AI makes it cheaper to build more software. But a better product isn’t just more software.
That’s what I liked about this post from Thariq. It also got me thinking about what else a company could afford to do with those savings.
the right way to use model capabilities is not to ship 10x more features to prod
it's to spend more time understanding your users, trying experiments, building prototypes, learning about things you don't understand so that you can ship things that actually work
Suppose AI handles most routine support questions well, at a fraction of the cost. We could spend some of the savings visiting customers in person. Sit with a team through their first day using the software. Help them get comfortable with it. Give them time to ask all their questions.
Those visits still take time. They still don’t scale particularly well. But customers get something valuable, and we have more room to offer it.
Working in healthcare, I’m familiar with Baumol’s cost disease: as wages rise across the economy, work with fewer productivity gains becomes relatively more expensive. An afternoon spent training a clinic’s team can still take an afternoon, however quickly we shipped the software.
That afternoon hasn’t become cheaper. The opportunity is to pay for more of it with savings elsewhere. AI can help with training and implementation too; the gains just don’t have to be equal across everything a company does.
We don’t have to make every part of the company equally efficient. Some of the things customers appreciate most involve giving them more of our time.
Getting much more productive in one part of the business could let us be more generous in another.