Legal AI CRM & Case Intel Platform

LEGAL AI CASE INTELLIGENCE PLATFORM
Technical Specification & Phase-Based Development Scope

Repost – Developers Must Quote Per Phase

1. Project Overview

We are building a solicitor-focused Legal AI Case Intelligence Platform, conceptually inspired by tools such as CourtReady.family, which was vibe-coded in 3 days.

However:

This system is NOT consumer-facing

It does NOT require step-by-step guided workflows

It is a CRM-style legal operations platform

It is designed for solicitors and legal teams only

This platform combines:

Case CRM

Evidence Vault

Timeline Builder

Legal Factor Mapping

Gap Detection

AI Query Engine (RAG-based)

The end goal is a structured, intelligent case preparation system — not a chatbot, not a document summariser, and not a client questionnaire tool.

CRM-style dashboard wireframes will be attached.

2. Mandatory Technology Stack

To avoid repeated clarification questions, the following stack is preferred and largely mandatory.

Preferred Stack

Frontend:

React + TypeScript (Next.js preferred)

Backend:

Node.js (NestJS or Express)
OR

Python (FastAPI)

Database:

PostgreSQL (Mandatory)

Vector Store:

pgvector within PostgreSQL

AI:

OpenAI API (must be modular so provider can be swapped later)

Storage:

S3-compatible object storage

Deployment:

Dockerised (Mandatory)

Non-Negotiable Requirements

Multi-tenant architecture (firm_id enforced at database level)

RAG-based AI Query Console (not simple ChatGPT wrapper)

Strict token control

Structured JSON outputs from AI

Caching layer to prevent repeated AI costs

Clean codebase with documentation

3. Architecture Clarifications (Previously Asked Questions)

To prevent confusion:

• Embeddings must be isolated per tenant at DB level (firm_id filtering).
• Physical vector DB separation is NOT required in Phase 1.
• Basic role structure required in Phase 1:

Firm Admin

Standard User
• OCR acceptable via Tesseract or similar open-source solution.
• WhatsApp (.txt) and email (.eml) parsing required.
• Hosting preference: VPS-based Docker deployment (Hetzner/DigitalOcean acceptable).
• Architecture must allow future dedicated deployment per firm (Phase 5).

4. AI Architecture Expectations (Important)

This is NOT a generic AI wrapper.

Phase 1 must include:

Proper RAG pipeline

Document chunking

Embedding once per document

Retrieval limits (top-k control)

Structured AI responses

Token limits

Caching of repeated queries

AI must NOT:

Re-process full documents unnecessarily

Send entire case files on every query

Produce long unstructured essays

We expect a cost-controlled architecture.

5. Dashboard Design Requirement

This platform must use a CRM-style dashboard layout, not a consumer-style guided wizard.

Attached wireframes illustrate the expected structure.

Dashboard must include:

KPI row:

Active Matters

Awaiting Documents

Upcoming Dates

Overdue Tasks

Matter list / pipeline view

Recent activity feed

Priority alert section

Global search

Matter workspace must include tabs:

Overview

Evidence Vault

Timeline

Factors

Gaps

Communication Scanner

AI Query Console

Exports

This is a professional legal operations UI.

6. Phased Development Structure
Developers MUST Quote Separately Per Phase

Each phase must be priced independently.

Phase 1 – Core Case Intelligence MVP

Scope:

Multi-tenant CRM foundation

Firm + user management

Evidence upload (PDF, DOCX, images with OCR, WhatsApp .txt, email)

Metadata extraction

Timeline builder

Irish legal factor mapping

Gap detection engine

Communication pattern scanner

RAG-based AI Query Console

Structured outputs

Docker deployment

Deliverable:
Fully working MVP deployed centrally.

Phase 2 – Evidence Integrity & Court Pack Layer

Scope:

Immutable version history

SHA-256 file hashing

Chain of custody logging

Exhibit designation

Court pack builder

Export-ready bundles

Deliverable:
Litigation-grade evidence system.

Phase 3 – Case Operations & Automation

Scope:

Task system

Status automation

Deadline alerts

Matter snapshot generator

Cross-case analytics

Deliverable:
Firm-level operational intelligence layer.

Phase 4 – Multi-Practice Expansion

Scope:

Employment law module

Personal injury module

Probate module

Custom factor mapping per module

Customised exports/templates

Deliverable:
Modular practice support.

Phase 5 – Enterprise & Security Layer

Scope:

Dedicated deployment option (VM per firm)

Role matrix expansion

MFA

Audit reports

Encryption at rest

Data retention controls

Export/import capability

Deliverable:
Enterprise-ready version.

Phase 6 – Client Portal & External Collaboration

Scope:

Client upload portal

Approval workflows

Secure share links

External counsel view

Activity tracking

Deliverable:
Full collaboration layer.

7. Important Instructions for Developers

Quote EACH phase separately.

Phase 1 will be awarded first.

Subsequent phases awarded based on code quality and performance.

Architecture must consider future phases to avoid refactoring.

Clean documentation required.