Map the perimeter
Inventory every model that touches pricing, selection or eligibility. Classify against Annex III and mark what is in scope.
/ Risk & compliance /
Your pricing model is high-risk under Annex III. The file has to prove it works.
Life and health pricing with AI is high-risk under the EU AI Act, and the obligations landed on 2 August 2026 — risk management, data governance, technical documentation, logging, human oversight. We build the governance layer around the models you already run.
The problem
The model shipped. The evidence did not.
-
Annex III caught pricing, not just chatbots
Any system that scores or prices life and health cover for an identifiable person is high-risk. Most insurers found out after the model was live.
-
Nobody can reconstruct a decision from six months ago
Article 12 wants logs that let you retrace an outcome. Notebooks and spreadsheets do not survive that question.
-
Human oversight exists on the slide, not in the workflow
Article 14 asks who can override the model and how. In practice the underwriter sees a score with no way to disagree on the record.
-
Fairness is asserted, never measured
Proxy discrimination in health pricing is the exposure that ends up in the press before it ends up in a fine.
How it works
Governance wrapped around the models you already have
Instrument the decisions
Logging, lineage and fairness metrics on the live path — not a report someone assembles by hand each quarter.
Put the file together
Technical documentation, risk management system and the oversight workflow, in a form an auditor and a regulator can both read.
The evidence is produced by the system, not written about it afterwards.
Your pricing model is high-risk under Annex III. The file has to prove it works.
What's included
What you get when you run this with Thinkia
A governed pricing perimeter with the documentation, logs and oversight the Act asks for—built on the models you already run.
Annex III scope assessment
Model-by-model classification with the reasoning recorded, so the boundary holds when challenged.
Technical documentation set
Annex IV structure, generated from the system rather than transcribed into a template.
Decision logging and traceability
Retrace any pricing outcome to its inputs, model version and the person who reviewed it.
Fairness and drift monitoring
Proxy testing on protected characteristics, with thresholds that raise an alert instead of a quarterly surprise.
Human oversight in the workflow
The underwriter can override, and the override is part of the record — which is what Article 14 actually asks for.
Regulator-ready reporting
One place to answer a supervisor without a three-week fire drill.
Powered by Thinkia Sentinel
Results
What changes when this runs in production
Results vary by context, data maturity, and scope. We scope honestly before we promise precisely.
<6 weeks
To a documented Annex III perimeter across pricing models
Orientative — confirmed in discovery; depends on the starting point.
100%
Of in-scope pricing decisions logged and retraceable
Orientative — confirmed in discovery; depends on the starting point.
Zero
Undocumented models left in the pricing path
Orientative — confirmed in discovery; depends on the starting point.
How we work
From first insight to production
Assess
Assess
Week 1–2
Inventory pricing models, classify against Annex III, and mark the gaps that matter.
Design
Design
Week 3–5
Define logging, fairness metrics, oversight workflow and the documentation structure.
Build
Build
Week 6–9
Instrument the live path, assemble the file, and rehearse the supervisor conversation.
Govern & scale
Govern & scale
Week 10+
Extend to adjacent models, tune thresholds, and keep the file current as models change.
Timelines vary by scope and context.
Go deeper
Governance, compliance, and what to read next
Ideas, trends, and tools to stay ahead
Get started
Shall we look at what is actually in scope?
A short session on your pricing estate: what Annex III catches, what evidence exists today, and what the gap really costs.