What changed
Google made Memory Bank profiles GA on Gemini Enterprise Agent Platform. Teams can define a static schema whose fields are populated and updated by models for low-latency access to evolving user or entity information.
Why it matters
Structured memory can avoid repeated retrieval and give agents consistent fields, while model-written profiles can also preserve errors, stale facts or sensitive attributes at scale.
What to do
Start with a synthetic profile schema, test extraction and updates across contradictory events, and require provenance, field-level retention, correction and deletion checks before storing real user data.
Caveat
Google establishes GA and the schema mechanism; it does not independently establish extraction accuracy, conflict resolution, privacy suitability or latency gains for a specific workload.