An Australian legal firm was losing thousands of hours a month to file notes. Every call - client updates, settlement discussions, initial consults - ended with a lawyer sitting down to type a structured summary. We built an AI-powered system that does it for them. Here's how it works and what it's worth.
The problem
Lawyers at this firm were spending 10 to 15 minutes after every call writing file notes. Not quick scribbles - structured documents with attendees, key discussion points, advice given, action items, and follow-ups. The kind of notes that matter in litigation and compliance.
Across the firm, staff were handling 40+ calls a day. That's over 400 minutes of typing daily - just on file notes. The notes needed to be consistent, thorough, and saved into the practice management system. But the quality varied depending on who wrote them and how rushed they were.
The real cost wasn't just the time. It was the opportunity cost. Every minute a lawyer spends typing a file note is a minute they're not doing billable work.
The solution
We built an AI-powered file note automation system using n8n for workflow orchestration and the Claude API for AI generation. The workflow is simple from the lawyer's perspective: finish a call, review a pre-written note 90 seconds later, approve it, and move on.
Behind the scenes, the system captures the call recording, transcribes it, and sends the transcript to Claude with firm-specific prompts. The AI extracts the structure the firm needs - attendees, key points, advice, actions, follow-ups - and formats it consistently every time. The finished note is saved directly into the practice management system, ready for the lawyer to review.
Instead of 12 minutes of typing, the lawyer spends 90 seconds scanning the note, making any small edits, and hitting approve.
The results
| Metric | Before | After |
|---|---|---|
| Cost per file note | ~$15 (lawyer time) | $0.25 |
| Time per note | 10-15 minutes | 90 seconds (review only) |
| Daily time saved (per lawyer) | - | ~2 hours |
| Monthly net savings (firm-wide) | - | ~$107,000 |
| Error rate | Variable | Consistent structure every time |
The $107,000 monthly figure accounts for the AI processing costs ($0.25 per call) and is based on the firm's actual call volume and average lawyer billing rates. The savings come from reclaimed billable time - hours that were previously spent on admin work.
How it works
Here's the technical flow from start to finish:
- Call completes, recording uploaded. The call recording is automatically captured and sent to the processing pipeline.
- Audio transcribed using speech-to-text. The recording is converted to a full text transcript with speaker identification.
- Transcript processed by Claude API with firm-specific prompts. The AI analyses the conversation using prompts tuned to this firm's note format and legal terminology.
- Structured note generated. The output includes attendees, key discussion points, advice given, action items, and follow-ups - formatted exactly how the firm needs it.
- Note saved to practice management system. The completed note is pushed directly into the firm's system against the correct matter.
- Lawyer reviews, edits if needed, approves. The lawyer spends 90 seconds checking the note, makes any adjustments, and approves it. Done.
The entire process from call ending to note ready for review takes under two minutes. The lawyer's involvement is reduced to a quick review and a single click.
What the firm owns
Everything. The n8n workflows, the AI prompts, the integration code. There are no ongoing licence fees and no vendor lock-in. If the firm wants to modify the prompts, add new note templates, or connect to a different system - they can. It's their infrastructure.
This is how we work at Workvolve. We build it, hand it over, and you own it. No subscriptions, no "managed service" fees that keep ticking after the project is done.
Frequently asked questions
Works with Actionstep, LEAP, Smokeball, and others via API. If your system has an API or supports webhooks, we can integrate with it.
Very high for structured data extraction. The AI pulls attendees, key points, advice, actions, and follow-ups with strong accuracy. The lawyer always reviews before approving - nothing goes into the system unchecked.
$0.25 per call on average, covering transcription and AI processing. For a firm doing 40+ calls a day, that's roughly $10 a day in running costs - against thousands in reclaimed billable time.
2-4 weeks including integration with your practice management system, prompt tuning to match your note format, and testing against real calls.
Yes - initial consultations, progress updates, settlement discussions, and more. We set up different templates for different call types so the note structure matches the context.