Case studies
in fixed-scope delivery.

A senior civil barrister in Auckland, operating as sole practitioner with a small support team. Correspondence was consuming six to eight hours a week of his own time — routine letters to solicitors, standard position letters to opposing counsel, and post-hearing follow-ups that were substantially identical across matters but being retyped from scratch each time.

He had tried AI tools informally — dropping client information into a public chatbot to "get a draft going" — but stopped once he thought through the privilege implications. He needed a system that genuinely worked but with hard boundaries he could explain to the Law Society if asked.

We started with a Readiness Review — a full audit of his correspondence patterns over six months. We identified four workflow families that accounted for roughly 70% of his routine writing time: first-contact letters to instructing solicitors, standard position letters, scheduling correspondence, and post-hearing updates.

Setup involved configuring an AI platform on his own desktop (not a cloud service), building a prompt pack from 40 of his own previous letters, and establishing explicit written operating rules: no privileged client material, no automated sending, every output reviewed by him before use. The system draft letters; he reviewed and refined them.

Three months in, his measured time on routine correspondence had dropped from roughly seven hours a week to under ninety minutes. He was using the time recovered to take on two additional junior briefs per month and was finishing work before 6pm for the first time in years.

An orthopaedic surgeon with a busy private practice was finishing clinic and spending the next two hours drafting referral letters, outpatient summaries, and admin correspondence. His practice manager was handling some of the admin but his clinical time was being eroded by documentation. Weekends were increasingly consumed by outstanding reports.

The enquiry came from the practice manager, not the surgeon. Her specific request: something that worked but absolutely could not touch patient records or the practice management system. The practice uses Medtech, and any AI tool that interacted with that was a non-starter.

We proposed the Clinical-Safe Setup package. The setup was entirely on the practice manager's workstation — a separate system from the clinical PMS. We built two non-clinical workflows: a referral letter drafter working from the surgeon's existing letter templates and his typed clinic notes (non-identifiable, no patient records), and an outpatient report structure builder.

Critically, we produced a one-page written operating document setting out exactly what the AI tools could and could not access. No patient identifiers. No clinical decision support. No Medtech integration. Every output reviewed by the surgeon before use. The practice manager operates the system under his authority.

Referral letter turnaround dropped from roughly two days to same-day. The surgeon reviews drafts the practice manager has prepared rather than writing from scratch. His after-clinic documentation time has dropped from roughly ninety minutes to under thirty.

A small Financial Advice Provider holding its own FAP licence had received informal indications from its professional indemnity insurer that its compliance documentation was "adequate but thin" — the kind of comment that often precedes more formal attention. The firm had policies in place but the documents were generic templates and hadn't been updated since the FMCA came into full effect.

They needed a compliance review that was grounded in how the regulator actually thinks, not one that produced another set of generic policy templates. They'd had one of those before and it hadn't helped.

We conducted a two-week fixed-scope review covering AML/CFT compliance, FMCA advice-giving obligations, Code of Professional Conduct obligations, and record-keeping practices. Rather than producing new templates, we stress-tested the existing documents against realistic supervisory scenarios.

The output was a written assessment identifying nine specific gaps — some significant, some trivial — with a prioritised remediation plan. Each gap included an explanation of why it would matter to a supervisor, not just a citation of the relevant section of the Act.

The firm addressed the three high-priority items within a month. When the next supervisory touchpoint came six months later, the firm's representatives reported that the conversation went better than any previous interaction. The engagement cost significantly less than the insurer's estimate for remediation work if a problem had arisen.