Latest Insights

In one meeting, a product owner asks two questions. First: "Which of our customers are about to leave?" A few minutes later: "What kinds of customers do we actually have?" Both sound like "let's use machine learning," but they are different problems, and they call for different approaches.

It's the third working day of the month. Your accounts payable team has a few hundred vendor invoices to get through: PDFs from email, scans from the warehouse, a couple of photos taken on someone's phone. Each one gets opened, read, and typed into the accounting system by hand. Somebody transposes two digits in an invoice number. Somebody else pays the same invoice twice because it arrived by email and by post.