03
Document intelligence for audit teams
PwC · AI Engineer · 12 months
The challenge
Auditors work with large volumes of contracts, financial statements and regulatory documents. Finding a specific answer, or checking documents against each other, used to mean hours of manual reading.
The approach
A document intelligence platform that lets auditors query large document sets in natural language.
- 01
Question answering
Retrieval-augmented generation (RAG) that answers questions with references to the source passages.
- 02
Information extraction
Automatic capture of key data points from unstructured documents.
- 03
Document matching
Identifying related and conflicting content across document sets.
- 04
Integration
Production-ready AI components delivered through APIs into the wider application.
The outcome
In production use by several hundred auditors, the platform makes thousands of audit documents searchable and turns hours of manual reading into precise, source-referenced answers.
My role
AI engineer and subject-matter expert in a cross-functional team. I built the AI components and worked closely with frontend, backend and data science teams to bring them into production.