CONCEPT Cited by 4 sources

Autonomy (enterprise agents)¶

Definition¶

Autonomy is the shift from prompt-response one-shot LLM calls to code-in-control end-to-end flow, where the LLM is empowered to make function calls, choose which database to read, choose the number of passes, and divide and conquer tasks — all without the developer writing explicit code paths for every possible execution trajectory.

Canonical statement (Source: Gallego 2025-04-03):

"Autonomy is where code is in control of the end-to-end flow vs explicit code paths you've written."

"Instead of sending a prompt to an LLM and getting a one-shot answer, you empower the model to make function calls, choosing which database to read data from, choosing the number of passes and how to divide and conquer tasks, etc."

Enterprise autonomy is the hard case¶

Gallego's framing: enterprise autonomy is not just the agentic-loop pattern everyone has seen; it is the specific case where private data — the thing that gives an enterprise "an unfair advantage in the market" — must be accessible to the agent without leaving the firewall. That constraint is what makes enterprise autonomy architecturally distinct from consumer agents:

"Autonomy is the codifying of business processes with private data along model advances. Think of it as public models already having consumed all of the public data and being suspended in their training because they didn't have access to your specific data. You can unlock this frontier either by fine-tuning or enriching context via RAG and other similar techniques that give models access to the specific context for the task at hand."

The architecture consequence is sending models to the data rather than sending data to model APIs.

What autonomy requires from infrastructure¶

Gallego enumerates three infrastructure requirements for enterprise autonomy:

Connectivity — to data, internal systems, and models via open protocols (his pitch for MCP as the integration layer).
Durability of execution — for agent-to-agent communication and long-running workflows; see concepts/durable-execution.
SDK ergonomics — a Ruby-on-Rails-style developer experience that glues it together; his pitch for the Redpanda Agents SDK.

Productization layer (2025-10-28)¶

Seven months after the April essay, Gallego's 2025-10-28 ADP announcement names the commercial shape of enterprise autonomy: the Agentic Data Plane (ADP). The productization adds two substrate layers the April essay only implied:

A governance layer — governed agent data access with [[patterns/on-behalf-of-agent- authorization|OBO authorization]], DLP hooks, per-agent consent workflows, and replayable audit trails. The April essay named governance as necessary ("autonomy is the codifying of business processes"); the October post names it as the primary product surface ("The fear from CIOs is not the code of the agent itself, it is governance").
A query-engine layer — the acquired Oxla C++ distributed query engine with PostgreSQL wire protocol + separated compute-storage + Iceberg-native. Positions SQL-over- Iceberg as the agent-context substrate: "SQL is the best mechanism to filter and aggregate while the model summarizes."

The April-vs-October distinction on the wiki: April canonicalises what autonomy is; October canonicalises what it costs to deploy safely (governance surface) and what it queries for context (analytical substrate).

concepts/agent-loop-stateless-llm — the planner-acts-observes- plans-again loop that autonomy canonicalises at enterprise scale.
concepts/agentic-data-access — Warfield's S3 Files framing of the agent-storage-friction question.
concepts/agentic-development / concepts/agentic-development-loop — the developer-facing sibling (code-authoring agents).
patterns/tool-call-loop-minimal-agent — the 10-line core of what enables autonomy.
concepts/sub-agent — the recursive-autonomy shape where agents spawn agents.

Seen in¶

sources/2026-02-10-redpanda-how-to-safely-deploy-agentic-ai-in-the-enterprise — Akidau talk-recap post canonicalising the risk-side dual of autonomy: enterprise agents default to the chaotic column of the D&D alignment chart ("the best outcome at this point is that of 'chaotic good' — because you don't know what you don't know"); governance
auditing + replay infrastructure is what moves them leftward toward lawful. Reframes Gallego's capability-side autonomy thesis with a rhetorical governance hook aimed at enterprise decision-makers.
sources/2025-10-28-redpanda-introducing-the-agentic-data-plane — Gallego's productization follow-up naming the commercial shape (Agentic Data Plane), the governance-first framing, and the Oxla acquisition as the query-engine substrate for agent context.
sources/2025-10-28-redpanda-governed-autonomy-the-path-to-enterprise-agentic-ai — companion 2025-10-28 governance-framing post. Reframes autonomy as requiring a governance layer to be enterprise- deployable; canonicalises AAC (no-long-lived-creds + pre-and-post-I/O policy check + fine-grained temporary access) and durable event log as audit envelope as the AAC + audit substrates under the autonomy capability.
sources/2025-04-03-redpanda-autonomy-is-the-future-of-infrastructure — Gallego's founder-voice framing positioning autonomy as the 20- year-arc successor to batch → streaming → continuous-computation for enterprise software, gated on private-data access.