Mural gives companies like IBM, SAP and Deloitte a shared visual workspace. Internally, though, a data question could wait a week or two for an analyst, and only a few people knew which of the thousands of tables in its Amazon Redshift warehouse to trust. After moving to the Databricks Data + AI Platform and rebuilding its data model, Mural opened Genie Agents to every team. Now employees get answers in minutes, often without involving the data team at all.
A small analytics team fielded questions from the whole company
When someone in sales wanted to know what a customer had done the previous month, they usually asked the analytics or revenue operations team. Mural's analytics team of five to seven people supports the entire business, with one analyst embedded in each major department. New queries and dashboards were added to a queue behind committed work.
"It would easily take a week or two to get to a request," said Wendy Lam, Director of Data Engineering at Mural. "The team had to decide whether to answer one person's question or build the dashboard they had already committed to."
The data itself was part of the problem. Mural's Amazon Redshift warehouse held thousands of tables, many tracking the same metrics with slightly different calculations, and it lacked consistent standards for data quality and access. Knowing which table to trust depended on a handful of people, and any AI assistant pointed at that data would face the same confusion.
"With thousands of tables and many duplicate columns across those tables, Genie wasn't going to know which one to use," Wendy said. "Humans couldn't figure it out either."
Building a governed data foundation for Genie Agents
To open self-service analytics to every team, Mural first needed a governed, trusted data foundation. Databricks gave it one, along with a clearer path from raw data to curated business metrics.
Transitioning to Databricks for data governance
When Wendy took over data engineering, Mural had already chosen to move to the Databricks Data + AI Platform. Her first assignment was to confirm the decision still made sense. After doing the due diligence again, she reached the same conclusion. Databricks closed Mural's governance and access gaps with built-in capabilities and gave it an AI-ready foundation.
Unity Catalog was the first component Mural adopted. The team migrated its data in about a year, with help from Databricks solution architects, then rebuilt the data model from the ground up using a medallion architecture. It organized tables into bronze, silver and gold tiers, retired duplicates and added descriptions to tables and columns. Mural also runs Databricks in separate data residency regions in the U.S. and Europe, with an APAC environment in the works. Most employees now have access to the curated gold layer, and Genie Agents respect the same permissions.
"We didn't have to build anything extra for Genie," Wendy said. "The permissions were already in Unity Catalog, so we just turned them on."
Empowering teams with Genie Agents
Customer support got the first Genie Agent, replacing the slow work of digging through a large activity dashboard with a user-friendly AI agent. From there, adoption spread organically, as anyone with access to a Databricks table could easily create a Genie Agent and share it with peers. Employees in product, sales, marketing and finance found Genie Agents in Databricks and started asking questions without a formal rollout. Early on, the data team reviewed those questions to ensure the agents were answering them correctly. Accuracy improved as usage grew, and employees learned which questions to bring to Genie Agents and which still call for an analyst.
Rapid data access for employees and enterprise customers
With a governed foundation in place, Genie Agents changed how fast Mural's teams get answers and what they think to ask. Employees now pull numbers themselves, often in minutes.
Ten minutes to the numbers for an analyst evaluation
A major industry analyst firm invited Mural to join a market evaluation on short notice. Jeremy Moskowitz, VP of Marketing at Mural, needed account-level figures on AI feature adoption. With much of the company traveling near the end of a quarter, he opened a Genie Agent instead of adding a request to the data team's queue.
He described the metrics in plain language. The agent hit an error on its first attempt, then identified the problem and fixed it on its own. When one number looked off, Jeremy clarified what he meant in everyday terms. Within about 10 minutes, he had what he needed, and Mural's Chief Product Officer confirmed the figures against her own dashboard.
"I got the numbers I needed without having to bug anyone on the data team," Jeremy said. "It got me out of a really tight jam."
Sales is finding similar value. At an internal AI talent show organized by the sales team, an account manager was named a third-place finalist for showing how he uses Genie Agents to research accounts, find Salesforce contacts the team hasn't reached yet and identify potential business sponsors.
Fewer routine requests and more curious questions
Before Genie Agents, Wendy's team fielded two or three data requests a week. Each one took several rounds of back-and-forth to understand, and requests often landed with the wrong team first.
"We get a lot fewer questions about where to find a metric," Wendy said. "That gives my team time to improve the platform, and other teams get the answers they need right away."
The data team is putting Genie Code to work as well, using it to help generate metric calculations for the gold layer and build a few internal apps.
The questions themselves have changed, too. Pre-built dashboards answer the questions they were designed for. With Genie Agents, employees ask follow-up questions, experiment and explore areas that no dashboard was built to cover. Product managers also build their own AI/BI Dashboards to monitor performance and spot new opportunities.
"There's going to be more focus on conversational analytics and how users can self-serve, as opposed to creating dashboards for them," Wendy said.
The same governed data reaches enterprise customers
Mural's enterprise customers want clear evidence of how their teams use the platform, especially ahead of a renewal. They get it from an in-product ROI report, built on the same governed data, that shows how many employees are active, which use cases are most popular and how teams collaborate. Security and compliance teams can pull audit logs through Mural's reporting API, which now returns more than 50,000 activity records per request, up from 5,000. Customers who once asked a customer success manager to compile that data can now self-serve.
Databricks as the home for analytics
Teams across Mural increasingly treat Databricks as the system of record for analytics, replacing a mix of Salesforce reports and spreadsheets.
Mural is piloting a new definition of a product-qualified account, which lives in a Databricks AI/BI Dashboard that a pilot group of reps uses as its starting point for outreach. The pilot reps have acted on those accounts faster than before, and Jeremy credits the shared dashboard with making the new process easier to roll out. Next, he wants product usage and sales data side by side, giving marketing a clearer view of the journey from first-time user to paying customer.
"I'm excited to get that single pane of glass, with both user data and buyer data living in Databricks," Jeremy said.
The data team is also using MLflow on Databricks to classify customer usage of Mural and predict usage trends, with a conversion model in progress.
FAQ: Mural and Genie on Databricks
Genie Agents are part of the Databricks Platform. They let people ask questions about company data in plain language and get answers without writing code or waiting for an analyst, using data governed in Unity Catalog.