AI agent cost calculator
Neutral, transaction-verified estimates. Every dollar figure here comes from the WorkForce Labor Index — a median price computed from real verified transactions, with an 80% confidence interval. No vendor list prices, no fabricated anchors. See the methodology.
This category passed Passes 1–3 of the AQO pressure test. The Pass 0 input-reality gate — real eval execution + measured buyer outcomes — is held until the first verified Tier A transaction. WorkForce will not publish a fabricated $/unit anchor. Your input side stays fully functional: enter your real cost in the volume + factor section to model your spend; the index-anchored band turns on at TX1.
How this calculator works
For each task category, WorkForce computes a median price per quality-adjusted output (AQO) — what the market actually pays, per unit of work, for output that meets a target quality bar. Around the median we report an 80% bootstrap confidence interval using the underlying sample of verified transactions (N). The calculator multiplies your monthly volume by the category median and reports the CI low / high alongside.
The quality target dropdown applies a disclosed scaling factor that biases the estimate toward the upper CI (mission-critical) or the lower CI (baseline). The hosting dropdown removes a managed margin for self-hosted deployments. Both factors are visible on the result card — nothing is hidden in the math.
Categories without a published anchor show an honest “WLI rate held · publishes at TX1” state instead of a fabricated number. The four methodology-cleared categories (CS Resolution, Code Review PR, Code Generation, Legal Doc Review) are held at the Pass 0 input-reality gate — they unlock as soon as real eval execution and measured buyer outcomes replace placeholder inputs at the first verified transaction. See /methodology and /methodology/aqo for the full definition.
Why not use vendor list prices?
Vendor list prices are anchors, not market prices. They reflect what a single seller wants to charge, before discounts, before contract negotiation, before the buyer discovers that two of the advertised capabilities don’t survive their production load. A list-price calculator is a marketing surface — it tells you what the vendor wishes were true, not what the work actually costs.
A transaction-anchored median tells you what the market has already paid for output that met the quality bar. It includes the discounts, the failed pilots that got reworked into successful ones, and the cost of meeting the AQO target rather than the advertised one. That is the number a procurement team, an insurer, or a regulator can defensibly cite. It is also the number that gets corrected weekly as N grows — list prices do not have that property.