A survey conducted on behalf of Basware found that 61% of organisations had deployed AI agents as experiments, and a quarter “did not fully understand” what an AI agent looks like in practice. The implication is that adoption remains uneven and, in many cases, exploratory. Basware’s would like to see its customers move from experimentation to operational use. The survey figures comprised of responses from 200 finance leaders in the US, United Kingdom, France, and Germany.
The question permeating agentic activities in financial platforms is one of governance. Finance functions will delegate tasks to AI systems only human operators retain control over authorisation, are assured of compliance, and have access to an audit trail. Basware’s agents actions pass through what the company describes as a central policy engine. This applies business rules and sets compliance requirements and risk thresholds, referring to such controls as autonomy ‘gates’.
Kurtz described the principle: “Autonomy without trust is just risk. Our platform is uniquely designed to ensure that every AI decision is explainable and governed through the same controls finance teams already rely on.” The company sees its agents integrating with established processes, rather than working in parallel outside governance frameworks.
Basware has several more agentic AIs in development. A Supplier Agent will manage invoice disputes and payment queries, able to contact suppliers and summarise discussions. An AP Pro Agent is intended to assist staff to resolve processing questions via a generative AI interface.
The company cites early user experiences from Billerud, a paper manufacturer. Jesper Persson from the company said there had been benefits. “Since day one, we’ve perceived the desired values from the project. The quality of invoices has improved considerably, and the AI continues to evolve and improve with each passing day. The efficiency gains we achieved translated directly into tangible cost savings.”
The company’s objective is to have finance teams delegate decisions and actions to agents in the future, and it plans to release more AI tools in 2026. The company states that AI is in its platform not an add-on feature.
