Celonis wants to be the roadmap to rebuild workflows for agentic AI
Celonis is probably not a household name, but the German process mining company helps large organizations uncover inefficiencies in business workflows and optimize operations using data and AI.
It's a tall order, but it could also be a key piece of the agentic puzzle. If the idea is to reinvent enterprise workflows for an agentic world, it requires understanding these workflows and where bottlenecks happen. This is where Celonis comes in, translating complex operations into an understandable visual map.
Company founder and CEO Alex Rinke believes process and intelligence are inextricably linked. “In our view, there's no AI without PI (process intelligence),” Rinke told FastForward.
He points out that consumers and businesses have profoundly different requirements when it comes to AI. “The enterprise needs the inputs to actually have the right context, and it needs to have the understanding of the process,” he said. But it doesn’t stop there. “You also need observability on the back end to really make sure you understand how the new process is working,” he said.
That is even more important when it comes to agents operating independently of human input. You have to be able to audit them.
Grounding agents with a graph
Last week at the company's Celosphere customer conference, it announced several major enhancements to the Celonis platform. This included updates to the Intelligence Graph, the company first announced two years ago. The graph helps companies see connections between different parts of workflows and across them, which could have tremendous value with agents.
"We now have over 60 different processes and combinations out-of-the-box in our process intelligence graph," Rinke said. This includes connectors to common enterprise software like SAP, Salesforce, Snowflake and Databricks — and the company is seeing momentum for this approach. "We tripled the number of customers to over 300 now that use our process intelligence graph, and I'm talking large enterprise, like Global 2000-type customers," he said.
Holger Mueller, an analyst at Constellation Research, says all AI needs grounding, a link to real-world data to ensure its outputs are reliable and relevant, and the graph is the way that Celonis is doing that. “A process intelligence graph can provide grounding for an agent, but it is more than that,” Mueller said. “It also can tell agents what needs to be done, turning agents from probabilistic to deterministic for the right processes.”
He says that execs want deterministic processes with high quality outcomes. The observability piece can ensure that happens. “You still need the observability across this new [agentic] process to really understand if it’s performing, where it is in the flow, where there’s an issue, and if you need to reconfigure it,” Rinke said.
That may not have been the original idea when Rinke launched Celonis in 2011, but as the enterprise tech landscape has evolved, being able to understand and observe how work flows through an organization has become increasingly important.
Along the way, the company has raised $2.4 billion, per Crunchbase data. Its most recent funding came back in 2022, $400 million at a $13 billion valuation. This was a lofty number at a time when big valuations were being tossed around liberally, but agentic has brought a new emphasis to understanding and redefining business processes, not just to root out inefficiencies, but to transform them in this new agentic context.
Featured photo by Unsplash Community for Unsplash+
