Heathrow executive says AI now checks every invoice, up from 3%

Heathrow’s accounts payable team once manually checked about 3% of invoices. It now checks all of them—around 4,000 a month—using AI, according to Alan Petrie, its head of Corporate Data and Analytics. In an Oracle Blogs interview published October 13, 2026, he described that shift alongside faster policy searches and plans to coordinate more AI agents.

The airport’s analytics foundation predates this interview. In a May 21 announcement, Oracle said Heathrow had implemented Fusion Data Intelligence for finance and human resources, alongside its Fusion business applications. The stated purposes included ready-to-use analytics, controlled access to sensitive data and embedded AI capabilities. Petrie’s new account gives concrete examples of how his team is using that foundation.

From spot checks to line-by-line review

The invoice workflow starts with Oracle OCI Document Understanding, which makes invoices machine-readable, including documents spanning multiple pages. Heathrow then checks each line against the corresponding purchase order—the record of what it agreed to buy. The process is designed to verify that the amount being paid matches the order, including value-added tax, or VAT, where appropriate.

Anomalies are automatically flagged. The workflow also checks invoice VAT numbers and supplier registration details against Companies House. Results appear in a Fusion Data Intelligence dashboard, where staff can see errors and investigate them. Petrie described significant time and money savings without putting a figure on either, and said the team had identified similar applications in other departments.

Faster answers, with accuracy still imperfect

Finance is also trying AI for a different job: answering procedural questions using policy documents and emails. Rather than search those materials manually for hours, staff can get an answer in about 20 seconds, Petrie said. He called the results very good, but acknowledged they were not always fully accurate. That qualification sits alongside the claimed speed gain.

Petrie sees scope beyond Finance because other departments also have procedures to follow, with HR policies an obvious example. Elsewhere, he described two uses already underway that bring AI and analytics into employees’ everyday work:

• Staff questions: AI agents embedded in Fusion Applications answer questions about remaining leave and workplace policies. Petrie expects the applications team to add more of these agents.
• Absence trends: A dashboard combines absence data held in Databricks, a legacy rostering system and HR data from Fusion. Departments can examine absences, lost shifts and comparisons with other teams. Petrie highlighted security, where Heathrow employs more than 4,000 staff and absences affect operations across the airport.

The next move is coordinating agents

Petrie’s next ambition is to make analytics conversational. Instead of knowing which dashboard contains an answer, a user would ask for a view—such as average departmental salaries over three years—and have the visualization created on demand. He wants that experience connected to the agents already embedded in business applications, rather than leaving operational questions and trend analysis in separate interfaces.

His proposed concierge AI would take a request and pass it to whichever agents were needed, potentially involving several agents communicating with one another. That creates a management task: ensuring agents work properly and directing each use case to the right one. Petrie said Heathrow initially plans to use Oracle AI Data Platform to manage agents connected to Fusion Data Intelligence and make them available elsewhere in the company.