N🚫T all business decisions require #machinelearning #ml or #predictiveanalytics of any sort. They need no prediction at all.

✨In fact, quite the opposite of what many think, we often apply predictive technologies to automate business decisions unnecessarily. Just because it's the latest trend, we get caught up in the hype, or we are lazy and don't want to do the hard work (using #data is easy), or we don't know the other approaches.

Many business decisions are purely rule-driven, and they are often formulated by domain knowledge such as engineering, legal, finance, product development, etc. Sometimes, a category of business decisions is constraint-related problems, e.g., transportation, routing, resource allocation, planning, scheduling, etc.

This is why it's vital not to begin with specific techniques such as rules, constraints, data, machine learning, optimization, etc. If it is about business decisions, 🫵you MUST start with business decisions.

🤔How do we start with business decisions?
🔹Select a specific business decision
🔹Use the decomposition technique to break down a complex decision into smaller decision units
🔹Identify the relationship and dependencies between those decision units
🔹Specify what inputs those decision units need
🔹Set the metrics of those decision units (when applicable)
🔹Understand what situations those decision units are in play

By this point, you have established a holistic view of business decisions by modeling them as a hierarchical Decision Graph.

🔎Now look at each decision unit, capture their conditions and actions (either rules or constraints) for each decision unit, and the Decision Graph will do the final job. It coordinates between all the decision units to execute them and create the final outcome.

In this process, you can also integrate #machinelearning #ml #predictiveanalytics or any other technology for that matter.

🚀Based on the nature of those decision units in the Decision Graph, you may choose the appropriate technique, whether it is #rules, #optimization #heuristic, or even #llms and #ml. But without decomposing and creating the decision model to unfold what you don't know, it is very easy to just be caught up in the hype of #ai #ml #llms and other trends.

#decisionintelligence

Posted here.

Published April 26th, 2024 at 07:30 am