How a digital intern built a legal AI platform. 69 commits. 2 weeks. Production.

A UK law firm drowning in case data.

Hundreds of active cases. Thousands of documents. A CRM that stores everything but surfaces nothing. Solicitors spend hours searching case files for facts they know are there — buried in PDFs, scattered across correspondence, lost in compliance timelines.

Every case needs statutory deadline tracking. Every document needs to be searchable. Partners need portfolio-level visibility across the entire caseload — which cases are stalled, which are at risk, which handlers are overloaded.

Off-the-shelf tools gave them pieces. Nobody gave them the whole platform. A custom build was required.

Hundreds of cases. Thousands of documents. A CRM that stores everything but surfaces nothing. Solicitors spend hours searching instead of practising law.

We deployed an intern that built the whole thing.

X-Arc trained and deployed CCCG — a digital intern tasked with building CaseGuard, a full AI-powered legal case management platform.

Not a chatbot. Not a dashboard. Not a feature request. A complete SaaS application. The intern designed the database, chose the tech stack, built the AI agents, created the UI, integrated the existing CRM, and shipped to production.

69 commits. 2 weeks. From zero to a production platform with multi-tenant architecture, AI agents with 17 specialised tools, CRM integration, and 7 rounds of UI polish.

CCCG didn't follow a prompt. It built a product. Architecture, AI agents, document processing, CRM integration, UI — the intern shipped what would take a dev team months.

The human secured the vault. The intern built the bank.

Legal data is confidential. Client data cannot be mishandled. This wasn't a case for full autonomy — it demanded a deliberate human-intern partnership.

The human owned the security layer. The intern built everything else.

Human Every database table needs Row Level Security. Tenants must never see each other's data. Proclaim API credentials need to be encrypted before storage.
CCCG Understood. I'll build the full application within those constraints. Multi-tenant architecture with strict data isolation. I'll handle the schema design, API routes, AI agents, UI, and CRM integration.
Human Authentication needs to be invite-only. No public signup. Role-based access: Super Admin, Admin, Member. JWT session management.
CCCG Done. Invite-only auth flow with email invitations via Resend. Role-based access enforced at both the UI and API level. I've also built the Super Admin panel for onboarding new firms.

What the human owned

The security layer. Row Level Security policies on every table. Fernet encryption for API credentials. JWT session management. Multi-tenant data isolation design. Proclaim integration security.

What the intern built

Everything else. The full SaaS application — 69 commits from zero to production. AI agents with 17 specialised tools. Document processing pipeline. Case timeline system. Dashboard with live KPIs. Complete UI/UX across 10+ views. 7 rounds of polish based on testing feedback. Database migrations. Automated test suite.

Would you trust an AI to build software that handles confidential legal data? You don't have to trust blindly. You partner. The human sets the boundaries. The intern does the work.

What CCCG built. A full platform, not a feature.

The Case Agent — a solicitor's AI partner

A solicitor opens a specific case and has an AI conversation with full case context. The agent has read the entire case file: documents, timeline events, correspondence, compliance milestones. It cites sources for every claim — no hallucination, grounded in the firm's own data.

01

Document search

Searches across all case documents — PDFs, Office files — using vector search. Finds relevant passages and cites the source.

02

Timeline analysis

Every event, correspondence, and milestone tracked chronologically. Monthly summaries generated automatically.

03

Compliance checking

Legal milestone tracking with working-day calculations. Statutory deadlines monitored automatically.

04

Similar case reasoning

Finds similar cases in the firm's history. Surfaces precedents and patterns across the entire caseload.

05

AI recommendations

Suggests next steps based on case analysis. Every recommendation cites the evidence it's based on.

"Ask it anything about the case. It's read the whole file."

The Portfolio Agent — a partner's overview

A partner or funder can analyse trends across the entire caseload in a single question. Stalled case detection, risk assessment, settlement probability, handler workload — all from pre-validated queries. No direct text-to-SQL. Security by design.

The Platform

Multi-tenant architecture — multiple law firms on one platform, strict data isolation. Invite-only authentication — no public signup. Proclaim CRM integration — connects to the existing legal CRM, imports cases, syncs data. Document processing — ingests case documents, vectorises for AI search. Dashboard KPIs — total cases, active/closed split, documents processed, AI session usage. Super admin panel — platform-level management for onboarding new firms.

Hire Your Intern

Numbers. Not promises.

69 Commits to production
2 wks Zero to production
17 AI agent tools built
8+ Database tables designed
10+ Pages and views shipped
7 Rounds of UI polish

Hire Your Intern

This isn't a prototype. It's a production SaaS platform.

CaseGuard is live. Real solicitors use it daily. Real cases, real documents, real compliance deadlines — all managed through a platform built by a digital intern in 2 weeks.

[ Platform screenshot — CaseGuard legal case management dashboard ]

Screenshot pending — platform access being arranged

The intern didn't just write code. It made architecture decisions, chose the tools, designed the database schema, and shipped a production system.

Zero trust assumptions. Every layer secured.

Legal data demands the highest standard. CaseGuard was built with security at every layer — not bolted on after the fact, but designed in from the first commit.

Row Level Security

Every database table enforces tenant isolation at the Supabase level. Queries physically cannot return another firm's data.

Encrypted Credentials

Proclaim API credentials encrypted with Fernet before storage. Never stored in plaintext.

Invite-Only Access

No public signup. Firm admin invites team members. Role-based access: Super Admin, Admin, Member.

Source-Cited AI

Every AI response cites the document or data it's based on. No hallucination. Grounded in the firm's own data.

Tenant Data Isolation

Multi-tenant from day one. Separate Pinecone namespaces, database-level isolation. Firms cannot see each other's data.

Security Patching

CVE-2025-55182 patched within days of disclosure. The intern maintains the platform it built.

The human secured the vault. The intern built the bank. Every layer is deliberate. Every access is controlled.

The intern that built this still knows every line of code.

CCCG is still maintaining CaseGuard. Every bug fix, feature request, and iteration is handled by the same intern that built the platform from scratch.

Context compounds. A new developer needs weeks to understand a codebase they didn't write. CCCG wrote every line. It knows the architecture decisions, the edge cases, the trade-offs. It doesn't need onboarding.

Six months from now, the intern that built this will still know the codebase better than any human developer who joins the project. That's the compound advantage.

Every correction, every bug fix, every feature request makes the intern permanently better. The intern that built CaseGuard will know this codebase longer and deeper than any human hire.

Hire Your Intern

Ready?

A digital intern built a production SaaS platform in 2 weeks. What could yours build?

Your intern is ready. Trained on your business, building your product, shipping to production.