Boring on purpose.

Measured always.

Method

Four phases, each with a fixed price and a written exit. You can stop after any of them and keep everything we made, including the eval set that tells you whether an agent still works.

01

Weeks 0-2 / $6,500

Audit

We interview the people who do the work, time a sample of real cases and map every system the work touches. You get a ranked list of workflows with hours returned, cost per task, risk and build time, plus a prototype for the strongest one.

You keep: ranked list / prototype / first eval set / fixed quote

02

Weeks 2-8 / from $12k

Build

One workflow at a time. We grow the eval set to a few hundred real cases, build the agent against it and ship behind a review queue. It goes fully live only when it beats the score we agreed in the audit.

You keep: agent / integrations / eval set / review queue

03

Monthly / from $1,800

Run

We host and watch the agent, sample its work every day and fix drift before you notice it. Every month you get one page: volume, accuracy, cost per task and the cases a person had to handle.

You get: monitoring / daily sampling / monthly report

04

Any time / included

Hand over

When your team is ready, we hand over the code, the prompts, the eval set and a runbook, and pair with whoever takes it on for two weeks. No lock-in and no licence to keep paying.

You keep: everything

[05] What a build looks like

Invoice matching, Northwind Freight