Next article
AI innovation on display: Insights from Celonis AI Lab New York City
A day after Celonis Process Intelligence Day (PI Day) in New York City, and a few weeks before AI Appreciation Day on July 16, six Celonis customers gathered at our New York office for AI Lab–our ongoing series of one-day AI hackathons. During the event, customers worked alongside Celonis solution experts to design and build impactful solution prototypes using our latest AI products–like Process Copilots, AI Annotation Builder, and Prediction Builder.
At the end of an intense day of solution development, teams had five minutes to showcase their projects. Each presented the business problem they aimed to solve, highlighted the potential impact of their solution, and demonstrated their creative use of AI and Celonis products. The AI Lab wasn't just about technology–it was about collaboration, competition, and co-innovation.
Here’s a round-up of what our teams built:
American Financial Technology Company
Challenge: Visitors to the company’s website use the “Contact Us” form for all manner of requests, be they complaints, customer service issues, or business inquiries. This forces Business Development (BD) teams to manually shift through and reroute inquiries, leading to delays and errors.
Solution: The team’s solution leverages the AI Annotation Builder to automatically classify and route submissions to the appropriate teams. For submissions sent to the business development team, the solution also connects to the company’s Salesforce instance, providing team members with the individual’s company information as well as the information from the “Contact Us” form. And by using a Celonis dashboard, the company can track performance, route accuracy, and enable Business Assistants to triage edge cases.
American Multinational Food Manufacturer
Challenge: When orders fail due to stock shortages, teams must manually investigate if alternative depots can fulfill them. During peak seasons, they frequently override default depot assignments to fulfill orders, rather than using existing technical solutions to direct stock.
Solution: The team used the AI Annotation Builder to streamline this process, reducing unnecessary stock transfers and manual effort. Their solution achieves this by enabling faster, smarter depot override decisions through embedded business rules. Those using the platform can access an operational view of these annotations directly within the Sales Order Execution app, allowing them to see recommended depots and execute necessary overrides for an order.
Major American Food Processing Company
Challenge: The company’s current deductions management and recovery process is time-consuming and tedious, requiring teams to pull data from multiple sources to process deductions. This fragmented workflow makes it difficult to quickly distinguish between valid and invalid deductions, delaying recovery of invalid claims and reducing operational efficiency.
Solution: Using the AI Annotation Builder, the team created a solution that extracts and structures data relevant to the deductions validation process from unstructured sources, such as emails and PDFs. Employees no longer need to manually pull data from different systems, streamlining the deductions validation process. The team also added AI-based validity recommendations to reduce decision latency and further accelerate the process.
American Telecommunications Company
Challenge: This company’s tasks are grouped into milestones that represent a collective group of tasks for one purpose, leading to manual categorization that can be time-consuming and inconsistent. Many tasks end up being “unmapped” because of this, limiting visibility and hindering order progression analysis.
Solution: Tasks can be classified into milestone categories using the AI Annotation Builder, leveraging LLM-based reasoning for intelligent automation. The team also created a solution using a validation dashboard to track annotation coverage, top unmapped patterns, and reasoning quality, allowing users to explore performance by task name, performer group, and automation flag.
International Development Bank Group
Challenge: This bank's Accounts Payable (AP) team spends a lot of time manually identifying inaccuracies and insufficient information in PO invoices. Their current process lacks transparency, making it difficult to track impacted invoices, pinpoint vendors submitting inaccurate POs, or assess the impact on payment timeliness.
Solution: By using the Analytical Process Copilot, the team created a solution that identifies and quantifies the impact of POs needing rework or approval. With this, the Operational Process Copilot also drafts notifications for vendors submitting inaccurate POs that require manual AP effort. Additionally, the team used the AI Annotation Builder to analyze free-text comments to identify rework reasons and establish initial PO priority.
Winner: American Multinational Investment Bank
Challenge: For this investment bank, the Know Your Customer (KYC) case resolution is currently bottlenecked by manual sorting and prioritization of outreach cases. A lack of predictive insights creates significant inefficiencies, making it difficult to consistently meet their one-year Service Level Agreement (SLA) and leading to ongoing challenges with workload planning.
Solution: By leveraging the Celonis Prediction Builder, the team created a solution that can identify the probability of a KYC case requiring outreach. To optimize team workflows, the team made a monthly case schedule to align with SLA constraints. This solution also provides a user-friendly operational interface that surfaces these predictions, guiding the actions of Customer Service Representatives (CSRs).
Up Next for AI Labs at Celonis
The solutions don’t stop here. Register your interest for our AI Lab at Celosphere, our ultimate Process Intelligence event. Our Applied AI team will review all submissions and follow up to confirm your participation and share next steps.
All AI Lab participants need a Celosphere ticket.
