Conversational Analytics: How Large Language Models (LLMs) Unlock Self-Service Analytics, From Questions to Dashboards

karla
Karla Nussbaumer
5 Min Read
/
/
/
Conversational Analytics: How Large Language Models (LLMs) Unlock Self-Service Analytics, From Questions to Dashboards

For years, the promise of self-service analytics has remained just out of reach. Business leaders want answers in real time, but often the data is locked away behind technical queries, ETL processes, or dashboards that do not quite answer the next question. Analysts spend hours writing code, and non-technical users are stuck waiting for insights.

What if you could just ask your database a question in plain English and get the answer immediately, along with a dashboard to visualize it?

That is exactly what happens when you connect a Large Language Model (LLM), such as Claude, ChatGPT, or a private model to Yellowbrick.

In this video, you will see how an LLM is connected to Yellowbrick SQL Data Platform to access all the data within the data warehouse. Check how a user without technical knowledge can get insights by asking plain questions.

Check the script directly on Claude.

A New Way to Ask Questions

In the video we show how a non-technical person asks a question in the LLM: “What was my revenue in Ohio in March 2025?”

The LLM explores the Yellowbrick database by examining schemas, describing tables, and assembling the correct query to find the answer to the question. It finds the table for revenue, the store table for locations, and the date table for time filters. Within seconds, you have your revenue for Ohio in March 2025.

What is the business impact? 

You do not need to know a single line of query language or wait for an analyst to create a new dashboard. Anyone in the business can ask the question and get the answer directly. This data accessibility shortens the cycle from curiosity to decision-making.

In the video, you can see how the user wants to get more information and the LLM keeps context, knowing the user still needs information from the Ohio revenue in March 2025, and it brings the customer table to generate the next query. This means that the analysis can follow a natural conversation, not a series of ticket requests or static reports. Each answer leads directly to the next question, the way business conversations actually work.

Now, business leaders and their teams can follow their train of thought in real time, explore the “why” behind the numbers, and make faster decisions. What used to take hours or days of back-and-forth with technical teams now takes minutes.

Beyond Queries: Blending Structured and Unstructured Data

Data does not just sit in tables. Some of the most valuable insights live in unstructured sources such as PDF files, formulas documented in wikis, or policies shared in emails.

In this demo, the large language model tapped into the Yellowbrick vector store, which holds unstructured documents. It discovered a file describing the company’s “loyalty-adjusted effective revenue” formula. The model reads unstructured content, understands the rules, and applies them directly to the structured data inside Yellowbrick. The result was not just a revenue figure, but a smarter measure of performance comparing effective revenue and true customer value.

The business impact: organizations are no longer limited to the data in rows and columns. They can merge the hard facts of structured data with the nuance of unstructured knowledge, creating a single source of truth that reflects how the business actually operates.

Visualizing The Data

Numbers are good, but visuals change the game.

When asked, “Can you give me this in a visual format?”, the model generated an interactive dashboard on the spot. It compared base revenue with effective revenue, broke it down by customer tier, and showed how rankings shifted for the top customers.

From question to query to business rules to visualization. All in one seamless flow.

Executives and managers can immediately see patterns, spot opportunities, and communicate findings to teams without waiting on business intelligence reports. This makes insights fast and actionable.

Why Yellowbrick Makes This Possible

LLMs can only be as powerful as the database they connect to. Yellowbrick is uniquely suited for this kind of dynamic and unpredictable querying because:

  • There is no need for pre-built indexes or aggregates. Yellowbrick handles raw, ad-hoc queries directly.
  • High performance at scale. Joins across fact, store, date, and customer tables return quickly.
  • Queries may be unpredictable, but the costs should not be. Yellowbrick handles these queries in an affordable and consistent way. 

The Business Benefits

  • Empower business users to ask questions without learning a query language.
  • Accelerate insights, no waiting on data teams to build dashboards or canned reports.
  • Unlock analyst productivity by letting them focus on deeper strategy instead of one-off requests.
  • Agile decision-making process. Go from question to decision in minutes, not days.

The Future of Analytics is Conversational

This demo is more than a proof of concept. It is a glimpse of how analytics will work going forward: natural, conversational, and dynamic.

With Yellowbrick, business leaders do not just get faster queries. They get smarter insights powered by both structured and unstructured data. And they do not need to wait for information technology teams to build the next dashboard.

Ask the question. Get the answer. Make the decision.

Ready to see it in action? Try Yellowbrick to see how it can power your large language model-driven analytics and unlock true self-service for your teams.

Sign up for our newsletter and stay up to date