GRDF: a time-management rules engine rebuilt with no regression
483 rules, 53 SAP flows, 12,000 employees, and no room for a single error.
Rebuilding the rules engine behind time and activity management, on a system where any mistake lands directly in someone's payslip.
Project Context
GRDF operates the French gas distribution network. Its time and activity management system, known as GTA, governs the working hours, schedules and pay of 12,000 employees.
The engine behind it had to be rebuilt: hundreds of business rules to reimplement, and dozens of SAP flows to reconnect and make reliable. All of it without ever interrupting payroll or scheduling.
On this kind of system, a regression is not a defect logged in a backlog. It is a shift assigned wrongly, or a salary paid short. The requirement was therefore not a low error rate, but none at all.
Strategic Objectives
Rules that carry the whole business
483 time-management rules to reimplement, each one encoding a agreement, an exception or a legal constraint accumulated over the years.
A perimeter that cannot be interrupted
12,000 employees whose pay and schedules depend on the system. The rebuild had to happen without any service interruption.
Flows to reconnect one by one
53 SAP flows to integrate and make reliable. Each one carries data that other systems consume downstream, so an error propagates rather than stays local.
Zero regression, not few
The acceptance bar was absolute. Proving the absence of regression on 483 rules requires a coverage that no manual test campaign can produce in a reasonable time.
Solutions Delivered by Adservio
A spec-driven approach: every rule is specified in an executable form, then generated and tested under engineer review. The delivery agents carry the volume, the engineers carry the decision, and the SAP integration is validated flow by flow before it reaches production.
Every rule specified in an executable form
Each of the 483 rules becomes an executable specification rather than a paragraph in a document. That specification is what the implementation and the tests are both derived from.
Implementation generated under review
Delivery agents produce the implementation of each rule, and an engineer reviews it before integration. The volume is absorbed by the agents, the judgement stays human.
Non-regression coverage produced with the rules
The agents also generate the non-regression coverage. Proving that 483 rules behave as before is only feasible when the tests are produced at the same pace as the code.
SAP integration validated flow by flow
The 53 flows are integrated and validated one at a time before going live, rather than in one batch. A discrepancy is caught on its own flow, not once everything is connected.
Results
The full set of time-management rules rebuilt, each one specified, generated and tested under engineer review.
Every flow reconnected and validated on its own before production.
The perimeter of the system: the working hours, schedules and pay of 12,000 GRDF employees.
Impact
A rebuild that stayed invisible
Payroll and scheduling never stopped. On a critical system, the measure of success is that nobody outside the project noticed the engine underneath had changed.
Coverage that can be shown, not claimed
Non-regression generated alongside the rules turns an assertion into evidence. The absence of regression is demonstrated on the full set of 483 rules, not sampled.
Rules that are readable again
Executable specifications leave behind a set of rules that can be read, tested and modified. What was buried in a legacy engine becomes a documented asset.
More Client Work
Augmented CIO: a sovereign AI foundation across the SDLC
SpeedBooking: a platform shipped for 600+ hotels
A Workday migration mapping accelerated by AI
Augmented CIO: GenAI in the data software factory
Augmented CIO: GenAI at the heart of the SDLC
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
EBIOS RM risk analysis on Azure
DevOps & DataOps: Azure cloud migration
Redesign & digital maintenance: web performance
Fabrik Digitale: cloud-native on Azure
TIV Test & Reporting: rail maintenance
GOC 2.0 project: critical architecture audit
Digital transformation: Soon & Connect
Real-time fraud detection and faster KYC
Rebuild a critical system without breaking it
Let's talk about your legacy engine, your business rules and your non-regression requirements. An Adservio expert gets back to you within 24 business hours.