Happy Sunday and welcome to InsideAI! I'm Rob May, CTO at Dianthus. We build custom AI/ML software for e-commerce businesses, then acquire the best businesses and use the software to grow them faster. I'm also a very very active AI angel investor, so if you have something new and interesting that needs an early check, please drop me a note.
One of the areas of building a business that is underexplored, in my opinion, is organizational design. People don't talk about it much but as companies get bigger a poor organizational design can lead to bad tradeoffs that work against your business model. No org structure is perfect. All of them require tradeoffs. The best way to start is think about the communication flows that need to happen regularly and the decisions that need to be made, and structure the org to optimize so that 1) those communications happen more naturally (as opposed to constant cross-team status meetings) and 2) decisions are made at the right point in the org.
As an example, in software companies, Product and Engineering teams sometimes come into conflict. Product has a feature they want built a certain way, and Engineering determines that way doesn't work well given how the product is architected. If your org structure is such that the Engineering group rolls up under someone who is a product person and also runs the product group, then your outcome will usually be a Product-centric outcome. If the Product team rolls up to the head of Engineering, you will get an engineering-centric outcome most of the time. If you want those conflicts escalated to you as CEO, then the head of Product and the head of Engineering need to be peers who both report to you.
The question today is - what does this mean for AI? I think most organizations use AI this way - they have a Data Science team that is like a shared services group. People in other departments go to the Data Science team with problems and the team finds solutions. The second most popular org structure is that Data Science teams are embedded in other departments or business units. In this way they more closely understand the problems they are working on, but it often leads to fragmented data sets and tools.
A good AI-centric org structure should do two things. First, it should unify data-centric workflows to avoid "data politics." This way you don't get different parts of the org doing different things with data, not sharing it, not insuring it's clean and complete, etc. The second thing it should do is lead with AI. Instead of Marketing coming to the AI group saying "could you build us a better model to recommend new products", the AI team should be generating ideas for the Marketing team to consider. While the AI team will inevitably generate some ideas that don't matter to the Marketing team, I believe overall you get a better outcome if AI leads with ideas and brainstorming.
The most recent trend has been that people are your core competency. Success is all about human capital and knowledge management and motivation and teamwork. That trend is peaking and the new core competency is going to be data. How you get it, clean it, manage it, and use it is going to matter more than ever before. You will need AI-first org structures to minimize data politics and maximize data driven innovation. This will ripple through all types of business frameworks in the coming decade. Best to adapt early and be a leader.
If you are working through problems like these, I'd love to hear your story. Thanks for reading.
@robmay
Comments
Post a Comment