Fixed-fee immigration boutique
A solo practitioner with a paralegal, drowning in visa-refusal appeals. We built a matter container, published fixed pricing, and put a first-pass drafting agent on every appeal.
Practice
Singapore · Solo + 2
Transformation
9 weeks
Track
AI-Native Transformation
Primitives moved
Matter, Actions, Time
The challenge
The practice was profitable but exhausting: every appeal was billed hourly, every client call reopened a negotiation about the bill, and the founder was personally re-drafting near-identical appeal letters from memory each time, with no written record of what actually worked.
The transformation, phase by phase
Phase 01
Name
Two weeks of read-only access to the case file system and calendar. We mapped what a 'matter' actually looked like across forty closed appeals and found five distinct action sequences hiding inside what the founder called 'one kind of case.'
Phase 02
Systematise
Each of the five sequences became a named playbook with a first-draft template. Deadlines - which had lived in a paper diary - moved into a structured calendar attached to each matter, surfacing automatically two weeks before any statutory cutoff.
Phase 03
Automate
A drafting agent now produces the first pass of every appeal letter against the matching playbook, citing the specific refusal ground it's responding to. The founder reviews and finalises - the same role she always had, minus the blank page.
The shift to AI-native
This is the manifesto's argument made concrete at the smallest possible scale. A solo practitioner has no associate layer whose billable hours fund the practice, so every hour AI removed from drafting became recovered capacity, not lost revenue - the exact opposite of what automation does to a leverage-model firm. Fixed pricing only became credible once duration data existed to back it; AI now runs first, by default, on every appeal, with the founder reviewing rather than originating from a blank page. The practice didn't get bigger. It got structurally different - the same one person, running a system instead of running on memory.
AI-native markers this firm now meets
- Pricing is transparent, not hourly
- AI runs first, on every matter, systematically
- Turnaround is fast, often same-day
- The firm gets measurably better over time
Outcome
Turnaround
5 days → 18 hours
Matters / month
12 → 34
Effective rate
+3.1×
By the fourth week we had stopped talking about AI and started running the practice. That was the point I understood what they'd actually built.