Home › Blog › ILMS: 168 workflow stages, live co-drafting and a maker-checker rule, for a university’s legal office

ILMS: 168 workflow stages, live co-drafting and a maker-checker rule, for a university’s legal office

Engineering · October 18, 2026 · 2 min read

Notices, cases, contracts, MOUs, opinions and RTI requests, each with deadlines, hearings and documents. Build notes on a legal management system where drafts are written together in real time and nobody approves their own work.

A university group's legal department deals with everything from a tenant's notice to a multi-crore contract, across several institutions. Before ILMS, that lived in folders, email threads and memory. The brief was to give every piece of legal work a home, a process, and a record.

Every matter gets a 360° page

A notice, case, contract, MOU, opinion or RTI request becomes a matter with one page that shows everything: the parties, the deadlines, the hearings, the documents, a risk score and a timeline of what happened when. The timeline is automatic; the system writes it from the actions people take.

168 workflow stages

Legal work is a sequence of steps that differ by type. A contract review is not a litigation, and an RTI reply has its own clock. ILMS encodes 168 stages across notices, litigation, contracts, MOUs, RTI and nine practice areas, each with an owner, a checklist and an escalation rule when a deadline nears. The point is not bureaucracy; it is that the next step is always visible and someone is always responsible for it.

Drafting together, with track changes

Drafts are written in a live workspace: several people in the same document at once, with comments, suggested edits, track changes and versions. This runs on Yjs, which merges concurrent edits without conflicts, over a NestJS backend. An AI drafting assistant in English and Hindi can propose text, and its output is always labelled as AI and needs a human to accept it before it becomes part of the draft.

Nobody approves their own work

A maker-checker rule is enforced by the system: the person who drafts cannot be the person who approves. Nine roles with a permission matrix decide who can see and do what, and a full audit trail records it.

Documents

Uploaded files are OCR'd and full-text searchable, and downloads are watermarked with who took them and when.

Stack

NestJS, PostgreSQL, Redis, Yjs, Docker on AWS. It is in production, used by the legal team day to day.

Lessons

Workflow software lives or dies on the checklists. Spend the time with the people who do the work to get the stages right, and the software mostly builds itself around them. And label AI output loudly; people trust tools that are honest about what wrote what. If your organisation runs on approvals and deadlines, a management system like this is scoped on a call first, then quoted.

Tags: NestJS, Yjs, legal tech, workflow, PostgreSQL

More notes

Contact

[email protected] · GitHub · LinkedIn · X