How Societe Generale scaled its workflow portfolio in record time

In less than three years Societe Generale created its own workflow platform — SG workflow — onboarding more than 7,500 users who have successfully deployed an astonishing 180,000 tasks in the first quarter of 2020 alone.

Why implement workflows for stock trading? The answer is simple. Societe Generale, one of the leading European financial services groups, still sends thousands of emails per day to carry out client orders. What the bank needed was a system that told users what tasks needed to be completed and why a particular process was pending.

“Scaling the modelling process was one of the biggest challenges for us,” Simon Letort, Chief Digital Officer and Head of Innovation for the Americas, explained in his presentation at CamundaCon 2020.

But with this impressive scaling of process modeling, in just six months Societe Generale doubled the total count of implemented workflows, grew the total amount of tasks performed sixfold and increased the total number of active users on the SG workflow platform tenfold. Key to this success was a sophisticated ‘DIY Modeler’ implementation strategy aimed at high user traction.

Today 1,200 individual users — DIY Modelers — model workflows on the SG workflow platform, empowering a workflow mindset across the company. This was made possible by 30 expert modelers who work within several centers of expertise. Their job was, and still is, to model complicated processes and simplify the work of the DIY modelers. For example, they have integrated additional tools such as form.ui and drag and drop components on the central platform. Each DIY modeler can easily use these components for the workflows they want to create, without having to deal with actual code. These components include, for example, API calls for applications that have to be integrated into a given process. Only expert modelers are allowed to call REST APIs directly.

Providing a user interface was key to getting management support for the whole DIY concept because banks tend to be “experts for back-ends but not for front-ends” Letort said. In other words, if you weren‘t a computer scientist, you couldn‘t build workflows. Removing this obstacle saved the company significant time, because DIY modelers were now able to help themselves instead of having to wait for the IT department to process change requests.

Of course, the DIY modelers need more time to model a workflow than the expert modelers. While the 1,200 DIY modelers have so far created around 300 workflows for 30,000 tasks, the 30 expert modelers have produced 250 workflows for 100,000 tasks. However Letort explained that the important metric is not how much time the DIY modelers took to model a workflow, but how much time they saved the experts. The experts then invest their extra time to focus on deeply complex workflows that are absolutely critical for business success.

The centers of expertise also make work easier for the DIY modelers by providing best practices and entering metadata within the individual components that are used to automatically generate documentation. This auto generated manual already contains more than 400 pages that are always up to date and appealing to the supervisory authorities. In addition, because of these instructions, and the components tested by the expert modelers, the development team also receives fewer inquiries and can invest more time in programming.

The next big goals for SC workflow are already set. By the end of this year, the whole platform is going to be migrated to the Azure cloud and prepared for process mining, with DIY modelling rolled out to more than 30,000 users.

If you’d like to learn more about Societe Generale’s workflow platform, watch Simon Letort’s CamundaCon Live presentation and download the presentation slides.

  • Should data be embedded in business processes?...

    In October, I had the opportunity to lead a session at Camunda Unconference. This was a rather unique event, with an “unconference” format designed specifically for the community; it was most conducive to great peer-to-peer discussions, and exhange of ideas, experiences and learnings. What was also interesting was that the topics for the sessions were sourced from within the community, and each topic was voted in to a final shortlist. The sessions themselves were discussion-led, to encourage collaboration and creativity. Data in Process I had voted for “Data in Process” as one of the topics; it has always intrigued me, and needless to say, I was thrilled at the opportunity to lead the session! I have been consulting for customers building...

    Read more
  • Testing Process Dependencies

    Welcome to the next post in our blog series: “Treat your processes like code – test them!” You can find the last post: “Testing entire process paths” here. Today’s topic is “Testing process dependencies”. For the execution of a model, there are often additional resources required. This might be source code or the dependency on other models. But how can we deal with this when testing our models? In this post we will take a closer look at the following dependencies: Models: Dependencies to BPMN diagrams referenced by the executed model Code: Dependencies on source code referenced in the BPMN. We will get to know another library that will help us with testing: camunda-bpm-mockito. The examples for this blog post...

    Read more
  • Hacktoberfest Retro

    Hacktoberfest is a wrap! We’re so pleased that we were able to connect with so many of you from around the world. We set out with a goal of increasing the amount of engagement we get in our GitHub repos and you all delivered! From fixing typos to diving head-long into big feature requests, it was inspiring to see the pull requests that came in. Here are a few of my favorite recaps from the month: Two separate community members translated Tasklist into Hindi and Nepali. A third community member translated Admin, Cockpit, Tasklist and our Welcome Message into Brazilian Portuguese… with proper updates for the 7.14 release Another community member closed a 2-year-old feature request for Camunda to support...

    Read more