From Manual SQL Entries to Instantaneous Insights on Sales Data

Customer Story

The client is a prominent name in the forestry, land management, and wood-products manufacturing industry. They oversee the entire lifecycle of their products: from planting and harvest to storage, finished-goods manufacturing, sales, delivery, and transportation. Their sales team relied heavily on their on-prem system (SQL Server)  to track order status, shipments, inventory contents, and sales figures. There was no centralized source to get all of that information, since their three main source databases covering different mills, product lines, inventory stores, and pricing had to be queried separately and reconciled with their all-company database before anyone could get a coherent view of the business. 

The company partnered with v4c.ai to design an intelligent assistant within Databricks that was built on their own data, trained on their business domain, and made accessible on any device through Microsoft Teams. It allowed the sales team to receive instant, accurate answers to any questions they needed to ask.

Problem Statement

The client's sales team often relied on information from their on-prem system, as there was no unified source of truth for them to understand the status of orders, shipments, inventory contents, and sales. What they needed was a single, simplified source of information they could come to for all of their data requirements. While that sounds straightforward, the complexity arises when you consider the 80+ tables spread across four key SQL databases that underpin their entire sales operation. Pulling a complete answer to even a basic question required touching multiple systems, writing queries across schemas that weren't designed to talk to each other, and stitching the results together manually.

Solution

The solution presented by v4c.ai was to build an intelligent chatbot in Databricks to answer sales queries. The process had three distinct requirements to be met: clean, unified data ready for business-grade insights; an analytical and AI layer established through the Genie Agent (formerly Genie Space); and a custom Microsoft Teams app that surfaces everything through a UI the users already know.

A - Medallion Architecture (Data Ingestion & Governance)

A connection was established to the source SQL servers by v4c.ai, and the required source tables were ingested into Databricks, then processed through a Medallion Architecture ETL pipeline:

• Bronze Layer: Raw Data: Ingested 80+ raw data tables from four key databases
• Silver Layer: Cleaned & Integrated Data: For the Silver Layer, the focus was on data cleaning and integration; for example, three different tables that set out the product characteristics were combined into one unified table, and the key tables were brought into standard form at this stage.
• Gold Layer: Curated Data for Genie: v4c.ai selected relevant columns and performed necessary aggregations to produce the final, refined data sources for the AI/BI Genie - 10 tables covering the full scope of sales operations and one table for some business terms such as internal product names, making it 11 tables in total

B - Genie Agent

The Sales Genie created by v4c.ai utilized 11 Gold Layer tables covering orders, shipping, products, pricing, and credit. Detailed metadata helped the Genie understand data context before querying.

Rigorous testing and client feedback encoded domain expertise into the system. This included defining answer formats, implementing SQL checks, providing example queries, and integrating a business glossary to ensure terminology accuracy.

C - Integration with Microsoft Teams

Note: This is a custom solution built out by v4c.ai for the client prior to DAIS 2026, when AI/BI Genie Spaces had limited integration options with apps like MS Teams. DAIS 2026 updates to Genie Spaces (renamed to Genie Agents) and Genie One have some of these features built-in, including a connector that directly connects MS Teams to Genie One.

v4c.ai integrated the Sales Genie directly into the client's Microsoft Teams tenant as a native app, putting it exactly where the sales team needed it. The resulting app was designed to offer users the same core capabilities they'd get in Databricks, and more:

• SSO Authentication: Users access Teams with their current credentials, and the authentication process in Teams makes use of both SSO and an On-Behalf-Of (OBO) token exchange, which means that the users do not need to handle separate credentials. To ensure strict security, token exchange succeeds only if the user's credentials have been granted access to the Genie Agent on Databricks. v4c.ai achieves this by managing access through a dedicated Azure Group linked to Databricks permissions, ensuring that only authorized personnel can access and interact with the app.
• SQL Transparency: The solution included the possibility of seeing the actual SQL query associated with any response.
• Feedback Loop: A feedback loop was implemented with thumbs up/down buttons, allowing users to flag incorrect answers and prompt the Sales Genie to regenerate.
• Export: An additional feature integrated by v4c.ai at the client’s request includes the ability to export the answers to Excel or CSV for use in reports.
• Monitoring Integration: User feedback surfaces directly in Databricks' monitoring tab; additional feedback comments and details, with timestamps and user information, were stored in a Databricks table for developer review.

Challenges 

• Matching the Databricks Experience in Teams: The biggest technical challenge was faithfully reproducing the Databricks Genie experience inside a Teams app. At the time of development, Databricks Genie Space API capabilities were limited, and features like written user-feedback storage required generating a PAT token on behalf of an internal service principal to collect feedback at submission time.
• Testing With Guest Credentials: The development team had to rely on sideloading since the client was using Microsoft Teams. The UI experience observed with guest credentials was at times different from that seen by end users with their standard accounts, which made it difficult to reproduce certain sign-in errors. Most of these inconsistencies were resolved by creating a temporary test account with standard credentials.

• Chart Rendering Limitation: Visual charts generated inside Databricks' Genie do not reliably reappear in the Teams app and result in display problems that impact other widgets and aspects of the app’s response. If users need charts, they have to log in to Databricks directly. Nevertheless, all tabular data renders properly and can be exported.

Impact

Prior to the deployment of this integrated solution by v4c.ai in April 2026, retrieving business insights required the sales team to manually write and execute complex SQL queries across disparate databases, which was a time-consuming and inefficient process that created a substantial barrier between raw data and actionable intelligence.

With the introduction of the integrated Genie Agents in the Microsoft Teams application, team members can now submit natural language requests, such as querying shipment volumes for specific products and locations, and receive instant, accurate responses. This seamless access has eliminated friction, empowering the team to focus on the insights they want.

Key Takeaways

v4c.ai's Sales Genie has succeeded in meeting user expectations and has integrated very smoothly into the user workflow. The clean, unified data provided by the Medallion Architecture was a great foundation, and the intelligence embedded in Microsoft Teams was a big driver for uptake. There were no new logins, no new interfaces, no friction at all. For many organizations, the problem isn't the data per se; it's the gap between the insights and the people that actually need them.

How v4c.ai Can Help

At v4c.ai, we help organizations transform their data into curated Genie Agents that deliver instant, business‑ready insights. By unifying siloed data into a reliable single source of truth, v4c.ai powers a business-trained Genie Agent customized with an organization's specific domain knowledge. By seamlessly integrating this intelligent assistant directly into everyday workplace tools like Microsoft Teams, v4c.ai delivers accurate, context-aware insights where teams already work, eliminating manual reporting and ensuring effortless adoption.

The result is a Genie Agent that doesn’t just answer questions, but reshapes how teams interact with data. This means faster decisions, greater confidence, and a culture where insights are accessible to everyone in the organization. The focus is on building solutions that are technically robust, aligned with the business, and easy to adopt.

‍