As payments become increasingly embedded into corporate travel, fraud prevention has become a balancing act between stronger security and a seamless user experience. AI agents, machine learning models, and real-time payment intelligence are reshaping how businesses detect fraud, automate expense management, and navigate increasingly complex global payment ecosystems. 

Ahead of Merchant Risk Council San Diego 2026, we spoke with Xiaoxiao Wang, Director of Fraud Risk Strategy and Data Science at Navan, about how Navan is using AI to strengthen fraud controls without adding friction for legitimate customers. She shares insights into Navan Connect, agentic AI through Navan Cognition, the challenges of cross-border payments, and why fraud prevention is moving beyond traditional rule-based systems toward adaptive, auditable AI. 

Read the full interview below.

BMW: Could you give us a high-level overview of Navan, and describe your role within the company?  

XW: Navan is the all-in-one travel and expense platform for businesses of all sizes, from high-growth SMBs to global enterprise clients. We strive to make the travel and expense experience delightful for users, effortless for admins, and financially optimised for the entire company.  

As Director of Fraud Strategy and Data Science, I design and scale automated, end-to-end fraud risk controls across Navan travel and expense products using AI and advanced ML techniques. My team’s mission is to keep bad actors off the Navan platform while maintaining a frictionless customer experience for legitimate business.  

BMW: How do Navan’s payments and card infrastructure influence the way you approach fraud risk, and how important are partnerships in building effective fraud prevention?  

XW Navan offers flexible payment options to businesses, including physical corporate cards, virtual travel cards, global reimbursements for out-of-pocket expenses, and Navan Connect, which allows companies to link their existing corporate card programs.  

Because our payment ecosystem spans both native issuing and third-party linked networks, our fraud strategy cannot operate in a silo. We collaborate closely with internal and external partners to maintain full-spectrum risk coverage across every touchpoint from business onboarding to real-time transaction monitoring.  

BMW: Can you explain what Navan Connect is and how it relates to the company’s payment ecosystem?  

XW: Navan Connect provides an innovative capability for businesses to link their current Visa, Mastercard, or American Express corporate card programs directly to the Navan Expense platform without switching card providers. Businesses preserve their banking relationships while gaining access to the Navan expense management tool, including automated transaction tracking, real-time spend visibility and policy enforcement.  

Navan Connect operates seamlessly alongside our native card-issuing program, giving us a dual-rail payment architecture that streamlines corporate spend management from initial booking to automated bookkeeping. 

BMW: How is Navan using agentic AI through Navan Cognition to automate corporate travel payments and expense management, and what role do specialised AI agents play?  

XW: Navan Cognition uses specialised, autonomous AI agents to manage complex, multi-step financial and travel workflows:  

• Ava (Virtual Travel Agent): Reinvents customer support by managing thousands of traveller interactions daily. Ava executes multi-step travel modifications—such as rebooking cancelled flights, modifying hotel reservations, and processing refunds—while continuously monitoring user sentiment to seamlessly hand off edge cases to human agents.  

• Expense Agent: Automatically populates descriptions, categories, and tax details for submitted expenses. Through AI-powered auto-coding, transactions map directly to General Ledger (GL) codes for instant, error-free accounting reconciliation.  

• Audit Agent: Continuously evaluates transactions against 45+ configurable policy and audit rules. It automatically clears over 73% of expenses without human intervention, flagging only true anomalies for managerial review.  

BMW: As Navan operates across multiple markets, how does the complexity of cross-border payments, currencies and regulatory requirements affect fraud risk? And how can data and AI help identify and manage those risks?  

XW: Scaling across international markets requires navigating fragmented regulatory mandates, local clearing rails, multi-currency processing, and compliance frameworks like Strong Customer Authentication (SCA) in Europe. In SCA-mandated markets, traditional issuer fraud liability shifts, requiring a highly optimised approach to authentication. 

To balance security with user experience, we partner closely with our payment partners to design adaptive 3DS step-up strategies. We continuously calibrate our AI models and risk engines using localised datasets, accounting for regional baseline spend behaviours, local Merchant Category Code (MCC) norms, and distinct international fraud vectors. 

BMW: At the MRC San Diego Event, you’ll be discussing how Navan navigates the challenge of preventing sophisticated fraud while ensuring a seamless onboarding experience for businesses. Without revealing too much, could you share a preview of the key themes or insights your session will cover? 

XW: I’ll share best practices for building a fraud AI Agent for onboarding under a 3-pillar company sign-up inspection framework. 

A core takeaway will be how to continuously adapt AI agents as fraud trends evolve. Rather than allowing a Large Language Model (LLM) make direct, opaque binary decisions, we tie the AI agent’s inspection findings to a deterministic scoring layer. This ensures strict mathematical consistency, full auditability, and the ability to tune risk thresholds rapidly without changing core prompts. 

BMW: From a data science perspective, how is Navan using data to distinguish between genuine customer behaviour and potentially fraudulent activity, particularly as fraud patterns continue to evolve? 

XW: We evaluate multi-dimensional contexts rather than relying on static, rigid rule thresholds. Instead of analysing an isolated booking or transaction, our machine learning models score activity against contextual baselines, including individual user travel/spending history, role-based spending expectations, and organisational spend norms. 

We deploy unsupervised clustering models to detect distributed card-testing rings and botnet activity across accounts. Furthermore, downstream chargeback notifications, merchant dispute data, and customer feedback pipe directly back into our risk engine feature store to drive continuous model retraining against emerging attack vectors. 

BMW: Looking ahead, which technologies and changing customer behaviours do you think will have the biggest impact on payments and fraud prevention in the travel industry?  

XW: The rapid rise of Generative AI tools presents the most significant shift. Organised fraud syndicates now use GenAI to bypass standard ID Verification (IDV) and Know Your Business (KYB) controls through synthetic corporate registration filings, deepfake video liveness bypasses, and hyper-realistic, cloned domain networks. 

Combating this requires moving beyond point-in-time checks toward deepfake artifact analysis and multi-layered risk strategies that continually evaluate entity legitimacy long after the initial sign-up. 

Technology and innovation are redefining payments and fraud prevention in corporate travel. Navan’s approach exemplifies the industry’s progress toward smarter, more secure solutions as the landscape continues to evolve. 

***

Author: Bernadette Walsh, Marketing Partnerships Manager, Melbourne, Payments Consulting Network  

Bernadette’s diverse experience has given her the opportunity to observe various approaches to business and problem-solving, helping her develop a unique perspective on building relationships and addressing challenges. Her resourcefulness and willingness to think outside the box have been invaluable in finding creative solutions to complex problems. 

*** 

Payments Consulting Network is a media partner of Merchant Risk Council, in San Diego September 13-16, 2026. 

Other upcoming MRC conferences:

• MRC Dublin 2026, 2-4 November

*** 

If you found this article helpful and would like to read similar articles, please subscribe to our newsletter.