Real-World DEM Application

AI Breast Screening Changes the Decision Structure

Type: Real-World DEM Application
Sector: Healthcare / Breast Cancer Screening
Deployment: Region Stockholm / Lunit INSIGHT MMG
Published: October 2026

The deployment
AI breast screening in Region Stockholm is changing the traditional double-reading model. Lunit INSIGHT MMG is being rolled out across the regional breast-screening programme, where the AI will act as an independent second reader alongside a radiologist.

The DEM question
When AI formally occupies a decision position previously held by a human expert, where does authority sit when the AI and the remaining expert disagree?

Get the Full DEM Application
AI Occupies an Expert Reading Position

This is not simply AI presenting information for a clinician to consider.

The operating model changes from two independent radiologists to one radiologist and one independent AI reader.

Where Authority Becomes Visible

What happens when the radiologist and AI do not agree?

The practical question is who can resolve the disagreement, what evidence they receive and whether a challenge can cause additional review or another authorised action while it still matters.

Authority Over the Decision System

The most important human may not be the person reading today’s mammogram.

At regional scale, authority also sits with whoever can alter thresholds, add review, restrict use or suspend the AI across the programme.

When the Decision Structure Changes

Once AI occupies a formal expert position, human oversight is no longer only about whether a clinician can override a recommendation. Organisations also need clarity on who can resolve disagreement, change the operating rules, introduce further review, restrict the AI’s role or suspend it when performance or evidence gives cause for concern.

Explore the Decision Environment Diagnostic for your own organisation.

The Decision Authority Question

Region Stockholm is scaling AI into a formal reading role across breast cancer screening. Can decision authority remain clear when AI and the remaining expert diverge?

200k–250k

Mammography Examinations Annually

3

Major Screening Hospitals

1 of 2

Reading Positions Occupied by AI

What Organisations Must Answer

Authority matters at two levels: within the individual decision and over the design of the decision system.

01.
WHAT HAPPENS WHEN THE AI AND RADIOLOGIST DISAGREE?
Which disagreements trigger additional review, and who has authority to resolve them?

02.
CAN THE RADIOLOGIST EFFECTIVELY CHALLENGE THE AI?
Can professional judgement cause the reading pathway to change without disproportionate delay or workload?

03.
WHO HAS AUTHORITY OVER THE OPERATING MODEL?
Who can alter thresholds, add review, restrict use or suspend the AI across the programme?

04.
CAN THE ORGANISATION ACT IN TIME?
Can a material concern about performance, drift or decision quality trigger authorised intervention before more screening decisions are affected?

Get the full DEM application.

See how the Decision Environment Method examines disagreement, escalation, evidence, intervention and system-level authority when AI occupies a formal expert role in a consequential clinical pathway. The full DEM application shows what organisations should test as these operating models move from controlled implementation to scale.

Explore the Decision Environment Diagnostic for your own organisation.

    Related Case Studies