Finance teams at more than 80 multinational organisations have been discussing how to control and audit artificial intelligence as it becomes embedded in invoice processing and payment operations. The companies, which have combined annual revenue of $1.4tn, were represented at a Dublin customer forum organised by Basware, the invoice lifecycle management software group.
Basware brought together its VIP and NextGen customer communities to examine the use of automation in critical finance processes. Delegates considered how to maintain traceable approvals when AI is used to classify invoices, route them through financial controls and identify potential payment problems. The $1.4tn figure refers to the revenue of the organisations represented at the gathering, rather than spending managed through Basware or revenue earned by the software provider.
Invoice processing is increasingly automated, but the decisions involved can affect who is paid, whether tax information is correctly recorded and how organisations demonstrate compliance. Incorrect supplier details, duplicate payments and unauthorised changes to bank accounts remain risks even when software reduces the amount of manual data entry. A reliable workflow needs records of both the system’s decisions and any human interventions.
Basware says its customer communities have helped shape more than 150 product and platform innovations. Participants in Dublin discussed future investment in AI, including the way automated agents should operate within defined permissions and business rules. The company has not attributed a separate deployment or efficiency result to each participating organisation; the event provided a forum for product development and customer feedback.
Jason Kurtz, Basware’s chief executive, said customers across different industries were asking how to increase the value of AI while keeping decisions controllable and defensible. The discussion covered the responsibilities of finance departments when automated systems handle work that previously required staff to check codes, suppliers, invoice values and approval routes individually.
The company pointed to its 2026 Finance Performance Report, which draws on transaction information relating to 2.5 billion invoices, 6,500 enterprise customers operating in more than 190 countries and over $10tn of combined spend. Basware reports that among organisations it classifies as best-in-class, 96.5% of AI-assisted coding and approval-routing decisions complete correctly without human intervention. That performance measure applies to the selected group and tasks used in Basware’s analysis; it is not an accuracy guarantee for every customer or payment workflow.
Finance departments introducing greater automation must decide which activities can be completed without review and which should require additional checks. Routine invoice coding may involve different tolerances from a first payment to a new supplier or a request to amend bank account details. Changes in an approver’s authority, data quality or supplier behaviour can also make an apparently ordinary transaction unusual.
Kira Mondrus, Basware’s chief marketing officer, said: “The questions customers raised throughout the event were not about what AI can do. They were about what happens when it gets something wrong.” She said the debate among finance leaders was increasingly focused on evidence and trust in automated decisions rather than the breadth of functions advertised by software suppliers.
The company also discussed its acquisition of Trustpair, the payment fraud prevention business, as part of a broader push to connect invoice handling with controls further along the invoice-to-payment process. The integration of verification and payment safeguards offers a route to reducing gaps between accounts payable systems and the steps used to authorise transfers, although the company has not provided a complete list of new integrated features available to all customers.
Basware plans to present further developments through its Innovation Showcase, including examples of how it combines automated decisions with compliance and financial control requirements. Customers assessing the products will need to establish what can be configured under their existing approval structures, what audit information can be exported and how exceptions are escalated. The Dublin discussions place those operational safeguards alongside productivity gains in the company’s product roadmap.
You must be logged in to post a comment.