Case study

How Santiane automated part of its support desk with AI

An entirely manual handling of incoming requests, on imperfect documents and within a strict regulatory framework. Standard cases are processed end to end, high-stakes cases always fall back to human review.

Published 30 July 2026

Industry

InsuranceHealth brokerage, three brands

Challenge

Incoming requestsHigh volumes, regulatory deadlines

Tandem service

Support automationAI extraction and routing

Constraint

Sovereign hostingHealth data, no third-party API outside the EU

The challenge

Imperfect documents, strict rules, and no room on the deadlines

Handling incoming requests was entirely manual. It tied up expert resources on repetitive tasks, with regulatory deadlines to meet on every case. The difficulty was not volume but the nature of the problem. Input data is imperfect by construction: scanned documents of varying quality, typos, heterogeneous formats. Business rules must be applied strictly, with no interpretation. And processing health data requires sovereign hosting, without routing through a third-party API outside the European Union.

manual processing

Every incoming request was read, qualified and handled by hand by expert resources, on repetitive tasks.

Scanned documents

at the front of the process

Input data that is imperfect by nature: scans of varying quality, typos, heterogeneous formats.

Health data

never to be exposed

A sovereign hosting requirement that ruled out any off-the-shelf solution routing through a third-party API outside the European Union.

The company

Santiane, France

Santiane

Santiane is a health insurance brokerage group bringing together the Santiane, Neoliane and Julia Mutuelle brands.

The group handles high volumes of incoming requests, in a sector governed by strict legal rules and demanding confidentiality requirements on health data. Two constraints that made automation both necessary and delicate.

Activity

Health insurance brokerage

Brands

Santiane, Neoliane, Julia Mutuelle

Market

France

Deployment

Self-hosted, on the group's own infrastructure

Our approach

Automate the standard path, always escalate the exception

The principle is simple: the workflow never makes a decision it cannot justify. Anything that leaves the standard path goes out as a draft to a human.

Deployment in a sovereign environment

A self-hosted workflow on Santiane's own infrastructure, so that health data processing never leaves the controlled perimeter.

Reliable reading of imperfect documents

Every incoming request triggers two parallel extractions, one tuned for scanned documents, the other for text files. Both streams are merged before analysis.

Structured extraction with safeguards

No value is ever inferred. A confidence score is calculated and uncertain fields are explicitly flagged, rather than filled in with an assumption.

Deterministic routing

Business and regulatory rules are applied by a deterministic module, never by the model. The case then goes either to automatic processing or to a draft for a human.

Results

Automatic for the standard, human for the exception

Compliant cases are processed end to end with no intervention. The workflow automatically switches to a draft as soon as a case leaves the standard path, so human validation is systematically preserved on high-stakes cases. All of it for a target cost below ten cents per case processed, API included, and eight weeks of implementation, without any health data leaving the group's infrastructure.

< €0.10

target cost per case processed, API included

0 wks

of implementation, in a sovereign environment

health data records leaving the controlled perimeter

of high-stakes cases kept under human review