Expense and travel reimbursement looks like a cheap process until you separate the reports that go through cleanly from the ones that come back wrong. The GBTA Foundation, in work with HRS, put the cost of processing a single expense report at about $58 and the time at about twenty minutes. That figure is the one most operators quote, and it understates the problem, because it averages the clean report with the broken one. Roughly nineteen percent of reports contain an error or missing information, and each of those costs an additional $52 and eighteen minutes to correct. The cost of the function concentrates in that fraction, not in the routine submission, and a redesign that does not target the fraction recovers very little.
Where the money actually goes
The arithmetic the GBTA Foundation ran is worth following because it shows how a small error rate compounds into a real number. A company that processes an average of about 51,000 reports a year, with roughly nineteen percent of them needing correction, is correcting on the order of ten thousand reports. At $52 and eighteen minutes each, the GBTA Foundation estimated average annual spending of roughly half a million dollars and nearly three thousand hours on error correction alone. None of those three thousand hours buy any business outcome. They go to reading a report, finding that it does not tie to policy or lacks a receipt, sending it back, and reprocessing it once the employee responds. The figures come from an older study that has been cited as durable for years, so treat the exact dollar amounts as indicative rather than current, but the underlying pattern is stable: a minority of reports drives a majority of the labor.
The category is large enough that the labor matters in absolute terms. GBTA's Business Travel Index 2025 projected global business-travel spending at about $1.57 trillion for the year, with the United States at $395.4 billion. Every dollar of that flows through a reimbursement or card-reconciliation process somewhere, and at scale the nineteen percent error tail is a standing cost on the finance team.
The compliance constraint sets the shape of the work
Expense processing is not a pure cost-minimization problem, because the IRS attaches conditions to keeping reimbursements tax-free. Under the IRS accountable-plan rules, a reimbursement is excluded from the employee's taxable wages only if the expense has a business connection, is substantiated, and any excess advance is returned, with substantiation generally required within sixty days. Substantiation means the amount, date, place, and business purpose, supported by a receipt where one is required. If a report is not substantiated inside that window, the reimbursement can become taxable wages, which pulls in payroll tax and withholding. The IRS also publishes per diem rates, restated each year in a notice such as Notice 2025-54, that let a company substantiate certain travel costs at a fixed daily amount instead of receipt by receipt.
That constraint is why the error tail is expensive rather than merely annoying. A missing receipt or an unexplained charge is a substantiation gap that, left open past sixty days, changes the tax treatment of the payment. The correction work exists because the document has to be made compliant, not just approved.
Walking the redesigned workflow
A redesign starts at capture. An agent reads the receipt or card-feed line, extracts the amount, date, merchant, and category, and attaches the image to the report, so the substantiation record is assembled at the moment of submission rather than chased afterward. This is the same document-variability problem that shows up across finance operations: receipts arrive as photos, PDFs, folded paper, and foreign-currency card charges, and the variation lives in a long tail of formats that no single template covers.
Next is the policy check. The agent compares each line against the company's rules and the relevant IRS test: is there a receipt where one is required, is the amount inside the per diem or policy cap, is a business purpose recorded, is the submission inside the sixty-day window. A line that passes every test needs no human attention. A line that fails one carries a specific reason, which is the information the correction step has always depended on.
The agent then maps each compliant line to the general ledger, choosing the account from the category and the historical coding of similar expenses, the same mechanical matching that drives the month-end close and the three-way match in accounts payable. The clean majority of lines flow through coded and substantiated, with no person in the path.
Only genuine exceptions route to a human: the missing receipt, the over-cap charge, the expense that does not fit any category. Finance keeps approval authority throughout. The agent does the substantiation and coding; the controller decides on the cases that need judgment and on every payment. The redesign does not remove the approver. It stops the approver from reading the eighty-one percent of reports that were always going to be fine.
Why the tail is where the time is
The clean report is already close to free in labor terms, near the twenty-minute, $58 baseline that mostly reflects employee time rather than finance time. There is little to recover there. The correction, at an extra $52 and eighteen minutes, is where the finance team's hours actually go, and it is the part driven by substantiation gaps rather than by accounting judgment. A redesign that catches the gap at capture, before the report is submitted, prevents the correction cycle instead of speeding it up. The three thousand hours the GBTA Foundation counted are reclaimed by not generating the rework, which is a different result from making each correction faster.
A reasonable counter, answered
A reasonable counter is that the error rate is mostly a behavior problem, not a process problem, and that the fix is to train employees to submit clean reports rather than to automate the checking. There is something to this. Better policy and clearer rules do reduce errors at the source. But the GBTA Foundation's nineteen percent has persisted across organizations that all run training and all publish policies, which suggests the floor is structural: people travel, receipts get lost, foreign charges post in odd amounts, and the sixty-day clock runs whether or not anyone remembers it. The redesign does not assume disciplined employees. It assumes the tail exists and moves the check to the moment of capture, where the gap is cheapest to close and where finance keeps the authority that makes the engagement an aligned one rather than a handoff.