-
Build comprehensive data-collection policies with real-time, event-based collection for training AI systems to accurately predict outages and manage systems security and health.
-
Supplement internal data with external data to enhance predictive power and to increase the accuracy of predictive algorithms. For example, use social media chatter about a sports event combined with geolocation data to better predict peak network traffic and preempt an outage.
-
Fully automate the low-risk, high-volume processes. However, for high-risk, high-impact decisions, let AI recommend the action and have a human operator make the final decision.
-
Finally, make diversity and inclusion a core value and design principle when recruiting the teams responsible for building AI systems. This ensures that systems and predictions are not biased.