What does the pressure-test actually do?

It turns the draft against the question and names what the evidence does not support, what is missing, and what the brief is assuming. The weak part of the answer ends up inside the answer instead of being found afterwards.

Is this how you produce your own work?

Yes. Our own research runs this way: the question and the authorized sources are written down before gathering starts, the draft is pressure-tested against them before anyone reads it, and a named person accepts the brief before it informs anything. The same boundary applies to us as to you, since the brief informs the decision and a person still makes it.

The governed research brief, traced and pressure-tested

The answer came back confident and you cannot check it

Today

Two sources disagreed and got averaged into one clean sentence

The answer came back confident and you cannot check it.
The sources are somewhere in a chat log, or they were links that have since moved, and nobody kept a record of which claim came from which.
Two of them disagreed and got averaged into one clean sentence that neither of them supports.
And the thing the brief never says is what it could not establish at all.

With agents

What would answer the question written down first, and the draft argued against before you see it

A brief can be checked rather than believed.
Disagreement survives to the reader.
The brief says what it cannot establish.

How an agent runs it

A brief you can act on has to be checkable. This workflow writes down what would answer the question before it gathers anything, and it argues against its own draft before you ever see it.

01

Frames

Writes down the question you asked, what would answer it, and what would not, before it gathers anything.

02

Gathers

Takes only the sources you authorize, and records where each item came from as it goes.

03

Assembles

Builds the findings with every statement traced to its source, and names the sources that disagree rather than averaging them into one answer.

04

Pressure-tests

Argues against its own draft: what the evidence does not support, what is missing, and what the brief is assuming.

05

Human gate

The brief is not used or circulated until you accept it. Nothing is submitted outside the business and no source is contacted, the decision the brief informs is yours, and the whole brief can be assembled by hand at any point without the agent because the question and the sources are yours.

06

Records

Keeps the question, the sources and the brief, so the next one starts from the record rather than from a blank page.

Results

What changes when a brief has to be checkable

Checked rather than believed

Every statement carries the source it came from, so a reader can follow any sentence back to where it started.

Disagreement survives to the reader

Sources that disagree are named as disagreeing rather than flattened into a single confident line.

It says what it cannot establish

The pressure-test names what the evidence does not support before the brief is delivered.

Where this fits

Workflow / Automate a Workflow

Tier 1 through the Automate a Workflow door: one process end to end, from a question to a brief a person accepts. The question, the judgement and the decision the brief informs are yours, and nothing is submitted or sent outside the business. Where several standing questions later run under one set of gates, that direction is an administrative role or a back office rather than a promised state; How It Works shows how the acceptance gate is built.

Tier 1 / Prove

How It Works

Questions, answered

No. It gathers only from the sources you authorize, contacts nobody and submits nothing anywhere. The gate is closed by default, since the brief is not used until you accept it, and the question and the sources are yours, so the brief can be assembled by hand at any point without the agent.

The governed research brief