How Make AI Agents help Celonis lower annual expense auditing costs

• Industry: IT – Data Analytics
• Use case: Financial Reporting
• Country: Global
• Company size: 1,000-5000 employees
• Apps used: monday.com, Google Vertex AI (Gemini)

Celonis, a global leader in process intelligence, processes thousands of employee expense report lines monthly across its worldwide operations. Facing a significant five-figure annual cost for third-party auditing tools and limited T&E (Travel & Expense) team capacity, the company developed an expense report auditing automation. By integrating Make with Vertex AI (Gemini) and monday.com, Celonis created a human-in-the-loop system. Since they already had Make and monday.com licensed, the additional investment was only $150 for Vertex AI, decreasing annual auditing costs by 99.7% while improving accuracy and user experience. Joe Graboff, Principal Architect of Integrations at Celonis and the leader of this automation initiative, explains how Make’ AI transformed their expense auditing process from a costly manual bottleneck into an intelligent, scalable system that processes thousands of reports with impressive precision.

The challenges

Over the years of global growth, Celonis's finance operations encountered several bottlenecks:

Unsustainable auditing costs

With employees submitting thousands of expense report lines monthly, the company relied on a third-party auditing tool costing tens of thousands of dollars annually, plus integration support. This cost didn't scale efficiently with the company's needs.

Requiring high accuracy with a limited team capacity

The T&E team consisted of just a few people managing thousands of submissions. Policy violations, missing receipts, insufficient evidence, and potential misappropriation of assets required meticulous review, but manual processes couldn't keep pace. The team needed perfect accuracy when handling employee funds, making automation challenging.

Need for rapid adaptation to evolving policies

Travel and expense policies naturally evolve with changing business needs, new compliance requirements, and market conditions. The existing system made policy updates tiresome, requiring lengthy implementation cycles that slowed the team's ability to respond to business changes.

The solution

Celonis used Make to build a sophisticated AI agent workflow that transformed their expense auditing process entirely.

Expense report auditing AI agent with human oversight

Celonis built a human-in-the-loop expense report auditing system using Make AI Agents. When employees submit expense reports in Workday, Make receives the data via webhook and creates structured items in monday.com, giving the T&E team a centralized dashboard to review all reports.

When an expense report arrives, Celonis’ agent first performs initial policy checks like validating item amounts, dates, expense age, and category-specific rules like meal and event spending limits.

For items needing deeper analysis, the AI Agent organizes all the relevant information (policies, receipts, and employee data) and sends it to Vertex AI (Gemini), which extracts details from receipts and evaluates them against company policies. Make then performs additional validation checks, like verifying currency matches.

The solution comprises over a dozen interconnected Make scenarios with automated checks running every few hours. If no issues are found, reports get auto-approved in Workday and routed to managers. If violations surface, items move to "review needed" status in monday.com for human evaluation.

Maintaining a self-service policy management in monday.com

Recognizing that policies must adapt quickly, Celonis built policy management directly into the monday.com interface. The T&E team maintains policy lists, thresholds, and approval criteria within monday.com and can sync these changes to Make's data store at the push of a button for an immediate update. No more IT bottlenecks of logging into Make, searching for the scenario, editing the prompt, and replacing the policy text manually every time the policy shifts. The system also performs automatic weekly syncs every Sunday, ensuring the latest policies flow into the Gemini prompts without technical intervention.

Streamlined communication with custom send-back interfaces

For reports requiring employee revision, Make generates custom webhook response pages that function like standalone websites but live within the Make platform. The T&E team can select from pre-populated AI text templates explaining specific policy violations, then send professional feedback directly to employees via email or within Workday. This eliminates the need to draft repetitive messages manually.

Building for scalability and reusability

The modular architecture Celonis created extends far beyond expense auditing. The team designed scenarios with reusability in mind, making it simple to adapt the code for new AI use cases.

They're already deploying similar workflows for purchasing card verification audits and legal document repository projects that categorize and organize PDFs, order forms, and Master Service Agreements using the same Make and Vertex AI foundation.

The results

Celonis’ expense report auditing AI Agent brought impressive outcomes across cost, efficiency, and user satisfaction:

• 99.7% reduction in AI processing costs – from tens of thousands for a specialized tool to just $150 per year for Gemini on top of already existing licenses
• Processing efficiency of approximately 1,000 documents per dollar with direct LLM integration
• Near 100% accuracy through context engineering and human-in-the-loop validation
• Successful 60-70 day parallel testing period confirming the agent’s spotless performance before full deployment
• After just 30-40 days, the T&E team felt confident to let the solution work even more independently

Make AI Agents: Redefining finance operations

Celonis and Make proved an ideal partnership for transforming finance operations. Make's visual workflow builder, direct Vertex AI integration, and monday.com connectivity let Celonis build a sophisticated AI agent system in-house at a fraction of previous costs.

The T&E team's enthusiastic adoption, including their growing comfort with expanded auto-approval, validates that human-centered AI design creates both operational excellence and user satisfaction.

This success has opened doors across the organization, with purchasing card audits and legal document processing already in development.

As Joe puts it, the expense report auditing AI agent is their "lighthouse project", illuminating the path for countless AI-driven workflows to come.