In my previous article, Many Companies Use AI. Few Know How to Build an AI-Native Enterprise Data Platform, I discussed how to integrate AI into enterprise data platforms and shared common challenges in applying AI to data engineering workflows, along with practical solutions. I highlighted three key elements of a practical enterprise AI architecture: data agents, AI-powered QA, and AI governance.
To dive deeper into data agents, I built a demo called the Avocado Sales Analytics Agent. In this article, I’ll walk you through the complete process of building it step by step—no code required.
What Is a Data Agent?
A data agent is an AI-powered conversational interface that allows business users to ask questions in plain language and receive accurate answers by querying data stored in a data warehouse.
Instead of waiting for data analysts to write complex SQL queries and generate reports, users can simply type: “How much is total TPV in Southeast Asia last year?” and get an instant answer like “$60 Billion”.
Choosing the Right Approach
There are two primary ways to build a data agent. The first is to build from scratch using open-source orchestration frameworks such as LangGraph/LangChain, CrewAI, or LlamaIndex. This approach gives you full control over the agent’s memory structures, strict business logic rules, and complex multi-agent execution loops.
For beginners, the second approach—deploying a data agent within a cloud data platform—is more practical and faster to implement. Most major cloud data platforms now offer native, out-of-the-box data agents:
- Snowflake: Snowflake Cortex Agents are low-code agent pipelines fully hosted inside Snowflake, allowing users to ask natural language questions over secure enterprise data warehouses via Snowflake Intelligence.
- Databricks: Databricks Genie is the managed conversational data intelligence tool within the Databricks ecosystem.
- Microsoft Fabric: Fabric Data Agents support direct data connections to lakehouses, warehouses, KQL databases, and Power BI semantic models.
- Google Cloud: BigQuery provides Conversational Analytics features, which I chose for my demo because they are fully accessible during the free trial and can be set up easily using a personal Google account.
For the Avocado Sales Analytics Agent, I used the Avocado Prices dataset from Kaggle, published by Justin Kiggins using data from the Hass Avocado Board and available under the CC BY 4.0 license.
Building a Data Agent With No Code
Google Cloud’s BigQuery offers the Conversational Analytics API, which allows you to build conversational data agents on top of BigQuery datasets. Here’s the step-by-step process I followed:
Step 1: Upload the Dataset
Download the Avocado Prices CSV file from Kaggle and upload it to BigQuery. Once uploaded, it’s crucial to understand the data schema of the tables—including table names, column names, data types, relationships, and business meanings. The agent needs this understanding to “learn” how to analyze the data correctly. Thorough data comprehension is essential before you can “teach” the agent effectively.
Step 2: Create the Agent
Navigate to BigQuery → Agent, click “Create Agent,” enter the agent name and description, and select your dataset as the Knowledge source.
Step 3: Write Effective Instructions
The instructions are the most critical part because they guide the AI in understanding the business context, defining response formats, and setting boundaries for query generation. This step requires careful crafting—a well-designed instruction set ensures the agent handles user questions accurately and consistently. In my demo, I included specific guidance on how to interpret date fields, handle null values, and format monetary figures to avoid ambiguity.
By following this no-code approach, you can have a functional data agent up and running in hours, not weeks, empowering business teams to self-serve insights from enterprise data. In the next part of this article, I’ll cover how to test the agent, refine its responses, and integrate governance policies to ensure safe and reliable usage.