What is the agentic loop, and how does business context improve it?

Summary

• An agentic loop is how an agent works: it reasons about the request with a model, calls tools to gather information or take actions, observes the results, and repeats until it can respond.
• Business context is what your data means: definitions, relationships, calculations, authoritative sources, and rules. Without it, the model "fills the gap with inference."
• Context improves the loop when the right, authoritative context reaches the agent at the point where it is reasoning. An ontology alone does not make an agent accurate.
• Genie Ontology, the Databricks unified context layer, is in Public Preview. The Genie One MCP server, which is generally available, lets agents ask business questions grounded in it.

What is the agentic loop, and how does business context improve it?
An agentic loop, which the Databricks documentation calls the agent loop, is the cycle an AI agent runs to achieve a goal. The documentation says the agent "reasons about the request with a model, calls tools to gather information or take actions, observes the results, and repeats until it can respond." Business context improves that loop by telling the agent what your data means, which sources to trust, and what the person asking may see. Databricks groups that context in Genie Ontology, which is in Public Preview.

What is the agentic loop?

Databricks defines an agent as "an AI system that perceives a request, decides what to do, and acts to achieve a goal." Three parts make up the loop:

Part	What it does
Model	The large language model reasons about each step and decides which action to take next
Tools	Functions that the agent calls to retrieve data, including your business data, or to change the state of another system
State	The current session, which is the conversation or task in progress, and memory, which is what the agent learns and recalls across sessions

Why do agents need business context?

A Databricks blog post on ontologies explains the gap. A schema describes how data is structured. An ontology describes what that data means in the context of the business. It connects technical assets such as tables, metrics, and queries to business concepts: definitions, relationships, calculations, sources of expertise, and rules about how those concepts should be interpreted.
In the post's revenue example, a schema might tell an agent that a table contains net_rev, gross_rev, and recog_rev. An ontology can help it understand which definition of revenue applies to the question, which source Finance considers authoritative, how the calculation is normally performed, and whether the person asking is even permitted to access it. When that context is missing, the model doesn't know which internal source contains the ground truth, so it fills the gap with inference.

How does business context improve the loop?

The same post states the limit: "An ontology by itself doesn't magically make an agent accurate. What improves accuracy is supplying the agent with the right, authoritative context at the point where it's reasoning."
Databricks advises that humans explicitly define and govern the small set of concepts that cannot be wrong, such as revenue, compliance rules, and core KPIs, while the broader ontology learns the long tail from how the organization operates.
With the Genie One MCP server, an MCP client or agent sends a natural-language question, and Genie searches your enterprise data, writes SQL, and returns an answer grounded in Genie Ontology with deep links back to your cited Databricks sources. Databricks recommends it for agents that answer business questions across the workspace in natural language. Unity Catalog permissions are always enforced, so users and agents can only query data they're allowed to access.

What is Genie Ontology?

Genie Ontology is the unified context layer that gives the Genie family of AI experiences a business-aware map of your organization. The docs label it Public Preview. It has two parts:

Part	What it is
Unity Catalog semantics	Modeled context that you define, govern, and certify through metric views, domains, and Pages. Pages are in Beta
Inferred context	A map of snippets that Genie automatically extracts and maintains from your existing assets and usage, including metric views, dashboards, SQL queries, and Genie Agents

One example snippet in the docs: a "qualified lead" only counts once a demo is booked. Each snippet gets an authority score based on where it was generated from, how often it is used, and how fresh it is. Snippets are gated by Unity Catalog permissions, so Genie only uses snippets extracted from assets you have permission to see. Genie One and Genie Code search the same ontology, and citation icons show which knowledge sources Genie used.

How do you put business context to work?

Databricks' guide to operationalizing Genie Ontology describes a layered path:

1. Add table descriptions and column comments in Unity Catalog, and apply tags.
2. Declare relationships, build a semantic model with metric views, and organize and document it with domains and Pages.
3. Certify trusted assets and deprecate stale ones.
4. Test a representative set of business questions for each priority domain in Genie One.

When an answer fails, trace it back to the underlying source or context and fix the root cause. The fix belongs in a Page when the problem is a business definition or synonym, and in a metric view when the problem is a governed measure, dimension, relationship, or calculation. Start with one high-value domain.

FAQs

Does Genie Ontology make an agent accurate by itself?

No. Databricks says an ontology by itself doesn't magically make an agent accurate. The agent needs the right, authoritative context at the point where it is reasoning.

Is Genie Ontology generally available?

No. The Genie Ontology docs label it Public Preview, and Pages are in Beta. The Genie One MCP server is documented as generally available.

Can an external agent use Genie Ontology?

Yes, through the Genie One MCP server. Databricks notes that external agents can access Genie's intelligence through MCP. Unity Catalog permissions still apply.