunities this assessment considers
- Model right-sizingRoute routine work to cheaper models.
- Prompt cachingStop paying full price to resend the same context.
- Batch discountsTake the ~50% discount on work that can wait.
- Context engineeringSend the excerpt the task needs, not the whole file.
- Output disciplineCap the expensive half of every call.
- Agentic guardrailsKeep loops and retries from burning the budget.
- Deterministic harnessGive agents scripts for the mechanical steps.
- When to skip AIMove rule-following work to plain code.
- Price sensitivityBreak the bill down so nothing expensive hides.
- Today
- $30,000 / month
- After the usual fixes
- $18,000–$24,000 / month
- Recovered
- $72,000–$144,000 / year
Most of your overspend comes from four things: repeated context that isn't cached, routine work sent to a frontier model, jobs running in real time that could run overnight in a batch, and the calls that never needed a model at all. Each one is a property of how the system is built, not how heavily it gets used. That's why the fix pays every month instead of once.
Start the assessment →16 questions · ~4 minutes
No account, no cookies, no trackers.