What LLM agent SaaS solutions are available for business use cases?

Summary

• Sort agent options by who builds the agent: business users, analysts and domain experts who work without code, or developers who write code.
• On Databricks, Genie One serves business users, Genie Agents is the low-code option over Unity Catalog data, and Agent Bricks is the developer platform for custom agents written in code.
• Knowledge Assistant and Supervisor Agent are earlier agent builders that Databricks no longer recommends for new agents.
• For any hosted agent platform, check model choice, tool and data access, memory, observability, governance and cost control.

What LLM agent SaaS solutions are available for business use cases?

Agent options for business use cases differ most in who builds and maintains the agent. Some are experiences for business users, some are low-code builders for analysts and domain experts, and some are developer platforms for custom agents written in code. Databricks offers one of each: Genie One, Genie Agents and Agent Bricks. This page compares those three, then lists what to check in any hosted agent platform.

How do Genie One, Genie Agents and Agent Bricks differ?

Option	Built for	What the Databricks docs say it does
Genie One	Business users	A simplified Databricks UI where users view AI/BI dashboards, ask data questions in natural language and use custom-built Databricks Apps, without navigating compute, queries, models or notebooks.
Genie Agents	Analysts and domain experts who configure agents, and business users who ask them questions	A domain-specific natural-language chat interface over data registered to Unity Catalog. Users get SQL queries, results tables and visualizations, and generated queries are always read-only.
Agent Bricks	Developers who build custom agents in code	The Databricks agent developer platform. You bring an agent built with any framework or harness and choose any model.

Knowledge Assistant and Supervisor Agent are earlier agent builders. Databricks no longer recommends them for new agents. It points new low-code agents to Genie Agents and new agents written in code to the Agent Bricks CLI (Beta).

What should you evaluate in any hosted agent platform?

Hosted agent platforms run agent infrastructure for you. On Databricks, Agent Runtime runs your agent server on Databricks Apps and Databricks-hosted serverless compute, so you deploy code instead of provisioning servers. Ask the middle-column questions of any vendor, then compare the answers with the Databricks column.

What to evaluate	Question to ask	How Databricks documents it
Model choice	Can you switch models without rewriting the agent?	Unity Gateway gives one API for frontier and open models, including models that Databricks hosts through Foundation Model APIs and models from external providers. You can switch models without changing agent code.
Tools and data	How does the agent reach business data and actions?	Databricks-managed Model Context Protocol (MCP) servers for Genie Agents, Databricks AI Search, Unity Catalog functions and SQL, plus custom and external MCP servers.
Memory	Does the agent recall earlier conversations?	Managed agent memory and managed agent sessions (both Beta) are stores backed by Lakebase that work with any framework.
Observability	Can you inspect every step?	MLflow Tracing records the inputs, outputs, latency, token usage and cost of each intermediate step. Evaluation uses LLM judges, custom scorers and expert feedback.
Governance	Who can use which model, tool and data?	Unity Catalog governs the data, tools, models and agents that your agents use. Unity Gateway governs access to models, MCP servers and skills.
Cost control	Can you cap spend?	Unity Gateway enforces rate limits and cost controls, and admins can analyze cost across models, principals and tags.

How does Databricks govern what agents can do?

Unity Gateway governs access to models, MCP servers and skills, and it enforces rate limits and usage tracking on the calls agents make to them. Tools can run with the permissions of the app's service principal or of the user who sent the request. Service policies in Unity Gateway (Beta) add guardrails by allowing or denying a request based on its content, and on an MCP service an input policy can also require human approval.

Which option fits which business task?

• Asking questions of governed tables in natural language: a Genie Agent. The Databricks docs use a sales manager who asks for the top-selling product over time and gets a visualization.
• One place for dashboards, data questions and apps for business users: Genie One.
• A workflow with custom logic that calls several tools: a custom agent on Agent Bricks. A custom agent can call Genie Agents or Genie One as tools.

FAQs

Should I build on a model API or use an agent platform?

The Agent Bricks docs name five areas that teams otherwise assemble and operate themselves: model access, compute, context, quality, and governance and control. If your agent will call tools, keep state or touch governed data, expect to cover all five.

Do I need to fine-tune a model to build an agent?

No. Agent Bricks takes an agent built with any framework or harness and lets you choose any model. Whether to fine-tune a model is a separate decision.

How do I keep agent spend under control?

Unity Gateway enforces rate limits and cost controls on model and MCP requests. Admins can set rate limits and spend caps, and analyze cost across models, principals and tags.

What is the difference between an agent and a single model call?

A single model call returns one response. An agent runs a loop: 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.