FIELD GUIDE

Ground it, gate it, close the loop

A 12-point self-assessment for engineering leaders whose teams already use coding agents.

Agents write good code. Trusting what they ship is the hard part. This field guide turns the essay by Pasi Vuorio (ModernPath) and Antti Rauhala (Aito) into questions for your leads, a scorecard you can fill in in 30 minutes, and the first 4 moves once you know your score.

Based on the essay by Pasi Vuorio, ModernPath, and Antti Rauhala, Aito

Free · PDF · 11 pages · No sales call needed

Three pages from the field guide Ground it, gate it, close the loop: the dark title page, a page showing the decision gate diagram, and the 12-point self-assessment table.

12

statements to score with your leads, 0 to 2 each

30 min

to score it together, after a 15-minute read

4

first moves that tend to pay back fastest

3 → 1

months to release, after we ran the Agentic Engineering OS on our own delivery

The last figure is ModernPath's own delivery: same team, same quality bar.

One pattern, two halves

AI now builds software and runs it. On both sides, ungrounded generation fails the same way: it is fluent and wrong. The guide is built on the 3 steps that fix both.

Ground it

Building software

Generation answers to a model of the real architecture, data, and past decisions.

Running software

Action answers to a model of the business's own data.

Gate it

Building software

An audit and a human approval sit where being wrong is expensive.

Running software

A confidence threshold decides whether the software acts, assists, or abstains.

Close the loop

Building software

What happens in production updates what the system knows.

Running software

Outcomes are written back so the next decision is sharper.

Who it's for

Fill it in with the leads who own each area. Each of them gets questions written for their side of the work.

Engineering leads

The question it helps you answer

Do our agents work from the real system, and can we trace what they ship back to a requirement and an approver?

Product leads

The question it helps you answer

For the decisions our application could automate, should it act, assist, or stay out, and can it show why?

Data leads

The question it helps you answer

Can a prediction see our tables and the links between them, at roughly the cost of a query?

What's inside

The pattern on one page

Why building software and running it now fail the same way, and the 3 steps that fix both.

Questions for your engineering leads

Grounding, context and access, coordination, and governance, each with the sign that you have a gap.

Questions for product and data leads

When your application should act, assist, or stay out, and the sign of a gap in each area.

The 12-point scorecard

Score each statement 0 to 2 and see what your total usually means.

Where to start

The 4 moves that pay back fastest, from grounding one codebase to gating one decision in your product.

How it maps to what we build

One page on how the pattern maps to ModernPath for the build half and Aito for the application half.

Try 3 of the 12 statements now

Score each one 0 if it is not true today, 1 if it is partly true or true for some teams, and 2 if it is true across the organization.

  1. Ground

    Agents work from a maintained model of our real architecture, not only the open files.

  2. Gate

    People approve at defined approval gates instead of reviewing every step.

  3. Close the loop

    We can trace shipped code to its requirement and its approver in minutes.

The guide has the other 9, what your total usually means, and where to start if most of your zeros sit in one step.

Get all 12 statements

Get the field guide

Free. Leave your email and the PDF downloads straight away.

Ground it, gate it, close the loop · PDF

We'll use your details to send the guide and may contact you about it. You can opt out at any time. See our Privacy Notice.

The download starts right after: no email confirmation loop.

About the authors

Pasi Vuorio

Pasi Vuorio · Founder & CEO, ModernPath

Pasi Vuorio is the founder and CEO of ModernPath, which builds the Agentic Engineering OS.

Antti Rauhala

Antti Rauhala · Founder, Aito

Antti Rauhala is the founder of Aito, a predictive database for applications that act.

Questions about the guide

Who should fill in the self-assessment?
An engineering leader together with the leads who own each area: engineering for the build half, and product and data for the application half. If the honest answer to a statement is "it depends on who you ask", the guide treats it as a 0.
How long does it take?
About 15 minutes to read and 30 minutes to score with your leads. The total matters less than where the zeros sit, and the guide tells you where to start from there.
Is it about ModernPath's product?
Mostly not. Sections 1 to 5 work whatever tools your teams use: the pattern, the questions, the scorecard, and the first 4 moves. The last section is one page on how the pattern maps to what ModernPath and Aito build.
Why is Aito part of it?
The guide is based on an essay by Pasi Vuorio of ModernPath and Antti Rauhala of Aito. ModernPath works on how software is built, and Aito on how software makes decisions once it runs, so each author covers one half of the pattern.
What happens after I leave my email?
The PDF downloads straight away. We may contact you about the guide, and you can opt out at any time.
Can I share it with my team?
Yes. The guide may be shared freely in unmodified form.

Grounding is where the Agentic Engineering OS starts.

The engineering OS reads your real system into one Knowledge Core, holds requirements and their evidence in the Ledger, and drives the coding harnesses your teams already use.

See how it works →

Field guide · Free PDF

12-point self-assessment

Get the guide