If AI Isn’t Your Competitive Advantage, Stop Building It

Jason Kurtz, CEO of Basware, warns that companies should stop building custom AI systems for standard business processes. He cites an example of an automotive manufacturer that wasted 18 months developing a bespoke AI solution for expense reporting. Kurtz argues that such specialized products could have been created in about 12 weeks at a tenth of the cost using existing specialist software.

While there are valid reasons to build in-house AI, such as when a company possesses unique proprietary data or when the use case is closely tied to its product, the math changes when the work is something that thousands of other businesses also need to do. Standard processes like accounts payable processing invoices can seem like they require deep specialist technology, but engineering challenges increase when integrating with ERP systems, supplier records, tax requirements, and handling exceptions.

Developing AI internally involves creating the surrounding workflow and maintaining it as regulations change and production data behaves differently from the training dataset. The quickest path to value is using AI embedded into processes and systems already in use by teams, leveraging large amounts of data to educate the AI. This tempered enthusiasm for custom AI suggests that the quickest path to value is leveraging existing solutions for common business functions.