Artificial intelligence is expected to change everything. Customer insight, compliance and business value still matter.
I used to believe, like many people, that AI would change everything about how work gets done. In accounts payable and payments, I expected fewer manual steps and more complex processes handled automatically. It seemed reasonable to think that as the technology became more capable, products would become more valuable. I still believe AI will reshape how work gets done. What I’ve reconsidered is the assumption that greater capability, on its own, produces a better product.
The Numbers That Sharpened the Question
One finding from Tungsten Automation’s eInvoicing and Payments Report 2026 brought that realization into focus. The share of business leaders who have modernized invoice processing is 77%, but only 30% fully automate payment reconciliation with invoice data. Meanwhile, 73% of respondents plan to automate reconciliation over the next 12 months.
For me, the figures show that automating one part of finance doesn’t ensure information and decisions carry through to the next. In payments, timing affects cash, controls must meet compliance requirements and payment decisions influence supplier relationships. The right payment decisions also depend on the circumstances. A team may need to weigh payment terms against cash priorities or consider whether a different method makes sense for a supplier. Product design has to reflect those trade-offs. A more capable system cannot decide what matters most to a particular business without understanding how that business operates.
That made me think past what AI can do. The important question is what a finance team needs the product to accomplish, and what evidence would show that it has done so. Automation has to fit the customers’ workflow and produce an outcome the business values.
What I’m Carrying Into Q4
Going forward, I’m asking product teams to make those requirements explicit before settling on a solution. Which part of the process needs to improve? What compliance obligations shape it? What business result should the customer be able to measure?
I want those answers to shape what we build as well as how we assess it. A product hypothesis should connect a real customer need to a practical outcome, such as reducing manual reconciliation work, improving payment traceability, or helping a team act on cash priorities. That gives us a clearer basis for deciding whether artificial intelligence is the right way to solve the problem. So, while AI will continue to expand what products can do, building products that finance teams trust will still require customer insight, compliance and a clear business case.
