Celonis Introduces AI Copilot for Enhanced Insights into Process Maps

Celonis, a German process mining startup valued at $13 billion, is embracing a multifaceted strategy for integrating AI into its offerings. While the company has utilized machine learning for years, the surge in generative AI has led to the introduction of an exciting new feature called Celonis Copilot, which was unveiled at this week's Celosphere, their customer conference.

Celonis enables organizations to analyze workflow efficiencies typically identified by expensive consultants. Last year, the company launched an innovative feature that visualizes multiple processes through a subway-style map, enhancing user experience.

This year, with the increasing influence of generative AI, Celonis is introducing Copilot, a companion feature that allows users to inquire about the data represented on the subway map. CEO and co-founder Alexander Rinke explained that they built Copilot on the OpenAI API. While the concept has been in consideration since the days of GPT-2, the recent advancements prompted the team to accelerate its development.

“It required significant orchestration, precise input, and vector database management. It wasn’t straightforward, but we recognized its huge potential, which drove our investment in this area,” Rinke shared.

In addition to Copilot, Celonis aims to enhance data accessibility within its platform, making it compatible with large language models (LLMs) utilized by clients. Instead of developing their own LLM, Celonis is concentrating on how to effectively interpret various data types tracked within their environment—a challenge due to the different terminologies customers might use to describe similar processes.

To address this complexity, the company is adopting a multi-faceted approach. They are launching a standardized process data model that allows customers to integrate all workflows and scenarios into a single, cohesive view—thus simplifying data interpretation for large language models.

Rinke emphasized the importance of defining key process elements, such as distinguishing between timely and late invoice handling. "We’re leveraging our extensive experience to establish clear business definitions for these elements," he stated. An accompanying API layer will enable seamless application development within the Celonis ecosystem and facilitate data integration with LLMs.

The strength of Celonis's ecosystem lies in the creation of a process intelligence graph, which interlinks various data types through the amalgamation of the process data model and a comprehensive dictionary of definitions. “By structuring process intelligence across systems and departments into one cohesive product, we provide a common language for process description throughout the organization, independent of underlying systems,” noted Rinke. This approach enhances customer value, expedites time-to-value, and positions Celonis as a robust network and platform provider.

These innovative offerings are currently in private release as the company refines and tests them with clients, with broader availability anticipated next year. Celonis has raised $2.4 billion to date, and its valuation reached $13 billion following a $1 billion funding round in October 2022.