Skip to content
Nupgrade

Vertical SaaS with an AI case assistant

Legal case management with an AI assistant

A shared workspace for a law firm's matters, documents, hearings and deadlines, with an assistant that organises facts and prepares referenced drafts for a lawyer to review.

A legal team needs its matters, documents, deadlines and reasoning to stay connected. A missed hearing or a document filed in the wrong place is a professional-liability problem, not an inconvenience.

The platform brings those records into one case workspace. AI helps organise information and prepare drafts, and every draft stays subject to legal review.

The problem

A missed hearing or deadline is a professional-liability event. Firms need one reliable calendar, a record of who did what and a way to build a case from many documents without losing where each fact came from.

What the system does

  • Matters with role-scoped access, document versions, a timeline and hearing and deadline tracking.
  • AI assistance for case building: organising facts and documents, spotting gaps and drafting referenced arguments for a lawyer.
  • A background worker that monitors calendar entries and raises tasks before a deadline.
  • An audit history of who changed what, and backups with a tested restore.

A representative journey

  1. Step 1: Open a matterCreate the matter, assign the team and set who may see it.
  2. Step 2: Organise the recordApproved documents and facts are filed. The assistant proposes a chronology and flags missing information.
  3. Step 3: Check the gapsThe lawyer sees what the timeline lacks: a missing date, an unsigned page, a contradiction between two documents.
  4. Step 4: Draft with referencesThe assistant prepares a draft in which each statement points to a document and page.
  5. Step 5: Lawyer reviewsThe lawyer edits, accepts or rejects. Nothing the assistant wrote is final until a person signs it off.
  6. Step 6: Track the next stepThe result becomes a task or a deadline on the shared calendar, with an owner.

Matter timeline · synthetic

1 gap
  1. 2026-01-12Contract signedDoc 1, p. 9
  2. 2026-03-04Delivery missedDoc 4, p. 2
  3. 2026-03-20Notice sentDoc 6, p. 1
  4. UnknownDate of reply not foundGap · ask the client

Draft summary

Needs lawyer review

The supplier did not deliver by the agreed date [Doc 4, p. 2]. Written notice followed within 16 days [Doc 6, p. 1].

The date of the client’s reply is not in the approved record. This is listed as a gap, not assumed.

Hearing: 2026-12-02Task: confirm reply date · owner: A. (fictional)
Concept view: matter timeline with referenced summary. Illustrative, with authored synthetic data. It is not a screenshot of a real system, and no real person, client, record or result appears in it.

Working prototypeUse a small working version of this on invented data, the way a person would.

Try the prototype

Inspect the system

This anatomy is a proposed description of how such a system is put together. It is not an audit of a specific deployment.

Inspect the system

How Legal case management with an AI assistant is put together

What it is
Matter workspaceThe web application where a team works on a matter.
What flows through it
People, matters and permissions go in. A shared view of the case comes out.
The decision
Access is scoped by role and by matter.
Read the whole system as a list
  1. Matter workspace. The web application where a team works on a matter. Flow: People, matters and permissions go in. A shared view of the case comes out. Decision: Access is scoped by role and by matter.
  2. Matter and document services. Stores matters, documents, versions and history. Flow: Uploads and edits go in. Versioned records and an audit trail come out. Decision: Every change is recorded with who made it, so the history can be reconstructed.
  3. Extraction and retrieval. Reads approved documents and finds relevant passages. Flow: Documents go in. Searchable passages with their source and page come out. Decision: Every passage keeps its document and page so a statement can be traced.
  4. Case assistant. Proposes chronologies, flags gaps and prepares referenced drafts. Flow: A matter's approved record and a question go in. A draft with references, or a list of gaps, comes out. Decision: The assistant prepares and a lawyer decides.
  5. Calendar and task worker. Tracks hearings and deadlines and raises reminders. Flow: Dates and owners go in. Tasks and reminders come out ahead of each deadline. Decision: A deadline has an owner and a reminder path, not just a date.

Key decisions

  • References over fluency

    A legal draft that cannot show its source is not usable. Each statement carries a link to the document it rests on.

  • The lawyer is the author

    The assistant prepares and organises. It does not decide, file or send, and the interface makes the review step unavoidable.

  • Calendar as a first-class record

    Deadlines live in one place with an owner and a reminder path, instead of in individual diaries.

Outcomes and limits

We publish no measured outcomes for this project, because we have none we can evidence. The description above is of what the system does, not of results it has produced.

  • The system does not replace a lawyer, guarantee legal correctness or integrate with a particular court system.
  • No compliance certification is claimed.
  • The AI model, retrieval method and deployment are not specified here, and no accuracy figure is claimed.

Guardrail

A lawyer reviews everything the assistant drafts. Nothing is filed, sent or relied on until a person signs it off.

Technical detail

Technology named in the portfolio for this kind of system. Which parts a given engagement used is not stated.

  • C# / .NET
  • PostgreSQL
  • Next.js
  • Docker
  • Playwright

One question

Which manual process currently takes the most time from your team?

Answer in a sentence or two. An engineer replies with where we would start, what it takes and what we would not automate.