This level represents a system in which ‘advanced techniques’ have been introduced to make recommendations, provide insights, and in some cases perform actions. The expectation is that predictive and prescriptive analytics are present, and that model development is due, at least in part, to machine learning. The result is that the system focus is inferred decision-making. An additional expectation is that these ‘advanced techniques’ are weighted towards recommendations and insights, so we refer to this type of automation as recommendation-driven automation.

Some characteristics of this type of automation are:

• Collects large amounts of telemetry data with the intent of using it as part of the decision-making process.
• Typically centralizes and curates the storage of telemetry data (cleansing, normalization, and so on).
• Uses data visualization tools.
• Uses analytics extensively to predict when a resource will be depleted, and in some cases automatically replenishes the resource.
• Uses analytics extensively to predict changes in usage patterns and in some cases, adjusts resources automatically to address the change.
• Uses analytics to detect anomalies, to predict what that anomaly could mean, and in some cases, to take action to address it.
• In many cases, the organization has embraced new operational patterns (MLOps, DataOps).

Take away: This level can be summed up by the statement “We have introduced the power of AI/ML into our systems”.