
What I like best about Diffco is their real AI depth and honesty, even when it could mean less work for them. Our product reads receipts and extracts line items – the AI is the product, not just a feature. Diffco spent four months working on our models and the strategy around them, and the most useful thing they did was establish a way to measure accuracy. We now evaluate each change against that measurement before it ships, rather than assessing it by trying a handful of receipts. They were also willing to tell us when a problem was in our data rather than our model, which showed us they were focused on solving the actual problem, even if it meant less model work for them. Review collected by and hosted on G2.com.
Nothing that affected the work. Some of the early sessions were more strategic than I expected when I thought we were starting on implementation. In hindsight, that order made sense, but it took me a little while to see why. Review collected by and hosted on G2.com.