Celonis, the German process mining startup with a $13 billion valuation, is taking a different approach to AI. Like many enterprise software companies, it has been using machine learning models for years, but with the rise of generative AI, it is adding a copilot feature, announced today at Celosphere, the customer conference in the company that happened this week.
The company helps customers understand how work moves through a company’s various processes, using software to find flow gaps, something traditionally done by expensive consultants. price. Last year it introduced a new feature that allows customers to view multiple processes by displaying them on a subway-style map.
This year, with the rise of generative AI, they added Celonis Copilot, a feature that sits next to the subway map and allows users to ask questions about what they see.
Company CEO and co-founder Alexander Rinke says the company is building Copilot on top of the OpenAI API. This is something they have been thinking about for a long time, since GPT-2, but when it started this year, they decided to build it into the product.
“It requires a lot of orchestration and proper input and prompting and working with the vector database. So it’s not easy, but we’re accelerating our investment in that area because we just see the potential,” Rinke told the TechCrunch.
Beyond Copilot, the company strives to help customers make Celonis data accessible to a large language model within their companies, while also making it easier for other third parties. party partners or customers to build applications on the process of data stored in the Celonis platform. Instead of trying to provide an LLM, they focus on how to process the different types of data tracked within Celonis, which can be challenging for a large language model, especially if each customer has different methods to describe elements of the same type. in the process.
This is a complex problem, so the company decided to take a multi-pronged approach to solve it. For starters, they provide their customers with a standard and structured way to process data within Celonis, which they call the process data model.
“We launched this process data model, so that customers can combine all their processes and scenarios in one view, so they are naturally connected,” he said. That should make it easier for large language models to understand this data because it is combined into one entity.
The second entity includes the definition of different elements of the process such as what you mean on time or late for an invoice, for example. “We’re taking all the knowledge we’ve accumulated over the many, many years we’ve been doing this to figure out these business definitions,” he said.
Then there’s a whole API layer on that to expose it to the Celonis ecosystem to build applications on top of that, or expose data to LLMs.
The power for the ecosystem, however, comes when you combine the process data model and the dictionary of process definitions to build what they call a process intelligence graph, which reveals the connections between the various different from the data.
“We’re doing this to form a business’s intelligence process across systems and departments into a connected product,” he said.
“And basically it gives you a common language to describe the processes of the whole company, and is completely independent of the underlying systems. And it creates a lot of added value for customers. faster which is a time of appreciation, and also sets us up to build a network and platform,” he said.
It also makes it easier for customers or third-party partners to use the data in their own large-scale implementations of the language model.
These products are usually in private release for now, as they build them and test them with customers, but should be released sometime next year.
Celonis raised $2.4 billion, per Crunchbaseand was worth $13 billion when it raised $1 billion in October 2022.