TL;DR
Get home office essentials delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Bottomline risk and fraud officer Katie Elliott told PYMNTS that AI is helping criminals make B2B fraud attempts broader and faster, while finance teams cannot scale manual checks at the same rate. She argues that payment controls should combine multiple signals, automate routine transactions and send unusual activity to human review; the cost and effectiveness of shared verification tools remain open questions.
AI is making B2B fraud attempts cheaper to scale, while finance teams still face the cost and workload of checking payment requests, Bottomline senior risk and fraud officer Katie Elliott told PYMNTS. Her account points to a shift in fraud prevention: businesses may need to verify suppliers and payment instructions using multiple signals before sending funds, rather than relying mainly on manual review or trying to recover money afterward.
Elliott said attackers now use AI and other available tools to make attempts “bigger, broader, faster.” She described a move from more targeted attacks toward mass phishing and spam. The report frames the imbalance as an economic one: criminals can make many attempts while needing only a small number to succeed, but companies cannot add manual checks to every transaction without slowing legitimate payments or expanding finance teams.
The risk is sharper when payments move quickly. “Once you click that send button,” Elliott said, an authorized payment can move in an instant, and a recipient may take the funds quickly. That makes pre-payment checks more valuable than relying on recovery after a mistaken or fraudulent transfer, though the report does not quantify recovery rates or the scale of losses.
Elliott recommended against making payment decisions from a single data point. She cited digital identity, phone details, email-domain history and other signals as information that may help establish whether a payment request fits the supplier relationship. She also said third-party data sources, APIs, verification tools and specialist expertise can be expensive, particularly for smaller companies. Payment networks, she suggested, could spread those costs across a larger transaction base.
Why Payment Checks May Move Upstream
The immediate operational issue is how to screen payment instructions without creating a manual bottleneck. If transfers can be difficult to stop once authorized, confidence before payment carries greater weight. That matters to CFOs and accounts-payable teams responsible for both preventing fraud and keeping legitimate suppliers paid on time.
The interview also points to a possible change in the role of payment networks. Their value may extend beyond moving funds if they can provide verification capabilities that individual businesses find costly to build. That is a proposition, not a demonstrated outcome in the source: it provides no cost comparison, performance data or evidence that shared tools prevent a stated share of fraud.
For companies, the balance is between stronger checks and operational friction. A process that flags too many ordinary payments for people to review could delay suppliers; one that misses unusual instructions could expose the business to loss. Risk controls must scale economically, but the report does not establish one model as suitable for every company.
As an affiliate, we earn on qualifying purchases.
From Supplier Onboarding to Ongoing Checks
Traditional accounts-payable controls often rely on establishing supplier details during onboarding and reviewing payment instructions when changes occur. Elliott’s warning is that this approach may be less dependable if convincing identities and impersonation attempts become easier to produce with AI. She described the ability to generate a whole identity as her greatest concern, but did not provide examples or data on how often this has happened.
Her proposed approach assigns different jobs to automation and people. AI could process ordinary activity, while human review is reserved for anomalies such as changed bank details, unusually rapid transaction activity or a supplier seeking an abnormal amount. Elliott gave the example of a supplier that normally receives a regular payment suddenly requesting three times as much; she said the system should send that case for approval.
This approach treats trust as something that may need to be checked repeatedly, rather than settled once during onboarding. It also depends on reliable information about counterparties and payment behavior. The interview does not detail how firms should gather or govern those data, or how to handle mismatches between signals.
“They are using the technology that’s out there, the AI, every tool available to them in order to make their attempts bigger, broader, faster.”
— Katie Elliott, senior risk and fraud officer at Bottomline, speaking to PYMNTS
As an affiliate, we earn on qualifying purchases.
Evidence on Costs and Results
The source is an interview and does not give fraud-loss totals, incident rates or measured AI-driven increases. It does not establish how often attackers use AI in B2B schemes, which sectors are most affected, or whether the described tactics have produced more successful payments fraud.
There are also no figures comparing in-house verification with network-provided services, and no test results showing how accurately automated systems distinguish fraud from legitimate exceptions. The practical costs, data access requirements, privacy implications and effects on payment delays are not specified. Elliott’s remarks identify concerns and possible controls; they do not prove that a particular verification model will work for every business.
As an affiliate, we earn on qualifying purchases.
What Finance Teams May Test
The next step for finance teams is to determine which payment changes and behaviors warrant extra review, and whether they have enough reliable data to assess them. Based on Elliott’s examples, firms may focus on changed account instructions, unusual payment velocity and amounts that differ sharply from a supplier’s established pattern. The interview does not announce a product launch, policy change or implementation deadline.
Businesses evaluating network or third-party verification will need to compare the cost of shared services with their own controls and assess how exceptions are routed to staff. More detail on effectiveness, data safeguards and impacts on legitimate payment processing would be needed to judge whether these tools can meet the challenge at scale. The source points readers to the full PYMNTS TV interview with Elliott for further discussion; it reports no additional confirmed milestone.
business identity verification services
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the development described in the report?
A PYMNTS interview with Bottomline’s Katie Elliott describes how AI can help criminals scale B2B fraud attempts and argues that companies need payment verification capable of handling more activity without relying on manual checks for every transaction.
What kinds of signals does Elliott say businesses should check?
She cited digital identity, phone information, email-domain history and other signals, and cautioned against basing a payment decision on only one piece of data. The interview does not prescribe a specific tool or checklist.
Does the report show that AI has increased B2B fraud losses?
No. It reports Elliott’s assessment that AI is making attempts broader and faster, but provides no loss figures or measured change in fraud rates.
What role would people have in Elliott’s proposed approach?
Automation could handle routine payment activity, while human review would focus on exceptions such as a changed payment instruction, a sudden rise in transaction activity or an unusually large supplier request.
Is a shared payment-network verification service confirmed?
No service or rollout is announced in the source. Elliott said networks could potentially spread the cost of verification data and expertise, but the report gives no pricing or performance evidence.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
