Corporate Expense Management in 2026: Cards, Controls and Costs

The finance head of a knitwear exporter in Tiruppur has a ritual every time the merchandising team returns from a European trade fair. Six people come back with a month's worth of receipts in three currencies, a few lost hotel folios, and taxi bills written in German. Reconciling it all against the company's bank statements takes her team most of a week. The money spent is not large. The cost of finding out what it was spent on is.

That gap, between the cost of spending and the cost of accounting for spending, is what the corporate expense industry sells against. In 2026 it has become one of the most heavily funded corners of fintech, with one start-up valued at $60 billion and another bought by a top US card issuer. This piece looks at what the market actually is, what the numbers say about cost and fraud, and which controls work.

A market built on business travel and card rails

Start with the spend itself. Global business travel spending reached $1.59 trillion in 2025, up 8.4%, and the Global Business Travel Association forecasts $1.71 trillion in 2026, with the total passing $2 trillion by 2030, according to Hotel Management's report on the GBTA index. The US ($423 billion) and China ($403.7 billion) together account for nearly half.

Much of that flows over card rails. Visa reported $1.8 trillion in commercial payments volume for fiscal 2025, up 7% in constant dollars, while revenue from its commercial and money movement business grew 14%, according to a MediaNama summary of the company's results. Cross-border volume, which covers much of business travel, grew 13%. Faster revenue growth than volume growth tells you where the margin is: in the services wrapped around the card, not the swipe itself.

The spend-management boom, in two deals

Two transactions this year frame the market. In January, Capital One agreed to buy Brex, which combines corporate cards, spend software and business banking, for $5.15 billion in a cash and stock deal, according to Coverager. Brex had been valued at $12.3 billion in a private round about four years earlier. The deal was expected to close in mid-2026, subject to approvals.

In October, Bloomberg reported that Ramp had raised about $1.85 billion at a $60 billion valuation, up from $44 billion in June, as covered by PYMNTS. Ramp's annualised revenue topped $1.5 billion in June, and its customer base reached 70,000 businesses, up from 50,000 at the start of the year.

Our read: the two deals point in the same direction from opposite ends. A bank paid a large discount to a peak private valuation to buy software and customers it could not build quickly. A software-first company won a premium valuation by turning card interchange into a platform for bill payments, procurement and, increasingly, automated approvals. The card is becoming the cheap part. The software that decides whether a card should work for a given purchase is the expensive part.

The next step is already visible. Spend platforms are adding software agents that read a purchase request, check it against policy and budget, and issue a card or approve a bill without a human in the loop for routine cases. That moves the finance team from approving individual transactions to writing the rules the agents follow, and auditing the exceptions. It also raises a new governance question: who is accountable when an automated approval goes wrong? Our piece on agentic payments and AI agents looks at how card networks are building rails for this. For a mid-sized company, the practical test is simple. If a platform cannot show exactly which rule approved a given purchase, and who set that rule, it is not ready to be trusted with the budget.

The cost nobody books: processing the claim

The most cited benchmark on the cost of expense reports is old, but nothing better has replaced it. A GBTA Foundation study from 2015 found that a single expense report cost $58 to process and took 20 minutes. Nineteen percent contained errors or missing information, and fixing each one cost another $52 and 18 minutes. Wage inflation since then means the real figure today is almost certainly higher.

Worked example: 400 travelling staff

Take a mid-sized company with 400 employees who each file two expense reports a month. That is 9,600 reports a year.

• Base processing at $58 each: $556,800.
• Error correction on 19% of reports, or 1,824 reports at $52: $94,848.
• Total: roughly $651,600 a year, or about $68 per report, before counting any fraud.

Now suppose a card-led system captures receipts at the point of purchase and codes most transactions automatically, and that this halves both the processing cost and the error rate. Those are assumptions, not measured results, but they are the kind vendors claim. Annual cost would fall to about $278,400 for processing and $47,400 for corrections on 912 flawed reports, or roughly $325,800. The saving of around $325,000 is the real business case for spend software, and it is larger than the interchange rebates most mid-sized firms earn on their cards. Companies should test any vendor claim against their own report volumes and error data.

