Conversational execution is consolidated on Amazon Bedrock AgentCore, a managed runtime for AI agents.
Case study, Legal
Reliable, efficient GenAI for lawyers and legal departments
CoNDENSE.tech has worked with Lidia since the start, from UX and UI design to AWS, DevOps and GenAI. Today, as an Anthropic partner, it makes the platform's conversational AI, legal research and document analysis more reliable and more efficient.
- 3.2×
- monthly AI answers in 6 months
- 2k+
- questions on peak working days
- −51%
- response time on the slowest 1 in 10 answers
- −69%
- conversion time per document
Lidia production data, January–September 2026; document conversion from July 2026.
The challenge
Many models, one platform, already in production
Lidia brings conversational AI, legal research, web research and document analysis into a single platform for lawyers and legal departments, powered by several models, each dedicated to a specific area. Anthropic models power chat, legal research, web research and document-based answers.
Every language model is accessed through Amazon Bedrock, keeping data resident in the EU.
The collaboration began with UX and UI design. With the platform in production, the work now is to make it more reliable and more efficient, while keeping AI consumption measurable and under control.
At a glance
- Customer
-
Lidia AI platform for lawyers and legal departments
- Industry
- Legal, AI for lawyers and legal departments
- Use case
- Conversational AI, legal research and document analysis
- Scope
- Solution development: UX/UI, AWS, DevOps and GenAI
- Engagement
- 2025 – ongoing
- Anthropic models
- On Amazon Bedrock, EU data residency
The solution
Reliability and efficiency across the whole GenAI chain
The work spans agent execution, user experience, context, documents and usage accounting, all towards the same goal: a GenAI platform with greater continuity of use and more controllable consumption.
-
Agent runtime
-
Recoverable answers
Long answers show their progress as they are built, and users can reconnect without losing the response.
-
Context reuse
Prompt caching reuses conversation and document context instead of processing the same material again.
-
Document ingestion
Retries, fallbacks and better table handling make the documents behind every answer more reliable to extract.
More accurate usage accounting and organization-level budgets keep AI consumption measurable and under control.
Partners on this project
Anthropic
Our Anthropic-certified engineers build Claude agents and workflows grounded in your data and run them in production, on the Claude API or on Amazon Bedrock in the EU.
AWS
Engineers with the AWS Forward Deployed Engineer Validated credential migrate, modernize and run platforms on AWS, as code, on containers or serverless.
More stories
-
Erga Tourism
4 / 5 customer replies written by a production Claude agent
A blocked WhatsApp inbox, turned into a 24/7 AI sales channel
-
Garipalli
~200 City Escape games completed a month
From a chatbot builder to a game platform of its own
-
Attorney Share
400 ms p95 response time, down from 1.5 s
From hours-long releases to five-minute deploys
Contact
Building something similar?
CoNDENSE.tech designs and ships production AI agents wired to real business systems, not demos.
Talk to CoNDENSE.techOr write to us at
Claude Select Services Partner, AWS Advanced Tier Services Partner, Directus Partner