Product descriptions:
Coop is a retailer with a unique mission: as a cooperative owned by over 4 million members, its goal isn't just profit, but feeding value back to its owners while prioritizing sustainability. To remain competitive during times of high inflation and shifting consumer habits, Coop needed to make data accessible to everyone, not just technical teams. By leveraging the Databricks Data Intelligence Platform, Coop transformed an internal hackathon idea into "AskCap", an AI-powered assistant that allows employees to query enterprise data via Microsoft Teams using natural language.
From "Cappuccino" to "AskCap", A Hackathon Story
The journey to democratizing data didn't start with a top-down mandate; it started at Coop's annual internal hackathon. Ankit Pratap Singh, Tech Lead on the Data Platform, led a team with a bold idea: "What if we cut the middleman when it comes to data?". The goal was to build a tool where business users could ask questions and get immediate answers without waiting for analysts to build reports.
The team brainstormed several creative names for their bot, including "CAPaccino," "CAPbot," and "CAPiche," before settling on the more professional "AskCap", short for Coop Analytical Platform. Despite Ankit having to manually patch the Databricks SDK overnight to bypass a bug just before the demo, the project was a hit. Although they finished in second place behind a sustainability initiative, the potential was undeniable, and the team moved quickly to turn the prototype into an enterprise-reality.
"We thought, 'What if we cut the middleman?' People can ask questions to a bot that appears real and gives you as good, if not better, answers than an analyst would," says Emil Nilsson, Head of Analytics at Coop and the one who facilitates data decision-making at the company.
A "Meta Layer" Powered by Genie and Unity Catalog
Turning a hackathon project into a trusted business tool required a robust engine. AskCap acts as a "meta layer" built on top of the Databricks Data Intelligence Platform, utilizing Databricks Genie for its native natural language capabilities. The bot is integrated directly into Microsoft Teams, meeting users where they already work.
Trust was paramount. To ensure the bot didn't hallucinate or reveal sensitive information, the team relied heavily on Unity Catalog for governance. The underlying data is structured in a "gold layer" with strict definitions, ensuring that when a user asks about sales or margins, the answer is accurate and compliant.
"The modeling is in the gold layer, and access is governed by Unity," explains Ankit Pratap Singh. "So the data we give to the end user has a higher standard and is trustworthy. Unity governance is in the grassroot level of the database, managing all kinds of privileges."
Changing the Culture of Decision Making
Since its launch to the head office, AskCap has achieved a "stickiness" rate of around 30%. It has become an essential tool for category managers performing "point lookups", such as checking the market share of specific products like bananas or honey, without navigating complex dashboards.
The tool is now evolving to become the "first responder" for data support. Instead of emailing the data team to ask why a report hasn't been updated, users can ask AskCap, which can instantly flag technical issues. This shift allows the engineering team to focus on building advanced models while empowering business users to explore data fearlessly.
"It’s better to ask 100 questions and maybe get one faulty answer than not asking a question at all," says Nilsson. "If you can create this network of many decisions happening all the time, we will be in a better place eventually."