Automotive · Production planning
Planning data from SAP BW, asked for in plain language
Product planners at a leading luxury car manufacturer spent hours querying SAP BW and several data warehouses to get the numbers they needed. We built a conversational assistant that retrieves planning data on demand, in German or English, by text or voice, connected securely to their on-premise SAP systems.
Key results
Client: Leading luxury car manufacturer (under NDA)
One of the world’s largest premium automotive companies. The manufacturer is under NDA, so it is not named here.
- Industry
- Automotive
- Use case
- Planning data retrieval and querying
- AI approach
- NLP and NLU, conversational AI
- Systems
- SAP BW and internal data warehouses
- Infrastructure
- Microsoft Azure with on-premise hybrid
- Engagement
- Design thinking into agile delivery
In short
The data existed. Getting to it was the job.
- Car product planning is one of the most data-intensive processes in manufacturing: production volumes, configuration options, market-specific variants, scheduling constraints.
- The problem was never that the data was missing. It was that reaching it required specialist knowledge of SAP BW and of which warehouse held what.
- So every planning question became a research exercise, or a request to a colleague who knew the systems better.
- A planner now asks in plain German or English, typed or spoken, and the assistant works out which systems to query and how to combine them. The data never leaves the company.
The starting point
The challenge
Planning data was scattered across SAP BW and several warehouses. Every query needed specialist system knowledge, which turned routine data retrieval into a slow manual process consuming hours of planner time every day.
Car product planning is one of the most data-intensive processes in automotive manufacturing. Planners pulled data from SAP BW and several other internal warehouses to make decisions: production volumes, configuration options, market-specific variants, scheduling constraints.
The problem was not that the data did not exist. It was that getting to it required specialist knowledge of the underlying systems: which warehouse held what, how to navigate SAP BW’s query structures, how to cross-reference across sources.
In practice every planning question turned into a manual research exercise, or a request to someone else who knew the systems better.
The build
What we built
We began with design thinking workshops to understand how planners actually work: what they ask, what data they need, where the manual bottlenecks sit. That shaped the solution directly.
Ask the question, skip the transaction
An assistant planners query in natural language, typed or spoken, in German or English. Instead of navigating SAP BW transactions or writing a query, a planner asks the question and gets the answer.
One question, several systems
Behind the conversation the assistant connects to SAP BW and the company’s other planning warehouses. It knows which system holds which information and how to combine data across sources, which is exactly the complexity planners used to carry themselves.
The data stays where it is
Planning data had to stay on-premise, non-negotiable for a manufacturer of this scale. The assistant runs in Microsoft Azure with a secure connection into the on-premise SAP infrastructure, and strict permission controls mean it only reaches data the specific user is authorised to see.
German and English, spoken and written
NLP and NLU handle both languages fluently in both forms, so planners across locations and teams work in whichever is natural to them.
Built on Azure services, delivered in sprints
The solution uses cloud-native Azure services, LUIS for language understanding, Bot Service for the conversational layer, CosmosDB for session data and AppService for deployment, delivered in agile sprints with the client’s IT department involved throughout.
What changed
The results
Before
Planners queried SAP BW and several warehouses by hand. Each request needed specialist knowledge or IT support, and a routine question took hours to answer.
After
Planners ask in plain language and get an answer in seconds. No system expertise needed, and the retrieval happens across every connected source at once.
The manual work of navigating data systems is gone for routine queries. Planners who had spent significant time pulling and cross-referencing data now spend it making planning decisions.
The hybrid architecture showed that enterprise AI does not require moving sensitive data to the cloud. The company kept full control of its planning data and still gave its teams a modern way to reach it.
It also showed a pattern we meet often: the biggest efficiency gains come not from replacing people but from removing the manual overhead that stops skilled professionals doing their actual work.
Questions about this project
Planning data retrieval and querying. Planning data was scattered across SAP BW and several warehouses. Every query needed specialist system knowledge, which turned routine data retrieval into a slow manual process consuming hours of planner time every day.
Time to retrieve planning data for one question: Hours → sec. German and English, in writing and by voice: 2 langs. Secure hybrid connection, data stays in their infrastructure: On-prem. Planners query complex data without IT support: Self-serve.
NLP and NLU, conversational AI. Technology used: Natural Language Processing, Natural Language Understanding, Microsoft Azure, SAP BW integration, Hybrid cloud, LUIS, Bot Service, CosmosDB, Voice interface.
Technology used
Talk to the people who built this
theBlue.ai comes back within one business day.
