Introducing the Celonis AI Annotation Builder: Unlocking the power of your data with AI

In today’s enterprise, data is king—but making sense of it can be complex. Some data is neatly structured, like transaction records, while other data—such as free-text service tickets or emails—requires interpretation. But beyond structure, the real challenge is applying business logic to data efficiently. Enter the Celonis AI Annotation Builder—a no-code tool that uses GenAI to reason through data, whether structured or unstructured, and generate decisions and action recommendations. By allowing users to define rules in natural language, it lowers the barrier to automating decisions, making AI-driven insights and process automation more accessible than ever.

Making AI and automation work for enterprise processes

Technologies are only as good as the input they get. For example, if you want an AI agent to automatically and accurately handle credit blocks, it needs to understand how your business actually operates and how it can operate better. It needs Process Intelligence. Celonis delivers Process Intelligence by combining process data (i.e., the objects, events, and relationships within your business systems) and business context (i.e. the process models, business rules, benchmarks and KPIs specific to your business) to create a living process digital twin of your organization—the Process Intelligence Graph (PI Graph).

The PI Graph shows you where value is hiding in your processes, and the Celonis Platform (with the PI Graph at its center) gives you the tools to capture it. That could be through a platform capability (such as the Process Explorer), a Celonis-built app (such as the Customer Service Control Center), a partner-built app (such as Celonis for Software Development Lifecycle Management - powered by Bloomfilter), or a custom-built AI copilot, assistant or agent.

One of the most exciting use cases for Celonis is AI-driven, intelligent automation.

Let’s say you’re managing a high-volume customer service operation. Your inbox is flooded with free-text service tickets, each carrying crucial but messy information. Some tickets are urgent, while others can wait. Manually triaging them is time-consuming and inconsistent. And, if you build an automation that incorrectly categorizes tickets as low priority that should be a high priority (or vice versa), you’re increasing rework for your customer service teams and will likely end up with angry customers. So, how do you extract the valuable signal from the noise of unstructured data? With Celonis’ AI Annotation Builder.

Announced at Celosphere 2024, the AI Annotation Builder leverages generative AI (GenAI) to interpret structured and unstructured data, apply business rules, and generate actionable insights. Essentially, you give it information and guidance in natural language on how you want it to make decisions and it uses its ability to process natural language to determine which decision makes the most sense. It produces AI annotations that enrich your data, making it more valuable for analysis and automation.

With the AI Annotation Builder, you simply define your prioritization rules in natural language—no data science skills required—and let AI classify tickets into high, medium, and low priority with clear reasoning for each decision.

How the AI Annotation Builder Works: From data to smart decisions

The AI Annotation Builder transforms raw data into structured insights in a few simple steps:

Step 1: Specify relevant data from Celonis data model

Identify the objects in your Celonis data model that you want to generate AI annotations for (e.g., purchase orders, customer service tickets, invoices). In the example below, the user has selected Customer Service Tickets (ID, Category, Description, Time Open).

Step 2. Write prompt including selected data

After specifying your data, you write out the instructions for the Large Language Model to generate the AI annotations. For instance, “If a sales order includes words like ‘urgent’ or ‘expedite,’ classify it as high priority.” Using the same example as Step 1, here the user asks the model to provide the category and sentiment of each ticket based on its description.

Step 3: Specify expected outputs

Next, you specify the expected output of the Large Language Model, for example categories and explanations for AI-generated annotations. In this example, the user specifies that they expect the LLM to return the category and sentiment for each ticket.

Step 4: Execute agent to generate annotations

With the configuration complete, you can execute the assistant. The LLM will process the instructions for each case (i.e. ticket) that have been filtered on and will return the respective outputs. These outputs are then stored in Celonis. Users can review the classifications and refine prompts as necessary.

Step 5: Leverage the annotations throughout Celonis

The generated annotations can be used like any other data in Celonis. You can use it to conduct retrospective analysis, create operational human-in-the-loop apps or even trigger automations.

Companies across industries are already harnessing the AI Annotation Builder to drive efficiency and intelligence into their operations:

• Finance & Order Management: AI categorizes credit blocks, analyzing order histories and financial data to recommend block removals or holds—cutting manual reviews by up to 5x​.

• Procurement: Large enterprises automate purchase order confirmations, analyzing supplier emails to flag missing approvals before they become bottlenecks​.

• Supply Chain: AI reads material transfer requests, assessing risk factors to classify movements as high, medium, or low risk—improving decision speed and reducing supply chain disruptions​.

• Customer Service: AI scans support tickets, assigning sentiment scores and categorizing requests, helping agents resolve issues 30% faster​.

Why this matters: AI-Powered Process Intelligence

The AI Annotation Builder is a game-changer for unlocking the full potential of your data. By enriching the data with AI-generated intelligence, it empowers teams with data-driven insights they can trust, enabling more informed decision-making and process automation.

And the best part? No coding required.

Ready to make your unstructured data work smarter? Contact us to learn more about the Celonis AI Annotation Builder.

Bill Detwiler

Senior Communications Strategist and Editor Celonis Blog

Bill Detwiler is Senior Communications Strategist and Editor of the Celonis blog. He is the former Editor in Chief of TechRepublic, where he hosted the Dynamic Developer podcast and Cracking Open, CNET’s popular online show. Bill is an award-winning journalist, who’s covered the tech industry for more than two decades. Prior to his career in the software industry and tech media, he was an IT professional in the social research and energy industries.