Deutsche Telecom modernizes their process automation architecture and maximizes their RPA investment

Monolithic problems

Deutsche Telekom IT GmbH is the internal IT service provider of Deutsche Telekom AG. With around 9,700 employees globally and a total budget of around €1.9 billion, it’s responsible for the design, development and operation of all its own and transferred IT systems to support business processes at Deutsche Telekom and its subsidiaries. In 2007, Deutsche Telekom IT began using an Oracle BPEL engine to build and run business processes. But this monolith created a number of issues that affected both the business and user experience:

• A lengthy time-to-market, lasting more than 12 months
• Vendor lock-in limited the implementation of new features; it took five days to set up or make changes to environments
• Releases took, on average, 1000 people-days or roughly three months to realize
• Regression testing took around two days to process all test cases

“We need to be on time, we can’t wait seven days to process an order, we can’t wait seven days to respond to something, we need to be always ahead and always responsible,” explained Willm Tüting, Managing Director of conology GmbH, who has worked with Deutsche Telekom IT for more than 10 years. In 2017, running a 10-year-old monolithic system, Deutsche Telekom IT took the first steps towards modernizing its processes by adopting a ‘partially agile’ development approach, working in three-month sprints. “We created several processes to deliver small fixes faster, but it made life much more complex,” said Willm. All fixes had to be delivered together with larger change requests, which took considerable people-hours to accomplish. In addition, they were still struggling with the monolithic BPEL system, which didn’t allow for true agility.

Laying the groundwork for change

In 2017, Deutsche Telekom began investing in fiber optic cables to deliver a better user experience. With this significant upgrade in hardware came the opportunity to revolutionize Deutsche Telekom IT’s outdated systems. This saw a complete change in both the operating system and DevOps approach, guided by three goals:

• Speed up: Change development from BPEL to Java and the modus operandi from waterfall to agile
• Cross-functional teams: Change from skill-based teams to international cross-functional teams
• Efficiency: Increase development efficiency by using Camunda, Spring, and other state-of the art technologies

From a monolith to microservices in the cloud

With these three goals in mind, Deutsche Telekom IT implemented a microservices-based architecture in the cloud, with Camunda workflow engines running in many microservices. Friedbert Samland, Project Manager IT Application at Deutsche Telekom IT, explained the new system is comprised of:

• Microservices: This approach partitioned the monolith and allows for cross-functional work. There is no GUI, instead it is now a pure BPMN system. “Inspired by Camunda Co-Founder Bernd Ruecker’s microservices cheat sheet, we run Camunda engines in many microservices, talking directly to a message broker,” explained Friedbert.
• Cloud: The power of the cloud dramatically reduced runtimes and enables a staged, fine-gained delivery approach. With no ready-made solution, Telekom IT designed its own Kubernetes-based approach.
• Scale Agile Framework (SAFe): Introducing a new organization with shorter end-to-end cycle-times has enabled flexibility and speed.
• DevOps philosophy: Automation and self-service are key to Telekom IT’s DevOps philosophy and ensure quality and speed.

Camunda has changed the game

One of the greatest advantages of Deutsche Telekom IT’s Camunda revolution has been enabling compliance by default. As a globally distributed business, with teams working across the world with multiple vendors and sensitive data, a highly automated solution was required. The result is an architecture that enables ensures data security. In addition, as well as supporting and operating a highly flexible DevOps philosophy, Camunda has enabled Deutsche Telekom IT to visualize complex process logic in one place, easily align human and automated tasks, and use the same language for business and development with BPMN. For those who are thinking of following in Deutsche Telekom IT’s footsteps, Willm has sage advice: “Choose your stack wisely, don’t start with a full stack, choose things that add the most benefits to your needs. Think big, start small, don’t try to get the holy grail in the first attempt; evolve to where you want to be.”

Improving RPA with process orchestration

After transforming their monolithic architecture, Deutsche Telekom adopted multiple robotic process automation (RPA) tools to automate tasks within customer service workflows. This rapidly became one of the largest RPA implementations in Europe with approximately 3,000 bots running in production by 2020. While RPA initially delivered significant time and financial savings by automating repetitive manual tasks, the implementation quickly became hard to manage, with costly bot maintenance impacting the original savings. In addition, automated tasks that were siloed in RPA bots limited visibility into Deutsche Telekom's end-to-end business processes. Deutsche Telekom introduced Camunda as a process orchestration layer to improve process visibility and reduce bot maintenance costs. Adding a process layer on top of bots has enabled the core IT team to develop APIs that replace individual RPA bots, delivering a more robust and future-proof task automation solution.

Conclusion

Today, Deutsche Telekom teams can:

• Visualize complex process logic in one place, including tasks executed by RPA bots and APIs
• Easily align human work and automated tasks
• Use the same process modeling language for business and development, along with native Java support
• Visualize real-time operations, seeing the status of any process at a glance
• Enable compliance by default across globally distributed teams and multiple external vendors