What it looks like done right

In Sweden’s MASAI breast-screening trial, AI sorted the scans by risk. Radiologists kept every recall decision, and regularly overruled the AI. That line was drawn before launch.

The result: 29% more cancers found, with 44% fewer reads.

The boundary didn’t hold the AI back. It’s why the AI could be trusted with the volume.

How I work

Getting there takes four steps. Most organisations start at step three.

1. Redesign. Rebuild the workflow around the judgement your experts already use, then add what AI makes possible. Skip this and you automate a workaround.
2. Spine. Decide which decisions AI can take, which always need a person, and what that person must actually do.
3. Guardrail. Enforce that boundary on every live case. This is engineering work, done by your team or your vendor.
4. Automation. The work that runs inside the boundary. It’s what most AI vendors sell.

Steps one and two are almost always missing. That’s my work.

How a decision flows once the boundary is set

Phase 1: The Decision Spine Diagnostic

One workflow. Two weeks. In writing.

We take one process where AI helps make decisions about people, such as screening applicants, flagging performance or handling claims. We map how it really runs, not how it’s documented. You’ll know:

• every decision in it, and which ones AI already makes alone
• which should stay with AI and which need a person
• who can say no at each step, and whether they really can
• what record proves each decision was made properly

You find the gap before a regulator, a lawyer or a headline does.

Founding-client rate available until the end of 2026.

Phase 2: Implementation

Three to six months putting the boundary in place: first in the workflow from Phase 1, then across your other high-stakes ones.

I work with your team on who holds authority, what they check, and the record that shows it held. Where AI decides alone, that means who owns the boundary and how you’ll know if the AI drifts outside it. Your engineers or vendors build the guardrail. I give them the boundary to enforce and stay involved while they build it.

I don’t simply hand over a diagram and leave.

Start here

If a wrong call in your system has real consequences, and you can’t say, by name and in writing, what it’s authorised to decide alone, that’s where most of my conversations start.

Free 30-minute conversation. Contact me: mattsheehan@spatialnext.io

Different systems. The same missing decision.

Cigna kept the human in: a doctor signed every denial. Robodebt took the human out: Australia’s welfare system raised 470,000 wrong debts because a gap in the figures was enough to send a notice. Opposite designs, same failure. Nobody had decided what the software could do alone, what needed a person, or what that person was there to do.

The same gap shows up wherever a signal turns into an action. In Aurora, Colorado, a family was held at gunpoint when only one of three checks on a licence plate matched. In Baltimore County, an AI system mistook a Doritos bag for a gun and a 16-year-old was handcuffed. In both cases, nobody had decided in advance whether the signal was enough to act on.

Every one of these organisations had the technology. None of them had the boundary.

If you use AI or automation in decisions about people, the same gap is probably in your workflows too. The only question is who finds it first: you, or an employee, a regulator or a court.

That’s the gap I close. Before the harm, not after it.

Start with 30 minutes

Bring one workflow where AI shapes a decision about people. You’ll leave with a first read on where its boundary is missing and whether a Diagnostic is worth doing. No pitch, no obligation.

From there, the Decision Spine Diagnostic maps that workflow, in writing, in two weeks. Then we set the boundary together.

I’m taking on three founding clients at a reduced rate until 31 December 2026. California’s new rules on automated decisions in employment start the next day.

Book a free 30-minute conversation: mattsheehan@spatialnext.io

Read the full cases: Cigna · Robodebt · Aurora · Baltimore County

A WordPress.com website