How Our Claude Developers Built a Legal Spend Analytics Platform for Proxure
TL;DR: Claude developers for legal tech built Proxure a benchmarking and data analytics platform, engineered end-to-end with Claude Code, converting natural-language questions into SQL and running benchmarking across legal spend rates, leverage, timekeepers, and fees. Result: 40-60% faster insights, 50-70% reduction in manual dashboards and SQL, and 2x-3x higher automation adoption.
Proxure is a legal spend management platform based in Calgary, Alberta, helping legal and procurement teams manage and benchmark legal spend.
Its teams relied on static dashboards and manual querying to analyse spend data, with workflows that required predefined filters or SQL expertise. Benchmarking was manual and time-consuming, and the system had no contextual continuity, so users couldn't refine a query or build on a previous insight.
We partnered with Proxure to replace static dashboards with a conversational, benchmarking-native analytics platform.
The Problem
Legal and procurement teams could see their spend data. They couldn't easily ask it questions.
- Static dashboards and manual querying methods were the default way to analyse legal spend data.
- Workflows required predefined filters or SQL expertise, limiting who could actually use the data.
- Users couldn't access insights quickly, increasing dependency on analysts and slowing decision-making.
- Lack of flexible querying prevented users from exploring data dynamically or automating repetitive analysis.
- Benchmarking was manual and time-consuming, reducing accuracy.
- The system lacked contextual continuity, so users couldn't refine a query or build on a previous insight.
Proxure needed a platform that turned plain-language questions into real answers, with benchmarking built in, not bolted on.
The Solution
Our Claude developers built a unified platform, engineered end-to-end with Claude Code, using GoML's AI Data Analytics Accelerator.
Conversational and Structured Analytics
A Benchmark API handles structured analytics, and a Chat Agent API handles conversational queries, both backed by real-time data access and CSV export.
Natural Language to SQL
The platform converts natural-language queries directly into SQL, with session continuity so users can ask a follow-up question without restarting the analysis.
Benchmarking Engine
The benchmarking engine covers overview, rates, leverage, timekeepers, and fees, with parallel computations for faster results.
The Architecture
Infrastructure
- PostgreSQL RDS for real-time data access
- AWS Lambda and FastAPI for the API layer
- Databricks SQL tables for data processing
AI and Security
- Amazon Bedrock for natural-language-to-SQL conversion
- Azure API for data enrichment and categorisation
- SHA-256 anonymisation for sensitive data
- Amazon S3 and Amazon CloudWatch for storage and monitoring
Quality Assurance
- Input validation, error handling, and secure authentication
- SQL validation and end-to-end testing
- API-driven architecture with session tracking and system monitoring
The Impact
- 40-60% faster insights
- 50-70% reduction in manual dashboards and SQL
- 30-45% better decision-making
- 2x-3x higher automation adoption
- Supports 3x-5x growth
Key Takeaways for Legal Operations Teams
Pitfalls to Avoid
- Gating analytics behind SQL expertise or predefined filters
- Treating benchmarking as a manual, one-off exercise
- Building a query interface with no session continuity
What to Prioritise
- Let natural language carry the query, not just the search
- Build benchmarking into the platform as a first-class feature
- Design for follow-up questions, not single-shot queries
Build the Same Platform on Claude
If your legal spend data is sitting behind static dashboards, our Claude developers can scope a similar benchmarking and analytics platform for you.