💡Are business rules and management of them still relevant in today's landscape?

The short answer is YES! but why?

✨Business rules are the foundational knowledge of how things should work. As we move toward automated machines, processes, AI agents, apps, etc., the more important business rules execution and management become.

💢The notion of #ai #agents #llms #machinelearning is that they just know things as they are trained either with business rules documents or trained by the outcomes driven by #businessrules.

This is an entirely wrong assumption and understanding of how business rules should be treated.

🔸Training data: The problem is that #ai #ml works by training with sample data. A collection of datasets that, at best, (a) reflects a subset of business rules, not all of them. (b) The specifics of circumstances in which a decision was made are always missing. (c) It has no context and has no definition of rules. (d) It cannot be validated against the actual source of rules. So, in these scenarios, data at based is produced by a mysterious and unknown set of business rules.

🔸The rules knowledge in #ai: When data representing a form of numeric information, e.g., used in vector databases, large language models (LLMs), and similar technologies, they may be understandable for machines but not readable, explainable for, and understandable by humans. Additionally, it is not traceable back to the source of the actual rules, and more importantly, the ways transformers architecture works with that data does not enforce business rules; it is based on weighing at best.

So, what are the options?

💫This is the first time in the history of computers that we actually have an open standard for business rules and decision modeling that allows a standard symbolic approach to modeling business rules and makes them executable!

💣We should use this open standard and build business rules packages in different contexts and industries. They are executable and interoperable modules of business rules. Then, we will need to embed them everywhere into automated processes, machines, endpoints, AI robots and agents, etc. In this way, business rules are not learned based on some samples of outcomes or numeric values but on the actual symbolic models created from specifics of business rules by experts. They are executable by machines, understandable by humans, and traceable to the exact source of business rules.

Learn more about business rules and decision open standard Decision Model and Notation (DMN) at here.

Posted here.

Published June 13th, 2024 at 07:30 am