Thoughts on the Software Ecosystem post Databricks DAIS (SNOW, PLTR, MSFT, Apps)

Research evidence excerpt

Thoughts on the Software Ecosystem post Databricks DAIS (SNOW, PLTR, MSFT, Apps)

Goldman Sachs Americas Technology: Software

o Postgres for agents: Databricks framed Postgres as increasingly central in an

agent-driven world, where workloads operate at machine speed and require

cost efficient, deterministic, and inherently branchable environments.

Lakebase introduces primitives such as near-instant database branching,

enabling agents to safely fork datasets, experiment, write changes, and revert

to prior states as needed. In our view, the emphasis on Postgres is critical, as

agentic applications require not just read access to enterprise context, but

governed, low latency transactional systems where agents can act safely, at

scale, and with repeatability.

o Data engineering: Lakeflow: New innovations in Lakeflow focus on

minimizing upstream complexity through a unified, declarative framework for

ingestion, transformation, and orchestration built natively on Spark. Lakeflow

abstracts pipeline creation via no code and declarative constructs, while

maintaining open, non proprietary execution under the hood. The innovation

shifts toward operationalizing data engineering at scale: automated

orchestration across >50 integrations, and tight integration with agents via

Genie Code and Genie Ops. These innovations are important in alleviating data

engineers’ ongoing maintenance burden, which has historically consumed a

disproportionate share of data engineering resources.

o Data science/AI: Databrick’s core lakehouse and data science positioning

remains central as it provides the enterprise context layer for AI. The

introduction of Genie Ontology and OntoRank create a structured graph

across enterprise data, documents, and usage patterns, improving agent