Basware report finds AI & automation gap in finance

Basware has published its inaugural Finance Performance Report, based on invoice and transaction data worth more than USD $10 trillion.

The study draws on production data from 2.5 billion invoices processed for 6,500 enterprise customers in more than 190 countries. The annual invoice value represented in the report exceeds USD $1 trillion, or roughly 1% of global GDP.

Rather than relying on surveys, the report uses data from transactions moving through Basware's network to assess finance team performance in areas including automation, AI-driven decision-making, working capital, controls and compliance.

Jason Kurtz, Chief Executive Officer of Basware, said the invoice remains a rich source of operational and financial data for large organisations.

"Basware processes more invoice data than any other platform. The invoice is one of the richest sources of intelligence in finance. It tells you what is happening across your processes, where AI is delivering real value, and where performance still falls short. This report turns that intelligence into something every finance leader can use," Kurtz said.

Automation gap

One of the central findings is a gap between companies that have automated parts of accounts payable and those that can run most of the invoice lifecycle with limited manual intervention. Basware measures this through what it calls a Lifecycle Autonomy Rate, which covers stages from invoice receipt and matching to transfer.

Best-in-class organisations recorded a 93% Lifecycle Autonomy Rate, compared with 81% across the wider network, a 12 percentage point gap.

The report also sets out an AI Decision Rate, which measures whether AI can make coding and approval-routing decisions correctly without human intervention. Among the top group, 96.5% of those AI-driven decisions were completed correctly, compared with 89.4% across the network.

Basware said the figures point to a distinction between task-level automation and broader process orchestration, with human oversight reserved for exceptions and accountability. The data suggests finance teams are increasingly testing how far AI can be used in routine operations while preserving auditability.

Control findings

The report also highlights differences in control performance. Across Basware's network, 1.4% of invoices were flagged as potential fraud or duplicates and diverted for manual review before posting.

Among best-in-class organisations, the comparable figure was 0.05%, indicating stronger controls earlier in the process and reducing the number of invoices that need a second review before moving ahead.

Donna Wilczek, Chief Product and Technology Officer of Basware, said the findings show that automation and control are being developed together in the strongest finance operations.

"This is what Governed Autonomy looks like in practice. Best-in-class organizations aren't choosing between automation and control. As AI transforms finance, that kind of visibility is critical to knowing where to automate, where to increase trust around AI, and where to keep people in control. That's the model every finance team should be building toward," Wilczek said.

Payment and compliance

The data also links invoice handling to broader finance priorities such as supplier payments and compliance. Best-in-class organisations paid 92% of invoices on or before the due date, compared with an average of 81.8% across the network.

That 10.2 percentage point gap suggests invoice process design can affect supplier relationships and working capital management, both of which have become more prominent concerns for finance leaders managing tighter cash positions and more complex supply chains.

On compliance, best-in-class organisations receive 99.7% of invoice volume through a single centralised compliance platform. Basware measures this through a Global Compliance Coverage metric that focuses on how consolidated a company's invoicing compliance approach is, rather than simply the number of markets it operates in.

A more centralised compliance model can reduce the burden of dealing with multiple local systems, outside vendors and direct government connections as e-invoicing mandates expand across jurisdictions.

How measured

The analysis covers 10 measures across invoice automation, AI decision-making, payment performance, matching, procurement compliance, fraud and duplicate exposure, e-invoicing and operational complexity. Unless otherwise stated, the figures are calculated using a volume-weighted cohort that meets invoice-volume, activity and data-completeness thresholds and represents about 92% of total network invoice volume.

Best-in-class results reflect the average performance of the top 5% of qualifying customers for each metric, meaning the group can differ from one measure to another. This creates a benchmarking structure that compares network-wide averages with the strongest performers on individual indicators.

The publication is set up as an ongoing reporting programme using data from the Basware network, with regular updates planned around specific themes, industries and metrics. Its first edition offers a baseline drawn from live invoice flows rather than survey responses, and points to persistent gaps in autonomy, control, payment performance and compliance across large finance operations.