The shape of a good workflow
General information, not legal advice. Check your own bar or regulator's guidance and your client engagement terms before putting any client information into an AI tool.
Every workflow below follows the same four steps: Input (what goes in, and in what form), AI step (one bounded job), Human check (who reviews and against what), and Record (where the result is stored). If you cannot name all four, the workflow is not ready.
Five workflows
| Workflow | Input | AI step | Human check | Record in |
|---|---|---|---|---|
| Intake triage | Enquiry text, anonymised | Categorise and list missing facts | Lawyer confirms category and conflict check | Matters database |
| Weekly review | Open matters and deadlines | Draft a digest of what is due or stale | Lawyer reads and corrects | Weekly review page |
| Meeting notes to actions | Your rough notes | Extract actions, owners, dates | Attendee confirms | Actions database |
| Document first-read | Non-confidential document | Summary and list of defined terms | Lawyer reads the source | Matter page |
| Client update draft | Status fields and notes | Draft a short plain-language update | Lawyer edits and sends | Matter page log |
Building one in Notion
A workflow needs somewhere to live. In Notion, use a database for the records and a page for the instructions and prompt. LawyerOS Pro has Matters, Actions and Clients databases and a Custom AI Agents page you can adapt; Starter has Matters and Deadlines. Neither performs the AI step by itself; that comes from Notion AI or another assistant you choose.
- Write the prompt and keep it on the workflow page.
- Add a Status property with values Draft, Reviewed, Approved.
- Add a Reviewer property so a person is always named.
- Run on dummy matters for two weeks and compare against doing it manually.
- Only then use it on live work, if your policy allows.
Where workflows fail
Skipping the human check because the first ten outputs were fine. Feeding in messy data and getting messy digests. Letting a tool send client messages directly. Forgetting that a deadline summary is not a calendaring system, so rules-based deadlines should be checked against the court rules and your own diary.
Frequently asked questions
- Which workflow should I start with?
- Meeting notes to actions is low-risk and quick to check. Start there with non-confidential meetings.
- Do I need coding skills?
- No. These can be run with a chat assistant or Notion features plus a well-organised database.
- Can AI calculate deadlines?
- Do not rely on it for that. Limitation periods and court deadlines depend on rules that must be checked by a person.
- How do I know the workflow works?
- Compare outputs with manual results on a sample and track the corrections your reviewer makes.
About the author
Raghav Handa trained as a lawyer, then spent years building legal technology, from enterprise legal products to custom systems for law firms.
Keep reading
- AIAI agents for law firms: what they are and where they fitWhat AI agents are, how they differ from chatbots, which law firm tasks suit them, and the supervision and confidentiality limits. Includes Notion Custom Agents.3 min read · 17 October 2026
- AIAI prompts for lawyers: 14 copy-and-paste prompts with placeholdersA library of 14 AI prompts for lawyers with [placeholders]: summaries, issue lists, email drafts, checklists and review prompts. Use them on non-confidential text.4 min read · 17 October 2026
- AINotion AI for lawyers: what it does and how to use it carefullyWhat Notion AI and Custom Agents can do for a law practice, example uses with matter databases, plan requirements to check, and confidentiality safeguards.3 min read · 17 October 2026
- Practice managementLegal case management in Notion: a complete setupSet up legal case management in Notion step by step: cases, contacts, deadlines, tasks, documents, time and notes, with a case page template and key views.3 min read · 17 October 2026