Six lessons on Decision Ops from our conversation with Carolyn Mooney
Vijay and I had a great conversation with the CEO of Nextmv, Carolyn Mooney, about Decision Ops. (Here is a link to the podcast.)
She has great insight into automating and optimizing decisions within an organization.
I learned a lot by re-listening to her episode. Here are a few highlights for me:
One: Decision Ops should be its own process. She defines and makes a case for the importance of decision ops.
Two: Testing is important. I like this theme throughout our podcast episodes. She stressed the importance of being able to replicate the decisions suggested (keep the inputs and outputs).
Three: Put engineers and end users together. To build effective decision intelligence tools, stay close to the end users. The end users have a lot of knowledge about the decisions being made.
Four: Carefully define what decisions are being made. I enjoyed her points about the importance of carefully mapping out what decisions are currently being made. It’s hard to improve a business if you’re unclear about what decisions are being made.
Five: Baselining is important. This relates to defining the decisions. Once you find these, you want to understand how they are being made today. This is before you attempt to make better decisions. She had some good advice about going slow— you can start to automate with very simple techniques before using more complicated optimization models.
Six: To make a decision, you may need to string together several algorithms. There is a temptation to solve a problem in one go. However, she emphasized the importance of being able to string together different algorithms to make informed decisions. I like to think of the space of Practical AI as additive— as new techniques are developed (like LLMs), they don’t replace old ones, your choices as an engineer grow.
Be sure to check out the full interview. I think you’ll find insights throughout.