Select your language (English only - some pages are automatically translated therefore some slight approximations)
CASE STUDY

ENGIE: Creating an AI factory to accelerate digital services

TESTIMONIAL


In this video, David Legendre, Chief Data Officer at ENGIE, explains how Artefact’s teams of Data Scientists are delivering multiple AI projects throughout the year.

The AI Data Factory dedicated to Engie’s data transformation 

There are major data projects at the heart of Engie’s digital transformation. David Legendre, Chief Data Officer at Engie France, presents the first season results of their AI Factory’s incubator deployed within Engie, that they called AI Data Garage, and recalls the ecological challenges which led the energy provider to reinvent itself.

 

Before, Engie was an energy vendor and was paid for that.

Today, Engie helps clients consume less and is paid for that.”

 

The group’s transformation is illustrated by the uses and challenges of data for the company. We want the group’s companies to comply with these standards, and data will enable us to attain this objective.

The goal of the AI Data Factory is to dynamically help resolve business problems we couldn’t crack before,”

explains David Legendre. 

 

The AI DATA FACTORY of Engie was configured like a large restaurant kitchen:

  • A place conducive to creativity
  • Enabling industrialisable recipes 
  • A staff organised in Feature Teams composed of business and data experts

4 use cases identified for season 1:

  1. Marketing(called e=mc2): personalisation of marketing actions
  2. Operations: maintenance of heating systems, for example
  3.  Satisfaction: call centre support
  4. Customer experience: use of visual recognition and pre-fill certain fields of Engie service subscription quotes


With regard to AI for maintenance actions, the challenge was considerable. A localised mesh had to be built for 200 centres to successfully submit predictions to enable optimised scaling of schedules and allocation of resources according to multiple factors (geography, weather, vacations…).

With the machine learning device developed, 85% accuracy on volume prediction was obtained after 30 days.

During the testing phase, this model will be available in 200 agencies in 2020.

 

The key success factors included:

  • The Feature Team: 1 Product Owner specialised in AI, 1 data scientist and 1 engineer who worked together on 1 use case one fixed KPI 
  • The data platform team: An IT team which is a powerful facilitator for ensuring access to quality data  
  • An early industrialisation logic from the outset: protocol documentation for quick and easy access by IT teams
  • Team support to ensure adoption and acculturation

The AI Data Factory is the best means of shifting from R&D to industrialisation,

with no major impact on the group’s balance,” 

concludes David Legendre.