One of the most interesting topics with our ODM and ADS customers right now is the integration of machine learning (ML) and AI models. Many customers start by asking "how do I replace my rules with ML" but rapidly realize that that's the wrong question! Instead, they should be asking "how do I use rules and machine learning together". ML models can complement and improve existing rules-based decisions and rules-based decision platforms like ODM and ADS are great platforms for operationalizing machine learning models by wrapping them with the guiderails and "so-what" logic that turns predictions into business decisions.
Successful projects have identified three main success factors in this area:
- Focusing on the decision first, building a visual blueprint of the overall decision such as a DMN Decision Model
- Not allowing the rules or ML to be their own separate projects but combining the technologies into a single business solution
- Not trying to solve the problem in one iteration, but instead enabling continuous improvement of both the rules and ML models.
The first is particularly important - beginning with the decision itself. Failed ML projects tend to start by looking at the data and seeing what models can be built from it. Successful projects begin with a business outcome - a decision to be improved - and work backwards to identify the ML models that will help. These projects can use ODM or ADS to develop a rules-based decision framework that offers immediate value and which can then be used to operationalize the ML models as they are developed.
If you're interest in this topic, post questions as responses or drop me a note. Meanwhile, @Lori Brown posted about a great webinar on this topic Use machine learning predictions with decisions to improve your business outcomes starring @Jerome Joubert coming up May 9. I also worked on a paper with @GREGER OTTOSSON called "Operationalizing AI: Beyond AI Pilots with Digital Decisioning" that you can find on our website or on IBM's here (thanks to @Lori Brown for the edit).
#ArtificialIntelligence(AI)
#MachineLearning
#ibmchampions-highlights-home #ibmchampions-highlights
#automation-featured-area-1