
How Unibail-Rodamco-Westfield Democratizes Access to Its Data with Generative AI
KEY FIGURES
~90% first-try query accuracy
–80% routine Data team requests
Seconds instead of days per answer
CHALLENGE
All of URW's data is hosted in Snowflake. It is available but practically inaccessible to most employees. Getting a simple answer ("Which are our busiest centers this quarter?") requires either mastering SQL or requesting assistance from the Data teams, creating a bottleneck that slows down decision-making at all levels. The same kind of bottleneck existed at Adelya, where sales reps relied on senior colleagues for technical questions.
Prior to the project, several key challenges were identified:
- Enable business teams to query data without technical skills
- Reduce reliance on Data teams for routine queries
- Accelerate decision cycles by eliminating intermediary delays
- Ensure Snowflake is adopted and usable by non-technical users
SOLUTION
Qolaig designed and deployed a Natural Language to SQL POC: a conversational interface that transforms a question asked in natural language into an SQL query executed directly on Snowflake, then returns the answer in a clear and understandable way without the user needing to know the database structure.
01. From Question to Data, Without SQL
The user asks their question as they would phrase it to a colleague:
- "How many stores do we have in the United States?"
- "What is the average occupancy rate per shopping center?"
- "Which are the busiest centers this quarter?"
The system understands the intent, identifies relevant tables, generates the appropriate SQL query, executes it on Snowflake, then interprets the results via a second AI layer before returning them in natural language.
02. A Business Glossary at the Core of Reliability
The project's main challenge wasn't technical; it was semantic. Every company has its own vocabulary: in-house indicators, internal synonyms, specific calculation rules, and undocumented table relationships.
Qolaig built a custom business glossary for URW, enabling the AI to understand the exact definitions of indicators, the synonyms used by field teams, and the group's specific calculation rules. This layer is what ensures the relevance and reliability of the generated queries. The same semantic rigour drives the learning assistant deployed at Liora.
RESULTS
- Natural Language Data Access for All Business Teams
- Reduced Workload for Data Teams on Routine Queries
- Easier Adoption of Snowflake by Non-Technical Users
- Accelerated Decision-Making by Eliminating Intermediation Delays
RETURN ON INVESTMENT
Every question that reached the Data teams consumed skilled time on low-value-added tasks. By delegating these routine queries to an AI agent, URW frees up its data engineers and analysts for higher-impact work: modeling, strategic projects, governance, all while accelerating the decision cycles of business teams. As at Dana Avocats, in-house expertise is focused where it creates the most value.
FAQ
How much does an AI workshop with Qolaig cost?
The AI workshop costs €5,000 excl. VAT for one day on site. 50% of this amount is deducted from development fees if the project moves forward.
How long does it take to deploy a solution?
The first use cases are usually live 2 to 6 weeks after the workshop.
Is our data kept confidential?
Yes. It is never used to train third-party models and stays in your environment, with hosting suited to your GDPR requirements.
What did Qolaig deploy for URW?
A Natural Language to SQL POC that turns a plain-language question into a query run on Snowflake.
Do users need to know SQL?
No. Users ask their question as they would ask a colleague and get a clear answer in plain language.
How does the AI understand internal vocabulary?
Through a custom business glossary that defines the group’s indicators, synonyms and calculation rules.

%20(1).webp)

















.webp)



%20(1)%20(1).webp)

.webp)





