Introduction to AI Governance

You worked hard to develop a collection of AI models and now you are ready to deploy them. Before you rush to deploy your application, let me ask you important questions.

How do we sense and interpret a poorly working AI system? Presumably, you have developed a high–quality decision–making engine. Have you provided a mechanism to instrument your AI system as system gets deployed and learns from its use? What is your criteria for a good system and when do you alert your management if the AI system starts performing poorly.

Your AI system is deployed in a 24X7 environment where users are using it round the clock. What is your mechanism to fix the problems without tearing down the system. Can you develop a continuous learning module of your system which constantly learns from use, adjusts the model and provides sufficient logs to your contributing experts to monitor its learning.

The model starts to learn and now your experts are baffled. The learning is exactly not what they designed in the original system and now as they redesign their model, they have differences in opinion. How do you reconcile these differences and bring the best of your corporation.

Is there a tool or a system for life cycle management and can you employ that tool to monitor your AI models like the way you developed process monitoring tools for your governed processes.