Skip to content
Shipped

Business intelligence · AI · Automation

Dashboards, a RAG assistant and automated data entry for a production company

A garment production company ran reporting from manual exports, kept its knowledge in documents nobody could search, and typed records in by hand. I built three connected pieces: dashboards on one trusted model, a RAG assistant over their documents, and automated data entry with validation — so the numbers refresh themselves, answers come with sources, and bad records are caught at the door.

Role
Freelance BI and AI consultant
Client
A garment production company
When
2025
Stack
Power BI · DAX · Power Query · Python · RAG
System

Dashboards

  1. Sales · creditpurchase exports
  2. Power Queryclean · shape
  3. Data modelstar schema · DAX
  4. Dashboardssales · credit · purchase
  5. Drill-downregion → invoice

Assistant and data entry

  1. Documentsreports · files
  2. Retrievalsearch · sources
  3. Cited answers
  4. Data entryvalidated on entry

01

The problem

Sales, credit and purchase data sat in separate exports that had to be pulled and combined by hand, so every report started late and was hard to trust.

Answers lived inside documents and reports; finding one meant opening files until something matched.

Records were typed in manually, and typos or missing fields only surfaced later, in a report.

02

What I built

  • Power BI dashboards for sales, credit and purchases, over one data model so every view answers from the same numbers.
  • Automated refresh — no more manual exports — and drill-down from a region to a single invoice.
  • A RAG assistant: ask a question in plain language and get an answer drawn from the company's own documents, with the source shown.
  • Automated data entry: a structured template with field mapping, and validation of required fields, formats and duplicates before anything is saved.

03

How it works

  1. Step 01

    Join

    Sales, credit and purchase records are cleaned with Power Query and brought into one star-schema model, with DAX measures defining each metric once.

  2. Step 02

    Show

    Dashboards filter from a region down to the invoice and refresh on a schedule, so nobody exports a file.

  3. Step 03

    Ask

    Documents are indexed for retrieval; the assistant finds the relevant passages and answers from them, citing where each came from.

  4. Step 04

    Enter

    New records go through a structured entry flow that validates them first; anything that fails goes to a short review list instead of through silently.

04

Outcome

faster reporting
0%faster reporting
  • Reporting runs off one model instead of a pile of exports — 40% faster.
  • Any number on a dashboard can be followed down to the invoice that produced it.
  • Answers are grounded in the company's own documents and show their source, so they can be checked in seconds.
  • Errors are caught at entry, not discovered in a report weeks later.

Want this kind of work on your data?

Get in touch