Celonis and Microsoft Host AI Lab, NYC: Eight real-world opportunities, seized in real time

In early June 2026, another group of innovative Celonis customers joined us in New York for the latest in our series of AI Lab hackathons. It was an incredibly productive day spent realizing real-world business opportunities, in real time, by building AI solutions on the Celonis Platform.

This time, the always highly collaborative event was co-hosted by Microsoft. Alongside their dedicated Celonaut value engineer, the teams had access to Microsoft experts, ready to help them take full advantage of solutions like Microsoft Fabric, Microsoft Teams, and Microsoft Copilot.

What business opportunities did the teams identify? And what solutions did they build? Here’s a round-up of all the AI-driven innovations.

Global pharmaceutical company: The “Hold Guardian” Agent

Opportunity: For any large company, leaving support incidents on hold can ultimately impact a wide range of KPIs, from breach of SLA response rates to average total cycle time. Simple automation can support the identification and escalation of on-hold tickets, but if the reason for the hold is very generic – for example, a request for further information (RFI) from the user – its power is limited.

Solution: The team created a “Hold Guardian” agent that uses an LLM to analyze free text support ticket fields, improving the classification of on-hold incidents. Based on this enhanced classification, another agent escalates the tickets appropriately, using Celonis Action Flows. When an incident is flagged for urgent attention, its owner receives a pre-populated form, as well as the AI’s rationale and recommended next steps.

Global pharmaceutical company: The “Routing Mismatch” Agent

Opportunity: In an industry with interlinked and tightly regulated workflows, there’s rarely such a thing as an isolated issue. More often, problems permeate across processes and create challenges downstream. AI agents can help to keep operations running smoothly, but to make accurate decisions, they need to understand the operational reality of the business.

Solution: A second team from this global pharmaceutical company worked to demonstrate how, by making cleansed and structured data from the Celonis Platform accessible through the company’s Model Context Protocol (MCP) gateway could improve outcomes. The result: it could build more accurate agents, faster. In just two hours, the team built an AI agent to resolve incident routing mismatches, a common cause of both on-hold incidents and duplicated work. The agent flags cases that are (or are at risk of being) incorrectly assigned, helping to prevent routing mismatches from snowballing into serious issues.

Global life sciences and diagnostics company: The “Supplier Intelligence” Agent

Opportunity: Whether it’s making specialized medical imaging instruments or at-home test kits, a life sciences production line can be brought to a standstill by late deliveries from suppliers. Even with commodity managers devoting significant amounts of time to researching and assessing risk, it’s possible for sudden changes in supplier stability or reliability to disrupt production, delay customer orders, and put sales revenue at stake.

Solution: The team built a Microsoft Azure agent to conduct automated web searches, scanning the internet for risk indications related to key suppliers. The agent synthesizes this information to provide a multi-dimensional risk assessment (across key pillars of risk, such as geopolitical, ESG, or operational risk), which commodity managers can view within their central vendor briefing sheet. But the team didn’t stop there. They also saw the opportunity for a Microsoft Copilot agent that combines this new insight with process intelligence, and existing supplier information such as on-time, in-full (OTIF) performance and price variance metrics. Commodity managers can use this agent to query, analyze, and deep-dive into vendor risk assessments, all while enjoying a holistic view of vendor performance.

Combined, these innovations promise to reduce manual effort, reduce production planning and manufacturing overhead from material shortages, while unlocking significant improvements in OTIF delivery from critical suppliers.

Global paints and coatings company: The “Invoice Hold” Agent

Opportunity: The Accounts Payable (AP) team at a global manufacturer may create thousands of new invoice holds in a single day. Investigating and resolving those holds can be extremely time consuming, requiring dozens of dedicated personnel and correspondence between multiple teams. The result? Significantly delayed payments and high outstanding payable balances.

