Next article
Operationalizing AI with PI: Five common AI use cases that businesses face
Celonis Process Intelligence Day returned to New York City on June 3, bringing together process professionals, business leaders, and AI experts at the Javits Center.
During the event, Senior Applied AI Engineers Grif Banta and Kiko Cortez explained five common AI use cases that businesses face during their roundtable Operationalizing AI With Process Intelligence for Real Business Impact. These are the areas within your enterprise’s processes where AI can be leveraged to help drive meaningful impact.
Grif Banta (Left) and Kiko Cortez (Right)
Let’s take a deeper look at these use cases and how the Celonis Process Intelligence platform has helped customers in real-life scenarios:
1. Unstructured Data Processing
Consider your current systems: Imagine how often you rely on comments collected by a webform, use free text fields and long form inputs on a PDF, or even manage a critical business process predominantly through an email thread. A lot of valuable information resides within these digital assets, but making it useful, especially for AI, often involves the labor-intensive, manual categorization and tagging of specific keywords or details.
With Generative AI (GenAI) however, we can make sense of unstructured and structured data alike. Celonis helps turn messy, unstructured content into structured, analyzable data using the AI Annotation Builder, a no-code tool that uses GenAI to reason through data (both structured and unstructured) and generate decisions and action recommendations.
Celonis’ AI Annotation Builder produces AI annotations to enrich your data, making it more valuable for analysis and automation.
2. Conversational Questions & Answers
Every day, you probably find yourself responding to similar, repetitive questions. According to Banta, these inquiries can be bucketed into three categories:
Analytical Inquiries: These are high-level questions where a person is interested in how their process is running. These are often asked by managers, directors, executives, and others who own the process itself and want to manage the metrics and identify how they’re trending.
Operational Inquiries: Questions like these come from people on the front lines of your business processes–those that, oftentimes, have to go through multiple systems to identify the current status of where a payment is or why an order did not ship.
Process Guidance Inquiries: Unlike the other types, these questions are not necessarily focused on your data specifically, but rather how your processes run.
Celonis Process Copilots let you talk to your processes right from your regular workspaces and get answers fast about what’s working and what blockers need your attention.
A Celonis Process Copilot recommended questions
Process Copilots are GenAI chatbots that let you ask natural language questions and get real-time answers about key performance indicators (KPIs), status, or process steps–right inside Celonis or in tools like Teams and Slack using the Chat application programming interface (API). Celonis’ Process Copilots simplify and accelerate the process of identifying value opportunities across your business.
3. Decision Automation
Not all process decisions fit neatly into a standard business rule. Many repetitive decisions within your enterprise's processes rely on human decision making and aren't entirely predictable or straightforward. This makes them difficult to automate with traditional methods.
Consider handling credit blocks. When a credit block is set, a credit manager typically reviews several pieces of information to decide on the next steps. Whether they’re examining sales orders or a customer's payment history, the decision to keep, remove, or internally escalate a block is often subjective, relying on your company's existing business guidelines–leaving individuals to make a decision that is often subjective.
This use case is exactly where you can start to leverage GenAI applications and the reasoning models behind them to create solutions that provide recommendations about what to do with the block and the reasoning behind the suggestion. For example, a leading manufacturer of design and architectural surfaces, used Celonis to create an AI Assistant for credit block management. The assistant analyzes each blocked order, pulls together relevant data, such as order value and credit information, and makes a recommendation on what actions to take (along with its reasoning for the recommendation). Credit managers can accept or reject the recommendation with the click of a button and even provide feedback to the assistant to learn from.
Celonis helps automate judgment-based decisions like this by feeding third-party agentic systems with Process Intelligence–via Process Intelligence APIs or directly through the AI Annotation Builder.
4. Prediction & Forecasting
From avoiding inventory stockouts to accurately predicting customer deliveries, planning and forecasting are critical business functions. Think about how many times you reacted, for example, to a delay, a cancellation, or an escalation, when all of the signs were there that this could potentially have happened, or confidently predict your sales forecast for the following month.
These are scenarios that Celonis customers are familiar with, and Cortez says we can bucket them into two categories:
Outcome-Based Predictions: These are predictions where we want to forecast the likely result of an event or situation–like which orders will be delivered late based on activity data, which customers will churn based on behavior and activities, which tickets are likely to be escalated.
Time-Series Forecasting: In these problems, we will monitor the change of a variable over time, such as estimating invoice payment volumes per month to optimize cash flow or the number of expected service requests/tickets.
To tackle these use cases, Celonis developed the Prediction Builder–allowing you to train and deploy outcome prediction models, so you can anticipate issues like late deliveries before they happen and take proactive action.
5. Data Deduplication & Matching
Duplicate or inconsistent data can clog up your processes–causing rework, confusion, delays, and other negative consequences for your business. According to Cortez, there are two types of use cases that companies encounter the most:
Spelling Matching (Fuzzy Matching): This challenge often occurs with similar, but not identical, entities–like vendor names. For instance, if you have five vendors with similar abbreviations but different payment terms, it's tough to quickly pinpoint the right information. Fuzzy matching helps solve this by detecting duplicates in invoices, customer data, and vendor master data.
Meaning Matching (Semantic Matching): Unlike fuzzy matching, sometimes the problem lies with entities that are actually identical but appear different. This occurs when the same entity exists in various formats or even different languages. These variations, though minor, can create significant data integrity issues, leading to confusion and errors. Semantic matching helps detect material master data duplicates, clustering free-text POs, and matching customer complaints to those that refer to the same issue.
Celonis helps identify and cluster duplicate or similar records with solutions like the Duplicate Invoice Checker App (part of the Celonis Finance Solution Suite), which prevents overpayments and duplicate payments by detecting and managing duplicate invoices using AI-driven intelligent matching and real-time ERP integration.
Teaching AI how business works
AI may be the biggest buzzword in tech today, but PI gives it the context it needs to operate in the real world, solving use cases like these fast and efficiently.
In the second episode of our Trust the Process podcast, we discussed how PI fuels AI–giving it the context it needs to work in the real world. Because without it, AI is just guessing. But with it? It knows exactly where it’s going–and how to get there.
Listen to Trust the Process as we explore why AI without Process Intelligence is like navigating with an outdated map–and how real-time process context is the key to making AI truly work.
Because when processes work, everything works.