The card is becoming the cheap part. The software that decides whether a card should work is the expensive part.

Fraud: small amounts, long lives

Expense fraud rarely makes headlines because individual losses are modest. It is common, though. In the Association of Certified Fraud Examiners' 2024 study of nearly 2,000 cases, expense reimbursement schemes made up 13% of all occupational fraud cases, the third most common type after non-cash theft and billing schemes, according to a CaseIQ summary of the ACFE report. These schemes lasted around 18 months on average before detection, among the longest of any category.

The four classic patterns are mischaracterised expenses (a family dinner claimed as a client meal), overstated amounts (an altered receipt), fictitious expenses (a receipt that never existed) and multiple claims for the same cost. Corporate cards close off some of these, because the transaction data comes from the network rather than the employee. They do nothing about mischaracterisation, which is a judgment about purpose. And generative AI now produces convincing fake receipts in seconds, which shifts the defence from inspecting documents to matching them against card and merchant data. Our coverage of AI fraud in payments looks at that arms race in more detail.

Controls that actually work

The shift from reimbursement to cards changes when controls are applied. In a reimbursement world, policy is enforced after the money is spent, by a reviewer reading a claim. With programmable cards, policy can be enforced before the purchase is approved. In practice, the controls that do the most work are:

1. Merchant category limits. A travel card that cannot be used at jewellers or electronics stores removes a whole class of misuse.
2. Per-trip or per-project budgets. A card that expires when the trip ends, or carries a limit equal to the approved budget, limits exposure.
3. Receipt capture at the point of sale. Asking for the receipt within hours, not weeks, cuts both errors and the scope for fabrication.
4. Automated matching. Comparing card data, receipt data and travel bookings flags duplicates and mismatched dates without a human reading every line.
5. Random audit of approved claims. The 18-month detection window suggests most firms review too little after approval.

The multi-currency and India angle

For companies whose staff travel abroad, currency conversion is often the largest hidden cost on the card bill. A single-currency card converts every foreign transaction, usually with a markup over the wholesale rate. As an illustration, if a card adds a 2% markup, the Tiruppur team's EUR 3,000 a month in trade-fair spending costs an extra EUR 60 a month, or EUR 720 a year, before any bank charges. A card that holds euro balances, funded at a better rate in advance, can reduce that. Our comparison of multi-currency travel cards and crypto cards covers the consumer side of the same trade-off.

India's card market is large and still growing: credit card spending reached a record Rs 21.16 lakh crore in 2024-25, up about 15%, with 109.8 million cards in use by March 2025, according to RBI data reported by Outlook Money. The RBI does not publish a separate figure for corporate cards, which is itself telling. Much of Indian business spending still runs through owners' personal cards, cash advances and reimbursement, which is exactly the setting where expense software has the most to fix.

The next frontier is cards linked to stablecoin balances. Visa says it has 130 stablecoin-linked card programmes in 40 countries, with spending quadrupling year on year. For businesses holding stablecoins for payments, a card that spends those balances directly could remove a conversion step. See our explainer on how crypto cards work for the mechanics.

What to watch next

• Late October 2026: Visa's fiscal 2026 results, the clearest read on whether commercial volume growth accelerates from 7%.
• Through late 2026: integration of Brex into Capital One after the deal expected to close in mid-2026, and whether the combined company competes on price with software-first rivals.
• Following the October 2026 round: how Ramp deploys $1.85 billion in new capital, particularly in AI-driven approvals and international expansion beyond the US.
• 18 January 2027: the latest date for the US GENIUS Act stablecoin framework to take effect, which will shape how quickly stablecoin-linked corporate cards spread.

About this researchMuffinPay InsightsIndependent analysis of stablecoins, CBDCs, cross-border payments and digital public infrastructure. Every figure is linked to its source. Read our editorial standards →