Skip to content

Harness Design / three weeks / fixed price

Delegate valuable work to AI. Keep the outcome accountable.

Before you build, we choose the workflow worth automating, compare the ways to do it, and prove a first version against cases you agree on.

Sample Harness Spec · synthetic example

Invoice exception triage

Outcome
Route each invoice exception to the right owner, with the evidence attached.
Context
The invoice, the purchase order, the vendor record and the last three related tickets.
Authority
Read invoices and vendor records, draft the routing. A person approves anything that changes a payment.
Evidence
Forty past exceptions with known owners, scored before any build.
Budget
A cost and time limit per exception, and a cap on attempts.
Recovery
An unrouted exception returns to the existing queue, unchanged.

When to bring us in

Before the first agent is built.

A process everyone wants automated

Customers or staff keep asking for it, and nobody has decided what the agent may do.

Too many candidate workflows

Several ideas compete for the same budget, and you need the one worth doing first.

A build that needs a clear start

Leadership wants a first version with evidence before a larger commitment.

The handover

Six outputs. A clear build decision.

Harness Design ends in a build plan with its test cases written first, or in a clear recommendation not to build.

  1. The workflow, chosen by value

    Ranked against the alternatives on value, frequency and consequence.

  2. Options compared

    Your team, a product you already own, simpler automation, or an agent.

  3. Harness Spec

    One page: outcome, context, authority, evidence, budget and recovery.

  4. Test set

    The cases that decide success, written before the build.

  5. Scored prototype

    A first version, measured on those cases.

  6. Build plan

    Scope, sequence and acceptance, or a clear recommendation not to build.

The process

Choose. Compare. Specify. Prove.

01 / Choose the workflow

Value first.

Rank the candidates by value, frequency and consequence, and pick one.

02 / Compare the routes

The simplest route that works.

Your team, a product you already own, simpler automation or an agent, compared on the same criteria.

03 / Specify

One page, six questions.

Outcome, context, authority, evidence, budget and recovery, with the test cases written before the build.

04 / Prove and decide

A scored first version.

A prototype scored on the agreed cases, and a build plan or a recommendation not to build.

Before we start

What if an agent isn't the answer?

Then the recommendation says so, with the simpler route and the evidence behind it.

Who builds it afterwards?

Your team works from the Harness Spec and the test cases, or a forward-deployed build starts from them.

What do you need from us?

A workflow owner, examples of the work, and access to the systems and data it touches.

How is it priced?

Fixed scope and fixed price, agreed before work begins.

Start a conversation

What are you building—or deciding?

Tell us what you’re building, what is getting in the way, and the outcome you need.