Text decision tree

Decide which parts of an AI platform belong beside governed data and which can stay in external services.

Adapted from Chapter 122 of Decision Trees: Frameworks for Technology Choices. Sources reviewed 2026-11-05.

START

Q0 Critical AI data already governed in Databricks?
  YES -> Databricks-centered candidate. Keep valuable data and permissions close. Then Q1.
  NO -> Otherwise compare an external platform first and justify any new data movement. Then Q1.

Q1 Preparation, evaluation and lifecycle tightly coupled?
  YES -> Evaluate Mosaic AI capabilities. Compare the integrated data-to-AI lifecycle. Then Q2.
  NO -> Otherwise compare a hybrid boundary with a simpler external application layer. Then Q2.

Q2 Required model or tool missing from the candidate?
  YES -> Add an external component. Retain governed data and clear interfaces. Then Q3.
  NO -> Otherwise keep the candidate without adding another platform. Then Q3.

Q3 Custom training or model serving required?
  YES -> Evaluate the MLflow and serving path. Use the necessary training, registry and endpoint tools. Then Q4.
  NO -> Otherwise a supported hosted model interface may be sufficient. Then Q4.

Q4 Permissions, quality, latency and total cost pass?
  YES -> Pilot the measured boundary. Assign ownership of every cross-platform interface. Then END.
  NO -> Otherwise revise the failed boundary or component and evaluate again. Then STOP.

END: Adopt the boundary that passes the workload checks. A Databricks-centered data plane can still use an external model or application runtime.

AI-generated chapter illustration

BookWriter-generated chapter 122 illustration

This AI-generated illustration gives a visual overview. It can simplify labels or imply connections that do not preserve every decision. For exact branches, prerequisites and stop conditions, follow the video graph and the text tree above.

Reference links

Comuvia Tech on YouTube

Related decision trees