LEGAL AI PLATFORM

From inbox chaos to court-ready packs.

Tess is a case-file compiler: it turns exports + attachments into auditable outputs with stable citation coordinates.

Tess compiles raw matter material into a review workspace and production artifacts — so the lawyer spends time deciding, not formatting.

Runs locally on a Mac Studio. No data leaves the building. Outlook/M365-native.

HOW IT WORKS

It's just an email.

Email your question

Plain English. Attach files if needed.

System routes & retrieves

Picks specialist, searches all three databases.

Get reply with answer

With citations you can verify.

Outputs
FLAGSHIP OUTPUTS

Flagship outputs

Concrete deliverables with stable citation coordinates — ready to file, forward, or print.

Evidence Bundle (PDF)

A single consolidated PDF with everything needed to cite, print, or file.

  • Self-describing cover (date range, total docs, generation timestamp)
  • Table of Contents as document register (ID / Date / From / To / Subject)
  • Stable document IDs (e.g., LCC-0001 … LCC-0007) in register and on pages
  • Single pagination scheme ("Page X of N") — citations become Doc ID + Page
  • Normalized email rendering: consistent header (From/To/Cc/Date/Subject/Attachments) + body
  • Threads and forwards preserved in-body (older correspondence visible even outside declared date range)
  • Attachments recorded by filename (embedded as separate exhibits is optional)
Example: 7 documents → 43-page bundle with TOC + stable IDs

Collation + indexing only — no argument drafting in the bundle.

Bates stamping and exhibit expansion are optional configurations.

Email Review Report (.docx)

Structured analysis of an email corpus with triage, entity normalization, and issue buckets.

Report structure:

  • Cover metadata: generated date, totals
  • Participant Directory (email → name / role / message count)
  • Email Summary Table
  • Attachments Table
  • Issue-specific analyses (e.g., Privilege Waiver, Negligence)
Example: 730 emails → 516 relevant (70%), grouped into review buckets with counts

Example issue buckets (one matter):

Categories are determined by the system from the brief and case background — factual classification, not legal opinion.

Privilege Waiver Risk 23
Negligence Indicators 47
Contractual Breach 31
Regulatory Compliance 18
Key Communications 89

Tess flags risk; privilege calls stay with counsel.

IN PRACTICE

When the client is in the room

"During a live meeting, counsel asked for every communication between two counterparties."

We printed the full set in ~10 minutes.

A detailed communications report followed within hours.

Case Study • Enterprise Legal Team
View Case Studies
REAL OUTPUT

What you actually get back

Real system output from active litigation matters. All processed locally on a Mac Studio — no data leaves the firm. Names and identifying details changed.

CONTEXT

A senior solicitor emailed a two-paragraph matter summary and pointed to a folder of ~90 source documents — emails, contracts, lease agreements, court filings, agent correspondence. No templates, no structured prompts.

THE INPUT
THE OUTPUT
Participant Directory 30+ total entries 
Side        | Name             | Role                | Organisation
────────────┼──────────────────┼─────────────────────┼──────────────────────
Defendant   | [Landlord]       | Property owner      | Harbour View Pty Ltd
Plaintiff   | [Tenant]         | Lessee / operator   | Metro Waste Services
Third Party | [Agent 1]        | Selling agent       | LJ Hooker Commercial
Third Party | [Agent 2]        | Managing agent      | Ray White Commercial
Neutral     | [Conveyancer]    | Purchaser solicitor | Preston Chambers
Chronology of Events 60+ total entries 
ID   | Date       | Event                                                | Sources
─────┼────────────┼──────────────────────────────────────────────────────┼────────
E001 | 2022-12-01 | Lease executed between [Landlord] and [Tenant]       | [D003]
E012 | 2023-03-15 | [Agent 1] lists property without notifying [Tenant]  | [D017]
E027 | 2023-06-22 | [Tenant] exercises right of first refusal            | [D034]
E031 | 2023-07-10 | [Landlord] rejects ROFR, proceeds with third-party   | [D041]
E048 | 2023-11-03 | ⚠ Termination notice served — date unverified        | [D067]
Source Document Index 90+ documents 
ID   | Date       | Filename                              | Type
─────┼────────────┼───────────────────────────────────────┼──────────
D003 | 2022-12-01 | Lease_Agreement_HarbourView.pdf       | Document
D017 | 2023-03-15 | Email_Agent1_to_Landlord.msg          | Email
D034 | 2023-06-22 | ROFR_Exercise_Notice.pdf              | Document
~5 min
processing time
60+
events extracted
30+
participants
8
flagged for review

2–3 days of paralegal work, done in ~5 minutes on a local Mac Studio. Every event traced to source.

FOUR SPECIALISTS

Pick the output. Tess routes the pipeline.

Each specialist is optimized for a specific output type. The system routes your request automatically based on what you're asking for.

Detective Example

~15 seconds
YOUR EMAIL
AI RESPONSE
DETECTIVE_RESPONSE
Robert James Hadley
Middle name: James
Date of birth: October 15, 1949
Source: Source: Meridian Holdings Directors Register (2019), Page 3, Section 2.1

You don't prompt-engineer. You pick the output; Tess routes the pipeline.

Trust
AUDIT TRAIL

Built for defensibility

Defensibility

  • Scope is explicit: date range + what's included + generation timestamp
  • Citation coordinates are stable: Doc ID + Page
  • Primary facts vs. secondary context are separated ("registry extract says X; email corpus says Y")
UNDER THE HOOD

Tri-Database Retrieval Architecture

Most legal AI tools use simple keyword or vector search. We built a three-database system that mirrors how lawyers actually think — tracking entities, citations, and semantic meaning simultaneously.

Lexical Database

Exact names, dates, parties. Elasticsearch (lexical/BM25).

Vector Database

Semantic similarity. Legal-specific embeddings.

Graph Database

Citations, relationships, document versions. Neo4j.

ORCHESTRATOR
Cited answers
Research reports
Classification outputs
Evidence bundles
BOUNDARIES

What Tess does / doesn't do

AI surfaces information. Lawyers form judgment.

What Tess does

  • Collate, index, summarize, extract, cite
  • Draft internal comms and reports
  • Surface information with verifiable sources

What Tess doesn't do

  • × Replace legal judgment
  • × Guarantee correctness
  • × Give legal advice
DEPLOYMENT OPTIONS

Choose your setup.

eDiscovery Service

One-time document review and classification. You send us the files, we send back the analysis.

  • PST/email ingestion
  • Relevance + privilege classification
  • Chronology & timeline generation
  • Evidence bundle production
  • Per-matter pricing
Get a Quote

Integrated Platform

Full deployment on your infrastructure. All your emails, files, and matter data — searchable and AI-ready.

  • Local/private deployment
  • All four specialists
  • Email, SharePoint, DMS integration
  • Tri-database architecture
  • Custom workflows & training
Schedule Demo

Ready to see it in action?

Book a demo with real examples from your practice area.

Book a Demo

Factual analysis, not legal opinion. AI surfaces the information. You form the judgment. Names and identifying details have been changed. Scenarios reflect real system output.