AI for Retail

CHAPTER 25Buying vs. Building AI

“Why can't I just build this myself?”

Every Retailer

AI is hard. Let me emphasize that. AI is really, really hard. Building an AI solution is not like building a web app or some simple software product. Building such a simple web app is akin to building a house. You need to get the specs right (user stories / mock‐ups / architecture diagrams), you need to get a contractor (product manager), a plumber (database architect), an electrician (infrastructure engineer), a few construction workers (software engineers), and perhaps an interior designer (UI designer). You build the house, the inspector comes by to make sure it is to code (QA engineer), and you are done. People can start living in it. And maybe you need a single low‐salary person to come by every once in a while, when a pipe bursts or to change a light bulb, but that's about it.

If you want to build a simple salesforce CRM clone for one‐tenth the cost, go right ahead. I know I would if I were a CTO of a major retailer.

However, building an AI system is a bit more like building a nuclear power plant. You need everything else we just discussed, but on top of that you need to go hire a number of nuclear engineers (AI engineers) who are quite expensive ($400k–600k/year), there are only 10k in the world that are actually qualified and are heavily sought after, so much so, not even the biggest companies in the world can hold on to them. Then once you build the power plant, you need teams of ...