Back office ops · Production

zeb builds SuperInsight GenAI self-service reporting engine on Databricks, reducing data analyst workload by 80–90%

The problem

Enterprise customers of zeb had teams of data analysts manually managing a substantial backlog of data requests, limiting data access for smaller operational teams and creating bottlenecks for analyst capacity.

Workflow diagram · grounded in source
1
User submits data request
trigger
“End users can send a request through email, Slack or other communication channels”
2
DBRX intent classification
ai_action
“The request is fed through a DBRX model that classifies the intent”
3
Vector Search context retrieval
ai_action
“Databricks Vector Search provides relevant context from a knowledge base stored in Unity Catalog”
4
Industry-adapted generation
ai_action
“The Model Serving endpoint combines another DBRX model with a fine-tuned adapter based on the customer's industry”
5
Output delivery
output
“The final output is then either sent to a data warehouse to generate a CSV file, to a deployed AutoML endpoint or to a reporting tool to generate a visual report”
Reported outcome

SuperInsight reduced data analyst workload by up to 80–90%, delivered 40% cost savings for customers, and drove a 72% uptake in reports requested compared to the previous manual process.
zeb also reduced its own solution development time by 40% using the Databricks Platform.

Reported metrics
Data analyst workload reduction80–90%
Customer cost savings40%
Uptake in reports requested vs. manual process72%
Solution development time reduction40%
Reported stack
Databricks Data Intelligence PlatformMosaic AIAgent Bricks Custom AgentsMosaic AI TrainingModel ServingUnity CatalogDBRXVector SearchAutoMLSlackTeamsServiceNowJira
Source
https://www.databricks.com/customers/zeb
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

SuperInsight reduced data analyst workload by up to 80–90%, delivered 40% cost savings for customers, and drove a 72% uptake in reports requested compared to the previous manual process.

What tools did this team use?

Databricks Data Intelligence Platform, Mosaic AI, Agent Bricks Custom Agents, Mosaic AI Training, Model Serving, Unity Catalog, DBRX, Vector Search, AutoML, Slack.

What results were reported?

Data analyst workload reduction: 80–90%; Customer cost savings: 40%; Uptake in reports requested vs. manual process: 72%; Solution development time reduction: 40% (source-reported, not independently verified).

How is this back office ops AI workflow structured?

User submits data request → DBRX intent classification → Vector Search context retrieval → Industry-adapted generation → Output delivery.

Related back office ops cases