Expense Fraud in the Age of Generative AI: What the Data Shows

Expense reports look like a low-stakes administrative process. The data says otherwise. The ACFE's Report to the Nations 2024 — 1,921 fraud cases across 138 countries — found expense reimbursement schemes in 20% of cases at companies under 100 employees, and 12% at larger ones. The median loss per case: $145,000.

Why does detection speed matter so much?

Because it's the biggest lever on cost. The ACFE found a median loss of $30,000 when fraud is caught within six months — rising to $250,000 when it runs undetected for two to three years. Faster detection doesn't just catch individual cases; it caps how much damage any single scheme can do.

How has generative AI changed expense fraud?

It hasn't created the problem — it's removed the skill barrier. Altering a receipt used to require real photo-editing ability. Today, a generic AI tool can produce a convincing fake in seconds, often indistinguishable from a genuine document even to an experienced reviewer.

Deloitte's Center for Financial Services projects GenAI-enabled fraud losses in the US will grow from $12.3 billion in 2023 to $40 billion by 2027 — a 32% compound annual growth rate. In Europe, Sumsub recorded a 780% year-on-year jump in deepfake incidents in 2023 (300% in the UK alone).

identifAI's own Deepfake Intelligence Report, based on nearly 10,000 observed synthetic media incidents worldwide (2020–2026), shows the same shift from a different angle: fraud was the second most common motivation behind deepfake use, at 20.1% of classified incidents. Looking at how fakes are made rather than why, synthetic still images — the format most used to fabricate documents — accounted for 17.4% of all technical carriers tracked.

Three mechanisms now show up repeatedly: template-generated receipts requiring no skill at all, AI-generated receipts realistic enough to include convincing wear and lighting detail, and AI-edited alterations to genuine documents — changed amounts, dates, VAT rates, or a swapped supplier IBAN to redirect a payment.

Why can't manual review keep up?

The ACFE estimates fraud runs about 12 months on average before discovery — a figure that reflects the limits of manual spot-checking as much as it does the fraud itself. Faced with the expense volume a mid-sized company generates yearly, finance teams get pushed toward sampling instead of systematic review, trading control quality for throughput. The cost isn't only financial: slower reviews mean slower reimbursements and less confidence in the underlying data.

What does AI-powered fraud detection actually catch?

The same generative AI capabilities that make fraud easier to produce are now used to catch it. Rather than automating everything, the most effective model is human-in-the-loop: AI systematically screens every submission for signals a manual reviewer would likely miss — layout inconsistencies, mismatched fonts or logos, generative-image artefacts, amounts inconsistent with the claimed category, receipts from a country that doesn't match travel data, duplicate submissions, or an invoice IBAN that doesn't match payment history — and routes only genuinely ambiguous or high-risk cases to a human for a final call.

That's the model behind SmartEX, Havant's expense and travel management platform, now integrated with identifAI to detect AI-generated, manipulated, or altered content directly inside the approval workflow — flagging anomalies without asking finance teams to give up judgment on complex cases.

Expense document fraud isn't a future risk. It's already measurable, already growing, and already accelerated by generative AI — and the tools enabling it are, for now, also the best tools available to counter it.

FAQ

What percentage of fraud cases involve expense reimbursement schemes?

‍
According to the ACFE's Report to the Nations 2024, expense reimbursement schemes appear in 20% of fraud cases at organisations with fewer than 100 employees, and 12% at larger organisations.

How much does expense fraud typically cost a company?

‍
The ACFE found a median loss of $145,000 per fraud case overall. Detection time changes this significantly: $30,000 median loss if caught within six months, rising to $250,000 if the fraud runs for two to three years.

Can generative AI create fake receipts?

‍
Yes. Generative AI tools can produce visually realistic fake receipts — including convincing fonts, logos, and wear details — often indistinguishable from genuine documents in manual review.

How common is fraud as a motivation for deepfakes?

‍
In identifAI's Deepfake Intelligence Report, fraud accounted for 20.1% of nearly 10,000 classified synthetic media incidents tracked globally between 2020 and 2026, making it the second most common motivation after political manipulation.

How does AI-powered expense fraud detection work?

‍
AI systems screen every expense submission for signals like layout inconsistencies, mismatched fonts, duplicate claims, or IBAN mismatches, then route only high-risk or ambiguous cases to a human reviewer — a human-in-the-loop model rather than full automation.

What is SmartEX?

‍
SmartEX is Havant's expense and travel management platform. It's integrated with identifAI to detect AI-generated or manipulated documents directly within the expense approval workflow.

‍

‍

‍