The expense report has survived every wave of enterprise software modernization. ERPs automated procurement. AP platforms automated invoicing. But the expense report, with its chased receipts, manual GL codes and after-the-fact policy reviews, remained stubbornly human.
That is changing. AI agents are now handling the full expense workflow, from receipt capture to GL coding to policy enforcement to reconciliation, without waiting for an employee to open a form.
For example, a Navan article drawing on the Skift and Navan 2026 State of Corporate Travel and Expense report found that 29% of travel and expense managers still process expenses manually, and 71% of travelers spend 30 minutes or more filing a single expense report. The article says AI shifts expense management from reactive reimbursement to proactive control, enforcing policies at the point of transaction before a charge goes through rather than weeks after money is spent.
The article identifies eight functions where AI is already delivering measurable value: receipt processing, GL coding, real-time policy enforcement, fraud detection, reconciliation, compliance capture, VAT reclamation and spend forecasting. A 2025 Forrester TEI study cited by the article found that organizations using Navan saw employees save 24 minutes per expense submission and finance teams spend 40% less time on auditing and reconciliation.
The article also points to the audit problem AI resolves structurally. Traditional manual audits review a statistical sample of transactions. AI audit agents analyze 100% of submissions in real time, scanning for fake receipts, duplicate submissions and policy violations, delivering complete coverage without adding headcount.
AI Catches What Auditors Sample-Check and Miss
A Ramp article argues that the defining difference between AI agents and earlier automation tools is context. Traditional automation follows rigid if-then rules. An agent evaluates patterns, understands nuanced financial scenarios and adapts based on what it learns.
Ramp’s research across 50,000 businesses found that companies using agents for expense management saw out-of-policy spend event rates fall 62% and policy flag rates drop 60% over two years. The Policy Agent screens every transaction with 99% accuracy and declines out-of-policy purchases at the point of sale. Every decision produces an audit trail that holds up to internal review and external compliance checks.
The article also looks ahead to agent-to-agent commerce. When an AI agent books travel or renews software on an employee’s behalf, it needs its own payment credentials. The article points to Visa Intelligent Commerce, which replaces static card numbers with single-use tokens that agents request on demand. Ramp is building on this through its Visa partnership, issuing scoped virtual cards that external AI agents can use within policy guardrails. The finance teams that adopt it early, the article argues, will have a structural advantage when machine-to-machine commerce becomes mainstream.
How Mid-Sized Businesses Are Actually Using AI
An American Express Trendex survey of 513 financial decision makers found that 52% already use AI to manage expenses, most commonly for detecting fraud, categorizing expenses and capturing receipt data. Another 40% say they plan to.
The survey found 66% of businesses struggle to keep up with spending complexity and 88% say they need an expense management tool that can evolve as their policies change. The complexity is structural. Half of businesses surveyed are working with more vendors than two years ago, and 41% have already adjusted expense policies in 2026, citing growing volumes, a need to reduce errors and a need for stronger controls.
The survey’s main point is that AI adoption in expense management is not being driven by IT or transformation initiatives. It is being driven by finance decision makers who are working with more complexity than their current tools were built to handle.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
