This article is part of a series. In Part 1, I outlined the premise: ML/AI shops can borrow tips and best practices from how algorithmic trading ("algo trading") shops operate. The rest of the articles explore those ideas in more detail.

Simply building an ML/AI department in your company is not enough. Any such work should have a firm business purpose and a means to evaluate how well it performs.

This is straightforward in the trading world. Every algo trading shop can summarize their purpose as follows:

"Our company's goal is to make money through buying and selling market instruments, such as shares of stocks. We analyze various data sources in an attempt to develop investment algorithms (to determine what kinds of trades we'd want to place) and execution algorithms (to place those trades in response to market activity). We evaluate the quality of our work, and our benefit to the company, based on how much money those trades generate for us."

(For the traders in the audience: I've oversimplified for brevity.)

Investment banks therefore have a clear reason as to why they have algo traders in the first place, and this set of algo traders in particular. They can see how investing in those traders' training, headcount, and tools reflects on that team's outcomes.

What statement can you make about your company's ML/AI efforts? How do you connect your ML/AI work to the company mission, and improving the bottom line? And how do you know when the team is successful?

Granted, it's much easier to do this in the trading world -- it's very clear to connect a decision to an outcome, since traders are dealing with prices and transactions -- but you should still try to document why your company uses ML/AI. If you don't have answers to those questions, consult (or define) your data strategy to guide you.

The data strategy will guide you on what skills you'll need and what data projects to tackle. It will also shed light your technical decisions. Should you start using neural networks? Is it time to develop a new cloud-based data warehouse? How much effort should you put into testing whatever is the latest shiny new data tool or technique? If you've defined why you're using ML/AI, then you can make those decisions based on how they will help your company, and not just because they are popular.

Remember: tools don't matter; results do. So make those ML/AI decisions accordingly.