Query your ERP data in plain English — get instant insights on sales, inventory, and operations without writing SQL or waiting for reports.
Client Overview
A mid-sized manufacturing company running SAP ERP across its procurement, inventory, production, and sales functions had accumulated years of high-value operational data that its management team could not effectively access. Generating any non-standard report required a formal request to the business intelligence team, which was already managing a backlog that stretched two to three weeks for complex queries. By the time managers received the data they needed to make a decision, the business context had often shifted and the insight was only partially useful. The executive team and department heads were frustrated by their dependence on a small group of BI specialists who acted as gatekeepers to the company's own data. Sales directors could not quickly check inventory availability during customer calls, procurement managers could not run ad-hoc supplier performance comparisons, and the operations team could not monitor production throughput against plan without waiting for a weekly report that was already outdated by the time it arrived. Spreadsheet exports were proliferating as a workaround, creating version control problems and inconsistent figures circulating across departments. Leadership identified data accessibility as one of the most significant operational bottlenecks in the business. They wanted every department head to be able to ask a direct question about the company's data and get an accurate, instant answer without writing SQL, understanding database schemas, or waiting for the BI team. The solution needed to be secure, connecting only to authorised data sources with appropriate role-based access controls, and accurate enough to be trusted for operational decisions.
Our Approach
Zentric Solutions built a natural language analytics interface on top of the company's SAP ERP data using OpenAI GPT-4 and LangChain. The system translated plain English questions typed by business users into precise SQL queries that were executed against a read-only mirror of the company's SQL Server database, ensuring that the underlying ERP system was never exposed to direct query risk. The text-to-SQL layer was trained on the company's specific data schema, table relationships, and business terminology so that queries phrased in operational language — such as 'what was our on-time delivery rate for supplier X last quarter' — were accurately translated without requiring users to know field names or table structures. Role-based access controls were enforced at the query layer, ensuring that each user could only retrieve data within their authorised scope. A sales manager could query sales and customer data but not payroll or supplier financials. The system also maintained a query audit log, recording every question asked and the SQL generated, providing full traceability for compliance and governance purposes. Results were returned in a conversational format with auto-generated summary commentary, and the interface could produce simple charts and tables alongside text responses for questions involving comparisons or trends. The BI team's request queue dropped by over 70% within the first two months as routine queries were handled directly by the business users who needed the information. Department heads reported being able to answer questions in seconds during meetings that previously would have required a follow-up report days later, fundamentally changing the pace and quality of operational decision-making.
Tech Stack
Project Tags
Everything you need to know about this project and our approach.
Zentric Solutions delivers cutting-edge digital products that streamline operations, enhance engagement, and drive lasting growth.
