Taxation RAG
Case Study Summary
Client: Mid-size tax advisory firm Industry: Tax & Professional Services
Impact Metrics:
- 93% reduction in research time (8 hours to 30 minutes)
- 5,000+ documents queryable with source citations
- 60% improvement in faithfulness and accuracy over standard RAG
- Up to 90% margins on fixed-fee and limited-scope projects
Mid-size tax advisory firm cut income tax research time by 93% — giving advisors instant, grounded access to rulings, legislation, and ATO guidance in one unified system.
Challenge
Advisors were spending hours manually searching across multiple databases to find relevant rulings, legislation, and ATO guidance. Research was slow, fragmented, and inconsistent with different advisors relying on different sources and often missing relevant material.
Our Approach
We built a RAG system that unified rulings, legislation, and ATO guidance into a single queryable knowledge base. The system uses hierarchical chunking, reciprocal rank fusion, and reranking to surface the most relevant results with high precision. Guardrails and evaluation frameworks ensure responses are grounded in source material with full citations. We designed a custom chunking strategy tailored to each document type, improving faithfulness and accuracy by 60% over standard RAG implementations.
Results & Impact
- Research that previously took 8 hours now takes 30 minutes
- Advisors can query across 5,000+ documents instantly, with source citations for every answer
- Reduced inconsistencies in research quality across the team — now used as the primary research tool
- Clients receive responses to ad-hoc questions within the hour, greatly improving service quality
- Margins on fixed-fee and limited-scope projects rose to up to 90% while improving quality of deliverables
Solution Architecture
Knowledge ingestion and retrieval mechanism with SoTA knowledge expansion techniques for retrieval breadth and depth
Tech Stack
- OpenAI embeddings, Supabase with pgvector
- Microsoft Azure cloud infrastructure, Modal
- Redis cache, Python backend services
- FastAPI for RESTful endpoints
- LangSmith for observability and evaluation
- Docker containerization, GitHub Actions for CI/CD pipeline
Team Expertise
Our team, specialised in retrieval-augmented generation and legal document processing, built a system that unifies fragmented tax research across rulings, legislation, and ATO guidance — delivering a grounded, citation-backed research tool that advisors trust as their primary source of truth.
-
Let's have a virtual coffee together!
Want to see if we're a match? Let's have a chat and find out. Schedule a free 30-minute strategy session to discuss your RAG use case and explore how we can work together.