Free tool

Agent ROI & TCO Calculator

The monthly and annual cost of failures plus manual checking, from numbers you already know, and what cutting the failure rate is worth. Built to share with a CFO.

Runs in your browser. Copy or print for a business case.

your_numbers
§01 / THE BILLresult: monthly and annual
Cost of unreliability

What it's costing you

Failures / month$0
Manual checking / month$0
Total / year$0
§02 / THE LEVERreliability: one concrete change
What reliability is worth

Cut the failure rate

One lever: your failure rate. The saving is your failure bill today minus your bill at the target rate.

Failures today$0/yr
Failures at target$0/yr
You recover$0/yr
Measured, not estimated

That failure rate is a guess. Prefactor measures it per run and attributes cost per agent. Dev tier: 25,000 free spans a month.

§03 / METHODformulas: in the open
Methodology

How the numbers are made

Failures per month

agents × runs/day × failure rate × 30

How many bad runs reach production in a month.

Failure cost / month

failures/month × cost per failure

What those bad runs cost once they land.

Manual checking / month

hours/week × hourly cost × 52 ÷ 12

The human time spent watching output by hand.

What reliability recovers

failure cost now − failure cost at target

The saving from one lever: a lower failure rate.

A model from your inputs, not a guarantee. Prefactor measures the failure rate instead of estimating it. Agent evaluation

§04 / FAQfaq: what teams ask
Questions

Frequently asked questions

Why is the number so large?

Failure cost compounds with volume. If the total looks too high, lower the failure rate or cost per failure to match what you see. The tool makes the compounding visible, not bigger.

What counts as the cost of a failure?

Whatever a bad output costs once it lands: an escalation, a reversal, an engineer cleaning up, a credit. Use a blended average. Unsure? Start with the loaded cost of the time to detect and fix one.

How is the saving worked out?

Your failure cost today minus your failure cost at the target rate. Same volume, same cost per failure, one lever.

Is this a Prefactor quote?

No. It is a model of your current cost of unreliability, from your own numbers, with no assumption about any tool.

See how every agent performs, and make it better

Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.