Celonis provides AI agents with process intelligence data to improve their operational context

Celonis SE is ramping up its efforts to help organizations reshape their operations with artificial intelligence, today announcing new capabilities within its eponymous Process Intelligence Platform at its Celosphere 2025 annual user conference, positioning it as the foundation for AI-driven operations.

Celonis, with offices in Munich and New York City, is the creator of a process mining platform that helps companies identify opportunities to increase their operational efficiency. The platform can be used to detect if a retailer has more of a particular item than needed to meet customer demand, for example, while a shipping company can use it to predict and mitigate delayed deliveries.

The Celonis Process Intelligence Platform collects and analyzes massive amounts of operational data from business applications, devices and systems. It supports more than 1,000 connectors, reducing the need for teams to create custom code to connect those apps to the platform.

Once the data is collected, Celonis enriches it with business context to create a digital twin of an enterprise’s business activities. This is called the Process Intelligence Graph, which allows organizations to analyze, design and operate AI agents and other autonomous processes.

Operational context for AI agents

With today’s update, Celonis extends the Process Intelligence Graph to support additional data types from newer data sources. Celonis Data Core is a new offering that allows companies to integrate data lakes such as Databricks and the Azure Data Lake with a simple, zero-copy and bi-directional connection.

The integration with Databricks is significant as it means that live data stored on that platform can be used to create more comprehensive digital twins of business operations. Customers can feed intelligence from the Celonis platform back to Databricks’ Agent Bricks platform and create production-quality AI agents optimized with their own operational data.

To support the creation of more capable AI agents, Celonis has also added new capabilities for connecting desktop actions such as keystrokes, mouse clicks and screen scrolling with business processes that support improved task mining and AI-driven task detection. Additionally, companies will be able to integrate their architectural blueprints so that AI agents can understand which systems are used for which activities, so they can use the right tool for each task they need to automate.

All this is done to support more advanced and configurable AI-driven operations, Celonis said. To enable these activities, the company has introduced new object-centric process mining tools to help identify problems at key intersections of processes, where problems often occur. This ensures the smooth transition of data between business activities, such as the transportation, storage, packaging and shipping of finished products.

Meanwhile, the company is expanding the capabilities of the Celonis Orchestration Engine to support AI agents as well as people and systems. It works by turning operational insights into automated task and workflow execution.

Finally, Celonis said it is launching the industry’s first Model Context Protocol server specifically designed for process intelligence, which is essential for providing AI agents with the dynamic operational context they need to make appropriate decisions and complete their work effectively.

Constellation Research’s Holger Mueller said process intelligence could be the key to ensuring AI works as intended, something that has challenged many companies so far. He explained that their reliance on their own data causes all kinds of problems for AI agents in particular, as the quality of this information is often questionable, meaning their agentic outputs reflect these inaccuracies.

“Celonis offers a promising solution to this by carefully analyzing each organization’s business processes through a lossless approach that leverages the latest innovations in object-centric process modeling,” said the analyst. “The process intelligence graph is a valid and workable alternative to traditional fetch-enhanced generation.”

AI agents in action

To demonstrate what the platform is now capable of, Celonis has worked with several partners to develop a range of ‘composable’ AI agents that can be quickly customized and adopted by customers. These include Rollio Inc.’s new Process Collaboration Agent, which is designed to automate the management of information technology services, procurement and other common business processes. It is said to resolve process exceptions immediately by pulling the right people and context from the Celonis platform.

Meanwhile, Trullion Inc.’s new agent helps. automate the complex, manual work required for lease accounting tasks. The new Agent Miner app from Bloomfilter Inc. is designed to understand and control the behavior of AI coding agents.

Celonis said some early adopters have deployed AI agents that leverage the process intelligence and insights to improve their operations. Mercedes-Benz Group AG has implemented AI agents to speed up decision-making and improve delivery times, while Vinmar International Ltd. claims to have turned its cash-to-order process into a fully automated operation.

Celonis Chief Product Officer Daniel Brown said companies need to take a more structured approach to make agentic AI work at the level of accuracy and reliability needed. He said this means companies must first identify the right use cases and then redesign their business processes in a way that enables AI automation.

“(The final step is to) orchestrate the agents alongside your people and existing systems,” Brown explains. “Our enhanced capabilities enable customers and their partners to build AI solutions that take their operations to unprecedented levels of efficiency and flexibility.”

Image: SiliconANGLE/Meta AI