About Soma

We get the whole team in one room.

Soma is the standard that cross-functional automation frameworks are built on — and the workshop that puts it to work. It's built for the CTOs, CPOs, and heads of engineering whose teams each own a piece of an AI decision — and don't often sit together. We get product, risk, and engineering around one table, and help them leave agreed. Curious what your team would find there?

Book a workshop → Half day · product, risk & engineering · in person or online

Where the idea came from →

Who we get around the table

Five people who own the same decision — and rarely discuss it.

Each holds a real piece of how your AI decides. Apart, they make assumptions about the others. Together, for a day, they make one agreement.

Engineering

Sets the threshold the model actually decides on.

"I assumed someone signed off on 0.80."

Product

Owns the flow the person moves through.

"I didn't know that was a hard decline."

Risk

Carries the exposure when it's wrong.

"No one told me it auto-decides."

Legal

Weighs what's defensible later.

"Can we actually explain this one?"

Support

Takes the call from the person it lands on.

"I'm the one who has no answer."

Nobody in that list is wrong. They're just reading the same decision in five languages. Soma's job is to get them to one.

How it fits together

One standard, made practical.

The Standard is the fixed bar at the bottom. The Framework and your Style Guide are what your team builds on top of it.

iThe method

The Framework

A repeatable, cross-functional way to turn the four duties into questions product, engineering, and risk answer together. It's what the workshop runs on.

The workshop runs it

iiWhat you keep

Your Automation Style Guide

The four duties mapped to your real systems — a style guide for the decisions your AI makes.

Your team owns it

The foundationThe Standard

The Standard of Conduct for Automated Decisions — competence, candor, recourse, non-abandonment. Soma defines it, drawn from the duty-of-care best practices that govern medicine, law, and the professions.

Defined by Soma

Where the idea came from

It started as something written — not a service.

Before the workshops, there was a single note arguing one thing: that a duty of care doesn't disappear when a decision moves to a machine. Teams kept asking how to make it real on their own systems. The room came after.

The founding note · 2026

You automated the decision. You did not automate away the duty of care.

Long before machines, anyone whose work could alter a life was held to a standard. When the deciding moved to a model, that duty didn't vanish — it landed on no one. The argument the rest of Soma is built on.

The cornerstone · a 6-minute readRead the note →

Glimpses into the four duties

Competence

A system inherits the obligation to decide at least as carefully as a conscientious person would. Accuracy in aggregate is not enough; the duty is owed to each individual the decision lands on. A system that decides worse than a competent person is not a tool — it is a hazard that was chosen.

Candor

Disclosure is owed to the person affected — not only to an auditor or a regulator. It must be legible: a reason they can understand and respond to, not a notice buried in terms. "Proprietary" is not an answer a person is owed about a decision that shaped their life.

Recourse

Review must be real, not decorative: reachable without exhausting the person, and held by someone empowered to change the outcome. A decision no one can revisit has not been governed — it has only been imposed.

Non-abandonment

Care does not end at launch. The obligation is ongoing: to watch for the person the system begins to harm, and to reach those affected before the harm compounds. Abandonment — deploying and turning away — is itself a breach.