Analyst Agent is the simplest way to create and share AI agents to provide self-service analytics.

We’ve designed Analyst Agent to provide full control to the builder and to be able to work in complex, messy data environments (so every enterprise data environment). Here’s what makes it unique:

• Analyst Agent doesn’t query your data source. Instead, it leverages datasets that you’ve created and curated in Smartbooks.
• Even though Analyst Agent doesn’t go directly to the source, you can still keep the data fresh by scheduling updates
• Analyst Agent will not answer any questions for which it doesn’t have the data

​

How Analyst Agent works

Analyst Agent is an AI agent designed specifically for data analysis. Analyst Agent is the simplest and quickest way to build agents designed to handle self-service analytics request while giving the builder full control over what data the agent can access. Think of Analyst Agents as agents specialized in answering specific questions.

Rather than trying to build an AI that can answer any business question, you can create dedicated agents:

• A marketing campaign agent that only looks at marketing campaign data
• A product agent that only looks at product data
• A customer success agent that only looks at account data
• etc.

This modular approach makes it faster, safer and easier to leverage AI for data analysis.

​

Preparing data for Analyst Agent

When you configure Analyst Agent, you’ll simply be selecting which DataFrames you want it to access. So preparing your data is as simple as creating DataFrames.

In a start by uploading or querying your data. Then use SQL or Python to create the DataFrames that you want to make available to the Analyst Agent. This is where you can wrangle your data to get it just right.

Here are a few tips to maximize AI accuracy:

• Don’t aggregate data unnecessarily. Think about the types of questions your users may ask, and query the data at the right level of granularity.
• Only pull in the fields that are needed (hint: don’t select * from ...)
• Filter out bad records or missing values
• Rename fields to make them as clear as possible to the AI. If you have an amount field, this is a good opportunity to rename it opportunity_amount_usd

​

Configuring Analyst Agent

Once you’ve curated your datasets, it’s time to configure and share your Analyst Agent.

Go to Report Builder in the top navigation bar.

1. Configure your report as needed
2. In the right hand configuration panel, add or remove whatever data artifacts you want the Analyst Agent to be able to access. Remove any excess DataFrames and narrow it down to the core, clean datasets
3. Relabel your data artifacts and provide more detailed descriptions. These descriptions are important because they’re used by Analyst Agent and they’re also the description that your users will see.

Once you’re finished configuring the AI, schedule updates and publish your report.

​

Analyst Agent FAQ