Solutions

The right tools, set up right — and built, when nothing fits.

The core of my work is implementation and training. The custom tools below exist because client firms genuinely needed them.

01The Core Offer

AI implementation & training

A sequence, not a product — because adoption fails when any one of these steps is skipped.

  1. Assess & select

    I start with how your firm actually works — practice areas, document flows, who does what — and select AI software that fits. That has meant Claude for the teams I've worked with so far, but the tool follows the need.

  2. Implement & configure

    Installation, account structure, and configuration inside your firm's own accounts. Data residency, retention settings, and access controls are decided deliberately during setup — not left to vendor defaults.

  3. Train & enable

    Hands-on sessions for lawyers and staff: structured prompting, verifying AI output, and embedding AI into the daily work your team already does. Training is practical and specific to your workflows — not a generic AI seminar.

  4. Build only if needed

    Where off-the-shelf tools leave a genuine gap, I build. The two tools below both started this way.

02Custom Tool

CanLII citation verification

A stateless MCP server that lets your AI assistant check citations against the real database while it drafts.

Large language models can produce citations that look right but aren't. This tool gives your AI assistant a way to verify — a Model Context Protocol (MCP) server that acts as a translation layer between your assistant of choice and the CanLII API, checking case citations, case IDs, and metadata against the actual database during drafting and review.

Step 1
Your firm's AI assistant
e.g. Claude — asks to verify a citation
query
Step 2
Stateless MCP layer
hosted on Cloudflare — stores nothing
lookup
Step 3
CanLII API
verified citations, case IDs & metadata returned

Only public citation queries travel this path — no client documents, no confidential material.

How the engagement works. Your firm connects with CanLII and obtains its own API key — you own the API relationship. I deploy the stateless MCP server on Cloudflare using that key, and connect it to your AI assistant as a verification tool it can call while working.

The server is stateless and handles only public citation lookups. Your documents and client information don't route through it, and your firm keeps control of its own CanLII access throughout.

03Custom Tool

Voice transcription

Meeting notes, transcribed — with the data staying in Canada.

A transcription tool for legal teams that turns meetings and dictations into usable notes. It's built on Google Cloud's Canadian region, with data residency handled in the code — so transcripts are processed and stay on Canadian servers, in your firm's own cloud environment.

04Approach

Security & privacy, by design

Everything I set up follows the same principles: your firm owns every account and API key, custom tooling is stateless wherever possible, and systems are built with Canadian data residency and privacy obligations in mind.

No badges, no certifications-as-marketing — just an architecture you and your IT team can inspect and control. If you'd like to walk through the details, get in touch.