Company data warehouse queryable in natural language

The client

A marketing agency in the fashion and luxury events sector, with several company applications: ERP, HR system, spreadsheets and other operational tools. Data was scattered across separate silos, each with its own internal logic, with no unified source that systems and users could access securely and consistently.

The context

The company ERP organised data according to its own internal structure: information was spread across many tables, making it hard to consult for anyone who didn’t know the system’s logic. HR data (attendance, contracts, employees) lived in a separate application, unconnected to the ERP. Some information existed only in Excel files on Google Drive. The result was a fragmented picture, where accessing cross-referenced data required technical skills or long manual processes.

The objective

Create a centralised, normalised and accessible source of data - both for company systems (via secure APIs) and for internal users, with no need for SQL or technical skills.

• Data spread across ERP, HR system and Excel files: no unified source.
• Complex ERP data structure: information on a single product spread across many tables.
• No way to cross-reference HR and ERP data without technical work.
• Data access reserved to those who knew SQL or the applications’ internal logic.

The solution

We built a single digital archive that gathers all company data - ERP, HR and Excel files - and organises it into a format that’s easy to consult. The ERP data, previously scattered across many hard-to-read tables, was reorganised into a clearer structure. A system imports this data automatically and makes it securely available to other company applications too. On top of it all, an AI lets anyone in the company query the data by writing in plain language - with no technical skills - always respecting the permissions assigned to each person.

• Centralised data warehouse: data from Oracle NetSuite, JetHR, Google Drive and other applications in a single normalised source.
• ERP data normalisation: from a multi-table structure to a simplified, consultable model.
• Middleware to import data and securely expose it to external systems via API.
• MCP server connected to the DWH: query data in natural language through an LLM.
• Permission-based access control: each user sees only the data they’re authorised for.

The results

A single source of data

All company data - ERP, HR, operational - in one place, normalised and always up to date.

Natural-language access

Users query the data warehouse through AI, with no SQL skills or knowledge of the technical structure.

Secure interoperability

Systems access data through controlled APIs - no direct access to the ERP.

Granular access control

Each user sees only the data they’re authorised for, with role-configurable permissions.

Cross-referenced data, effortlessly

ERP and HR finally connected: a full view of people, costs and activities in a single query.