Solution: The team built a Microsoft Copilot agent to help its AP analysts identify and resolve invoice holds. Interacting with the analysts through Microsoft Teams, the agent surfaces relevant information and provides recommended actions, generated by the Celonis Annotation Builder. Actions are automatically recorded back into Celonis as they are taken, enabling the next step: mining the agent’s activity to measure its impact and optimize performance. The team expects its solution to improve productivity (thanks to fewer touches per invoice), while helping to ensure timely payment and maintain strong vendor relationships.

Global vehicle rental and mobility solutions provider: The “Operational Location” Scorecard

Opportunity: When you’re working in the subrogation department at a global vehicle rental company, having visibility across thousands of operational locations is a multimillion-dollar issue. But inaccurate reporting and inadequate documentation from local agents can lead to missed vehicle damage collections and lost revenue.

Solution: The team built a solution to improve data and documentation submission, identifying over a hundred new levers that affect collection, and generating an operational score for every location. This meant using Celonis Annotation Builder to pinpoint and categorize discrepancies in vehicle mileage information, and combining Celonis with the company’s own automation platform to streamline document tracking, and the capture of evidence needed to support successful claims. By the end of the hackathon, the team had not only reduced the manual effort involved in location-level analysis, but it had also significantly optimized its collection rate.

Global medical technology company: Intelligent Payment Reminders

Opportunity: Whatever business your company is in, one thing’s true: if your customers don’t pay on time, your cash flow suffers. But if you can effectively predict delayed payments, and take timely action to prevent them? Then, you can dramatically increase your working capital.

Solution: This team zeroed in on a very specific use case: reducing late payments for a single division of a single business unit. With the help of Celonis Prediction Builder, its solution flags upcoming invoices that have a high probability of becoming overdue based on the customer’s prior payment history. Whenever the flag gets raised, a Celonis Action Flow sends a message to the Collections team, giving them the option to escalate to a Collections Manager, or send an automated dunning email to the customer – with a form to log a promise-to-pay date or dispute. The solution promises to deliver gains in both productivity and working capital, which could grow exponentially when scaled to other parts of the business.

Multinational aerospace corporation: The “Material Dependency” Agent

Opportunity: When disruption strikes your manufacturing process, understanding its internal dependencies is the key to mounting a smart, strategic response. Building an aircraft is a prime example. If one of the 15,000 – 20,000 required parts has broken or failed to arrive, do you roll the production line, and accept the consequences of having to incorporate it later, at a different workstation (or ‘travel work’)? Or is it better to hold the line, pay for next-day delivery of a replacement part? These decisions have a significant impact on productivity.

Solution: The team wanted to create a clear, data-driven view of the downstream impact of materials shortages – one that would enable faster, better-informed decisions, reduce downtime, and maximize labor productivity. Using Celonis Process Copilot, they created an agent that builds on the company’s existing precedence network. (This, as one team member put it, is like their “Lego manual for building a plane”). The agent can assess the impact of a part shortage, letting decision-makers drill down into the additional hours of work created at subsequent stations, and helping them to make the right call on whether to hold the line.

American food manufacturing company: The “Ringer Orders” Agent

Opportunity: The push to recognize revenue in the right period is one every Order-to-Cash (OTC) team knows well. It can easily become a major time sink: analysts often manually collate orders expected to ship after period close, then work (and rework) the results in a single shared spreadsheet.

Solution: This team built a solution to strengthen period-end revenue realization. Their “Ringer Orders” agent helps to maintain a dashboard for tracking orders that are at risk of shipping out of period. By prepopulating or recommending reason codes, it saves analysts time, and reduces the likelihood of accidental overwrites and rework. Another agent identifies high-value orders with a credit hold, and shares relevant context (like customer credit history) with analysts, to accelerate decision-making and order release. It’s a solution that has the potential to not only improve working capital, but deliver significant productivity gains for the OTC team.

Explore the past and future of AI-driven innovation

We’ll be bringing AI Lab events to even more cities in the coming months – watch this space for details. In the meantime, you can explore the growing history of AI-driven innovation with our round-ups of previous events: