STEF: Industrialising GenAI at the Heart of the SDLC
A hybrid LLM gateway and AI Act compliance from the very first sprint.
Adservio industrialises GenAI within the SDLC of a European leader in temperature-controlled transport and logistics.
Project Context
STEF is a European leader in temperature-controlled transport and logistics. Its IT department must accelerate delivery, contain technical debt, and industrialise GenAI without degrading its regulatory compliance (AI Act, NIS 2, DORA).
The answer: a hybrid LLM gateway routing each request by data sensitivity, versioned Skills and end-to-end observability, to industrialise GenAI within the SDLC while staying compliant with the AI Act, NIS 2 and DORA.
"We have around twenty AI POCs running, but none of them sustainable in production. We need a framework that industrialises without adding weight, and that holds up against the AI Act from the very first sprint.", Group CIO, STEF
Strategic Objectives
Business delivery pressure
Accelerate cycle time without adding to technical debt, to meet growing business demand without degrading the maintainability of the information system.
Multiplicity of LLMs
A need for routing governance based on data sensitivity, to orchestrate several cloud and on-premise models according to the criticality level of each request.
Strong regulatory compliance
AI Act, NIS 2 and DORA from the first sprint: compliance cannot be a layer added afterwards, it must be built in by design.
Cross-functional mobilisation
From C-level to developers, across 3 clocks: aligning the strategic vision, the pace of the teams and the rhythm of the organisation's transformation.
Solutions Delivered by Adservio
Adservio structured the transformation around a hybrid LLM gateway routing each request by data sensitivity (cloud when needed, on-premise when needed). Strategic and steering committees, a technical cell and a community of practice work together with the Adservio Academy. Three waves (MVP, Scale, Run) aligned with the 3 clocks (AI, humans, organisation).
Framing : 1 week
Priority use cases, data classification, target architecture.
POC : 2 weeks
Isolated test case, Skill demonstrator, baseline measurement, Go MVP decision.
MVP : 3 months
Hybrid gateway in production, 2-3 Skills, observability, Go/No-Go Scale file.
Scale · Run
Domain-by-domain industrialisation, Academy by waves, STEF autonomy above 80%.
Results
Delivery cycle time reduced thanks to the industrialisation of GenAI in the SDLC, without adding to technical debt.
Time between business need and production release cut, with a hybrid gateway live in production from the MVP.
Measured adoption on the rise, driven by the Academy and community of practice that make teams autonomous.
AI Act and NIS 2 compliance validated at every milestone, with 100% of LLM calls traced for observability and FinOps.
Impact
Sovereign hybrid gateway
A sovereign hybrid LLM gateway in production, routing each request to cloud or on-premise according to data sensitivity, without ever leaving the controlled perimeter.
Contained debt
Higher test coverage and contained technical debt: GenAI accelerates delivery without degrading the maintainability of the information system.
Observability and FinOps
Full LLM and FinOps observability, with 100% of calls traced to steer costs, performance and compliance continuously.
Catalogue of Skills
A reusable catalogue of versioned Skills, capitalising proven usages and accelerating each newly industrialised domain.
Autonomous teams
An Academy and community of practice that make the STEF teams autonomous: target autonomy exceeds 80%, with no dependency on a supplier.
“Lead-to-prod halved, AI Act validated at every wave milestone. What used to take six months in POC now fits into three operational sprints.”
More Client Work
Augmented CIO: a sovereign AI foundation across the SDLC
EBIOS RM risk analysis on Azure
TIV Test & Reporting: rail maintenance
GOC 2.0 project: critical architecture audit
Augmented CIO: GenAI in the data software factory
INES programme: AI-augmented depositary control
Interbank flows & SAP S/4HANA modernisation
360° Performance & Observability
Dentaline: deep learning on panoramic X-rays
AI FastPass & Mobile Order
EBIOS RM risk analysis & GDPR
Application Performance Centre
DevOps automation for life insurance
Building an educational nonprofit information system
Cinké Évolution : data visualisation steering
SIAG modernisation via SAP Fiori
DevOps & DataOps: Azure cloud migration
Redesign & digital maintenance: web performance
Fabrik Digitale: cloud-native on Azure
SpeedBooking: a platform shipped for 600+ hotels
A time-management rules engine rebuilt with no regression
A Workday migration mapping accelerated by AI
Real-time fraud detection and faster KYC
Industrialise GenAI in your SDLC
Let's discuss your LLMOps, hybrid gateway and AI Act compliance challenges. An Adservio expert gets back to you within 24h.