Executive Summary
For one of the world’s largest restaurant chains, getting insights into how customers behave across regions, stores, and channels is critical. Business teams across the globe have a constant stream of questions – but only a small group of experts knows how to tease out answers – from the many data warehouses and operational systems spread across the globe.
By layering PromptQL on top of their data, the company is transforming how the business engages with the data, gathers insights and makes decisions.
Business users can now ask real questions in plain English, like “Why are delivery sales down in Munich?” In response, PromptQL automatically retrieves accurate answers with full reasoning and explainability.
PromptQL provides a conversational analytics interface that truly works. Not only does it understands their data but also the unique context of each business – empowering teams to act faster and drive revenue growth without overhauling the existing data stack.
The Challenge: Bottlenecked Access to Data and Insights
The company’s data warehouses and operational data are rich with signals – but only a small number of analysts and data experts understand the data and have the SQL expertise to glean insights from fragmented datasets.
“We’ve had Power BI and Looker for a long time. But business teams still can’t ask the questions they want.”
Even users with data access faced friction. Getting answers meant multiple steps: tracking down tables, navigating joins, writing filters – often taking hours or days.
Even a simple question would sometimes have lead to a complex ETL pipeline. This created delays, limited experimentation, and prevented teams from asking follow-up questions. AI projects were also blocked, because the data couldn’t be accessed or queried flexibly.
The Solution: PromptQL
To unlock this, the company turned to PromptQL after evaluating many alternatives, including trying to build something in house.
“First, we tried to build it. Then we evaluated 100s of vendors. Finally, we chose PromptQL”
PromptQL connected multiple sources (such as data warehouses, databases, and other systems scattered) into a unified semantic graph, which is the foundation of its accuracy.
Now instead of writing SQL, business and product users can ask questions like:
- “What are the top 10 selling items in drive-thru orders across the Midwest?”
- “Which store locations have seen delivery drop the most in the last 7 days?”
- “Which combo meals are trending with loyalty members?”
PromptQL interprets the question, builds a multi-step plan, joins relevant data, applies filters and aggregations, and delivers answers - with full transparency that even business users can understand.
“The coolest part was it showed its work. You could see the plan, the logic. It wasn’t a black box.”
When something goes wrong - like a missing value or field - PromptQL retries with a fallback plan. For example, one expert commented how it caught a null value in one of our filters, retried with a different filter, and then answered the question.
PromptQL Innovations Behind the Accuracy
It’s important to note that PromptQL doesn’t just work like a text to SQL generator for this customer (or others). It introduces three major innovations that are changing the game for obtaining reliable insights:
- Understands Your Business Context
PromptQL captures your organization’s unique terminology, data structures, and logic, encoding them into a planning language that LLMs can reason over. This allows the AI to answer complex questions as if it were an expert analyst deeply familiar with your business and data landscape. - Self-Correcting and Resilient
When a query encounters issues—like missing fields, null values, or schema mismatches—PromptQL dynamically adjusts its plan and retries. This built-in resilience ensures consistent, accurate answers without manual troubleshooting or delays. - Transparent and Explainable
Every result is accompanied by a clear, plain-English explanation of how the answer was derived. Users can easily understand the logic behind each insight—without needing to inspect SQL—building trust and enabling fast, informed action.
“What impressed us was the explainability. You could see each action it took in plain speak."
The Impact: Unlocking Revenue Insights at Global Scale
With PromptQL, business stakeholders aren’t just reading dashboards – they’re engaging with data as a conversation.
Marketing and revenue teams explore trends, identify what’s working, and adjust campaigns or offers to grow revenue - without waiting on analysts.
- Data scientists save hours of manual querying and prototyping.
- Field operators test hypotheses in real time.
What used to take three days and four meetings now happens in minutes – with full transparency and traceability.
What’s Next: From Queries to Agents
The company is expanding PromptQL access to more teams and use cases. They’re integrating operational actions – like adjusting promos and surfacing menu insights – directly into the AI loop.
They’re also building reusable PromptQL programs to power repeatable workflows like:
- Weekly store performance rollups
- Campaign impact summaries
- Forecasting delivery trends across regions
“We started with queries. Now we’re thinking about agents – things that watch, reason, and act.”
This is expected to further shorten the time between revenue insights and action – letting teams measure, test, and optimize revenue-impacting decisions faster than before. With PromptQL, the company is turning questions into answers – and answers into action.
