Why I Help Canadian Lawyers Adopt AI Responsibly

Disclosure and scope: AI assisted with drafting and structural refinement. I reviewed and revised the final content, verified its factual claims, and accept responsibility for its analysis and final form. I am an AI strategist, not a Canadian lawyer. This article addresses operational adoption and workflow design. It is not legal, privacy, cybersecurity, contractual, or technical advice.

Knowing the Procedure Is Not the Same as Performing It

Current as of July 2026.

The new cabin crew I trained had learned the procedures. They had passed the exams.

What they struggled with was sequence and timing.

English announcement, then French — and remembering the second once the first was behind them and three other things were competing for attention. Knowing which point in the flow called for which announcement, rather than knowing the announcement itself.

The manual told them what. It did not build the habit of when.

So I taught the order, not only the content. Once the sequence was automatic, their attention was free for the things the manual could not anticipate.

That is the point. Procedure is not the opposite of judgment. It is what clears the space for it.

I spent nine years on that problem — first as a lead cabin crew member, accountable for the conduct of a crew under regulated procedures, then as an onboard trainer, accountable for whether those procedures survived contact with people who were tired, rushed, or facing something the manual described in general terms and the day delivered in specific ones. My job was not to write the procedures. It was to find where they broke down in practice and make them easier to follow correctly.

The Same Problem, a Different Profession

Generative AI produces fluent, confident, well-organized text. Fluency is not accuracy. A polished paragraph can contain an unsupported conclusion, a misstated holding, or a citation to something that does not exist — and the polish is what makes those defects harder to catch.

Alberta's three courts addressed this directly. Their joint Notice to the Public and Legal Profession of October 2023 does not simply ask for a human in the loop. It asks that AI-generated material be verified with "meaningful human control," and it says what that means in practice: cross-referencing against authoritative sources until the citations and their content hold up to scrutiny.[1]

That distinction is where my work sits. A human becomes an effective control when four conditions hold: someone knows specifically what must be checked, knows which sources are authoritative, has the time and the mandate to check, and recognizes the point at which a question exceeds their competence or authority. Miss one and what you have is a signature rather than oversight.

Why This Lands Hardest on Canadian Lawyers in Small Firms

A solo practitioner faces the same questions as a national firm — confidentiality, competence, vendor terms, disclosure — without a privacy office, a security team, or an IT department to route them to.

And AI rarely arrives as a project you can govern. It arrives as a chatbot used for a first draft, a consumer tool used to summarize a document, a feature switched on inside software the firm already licenses. Use develops before anyone has decided what is permitted, what information may be processed, and what review is required before anything leaves the office.

The Law Society of Alberta's Generative AI Playbook grounds generative AI in duties lawyers already carry under the Code of Conduct, including the obligation to develop an understanding of technology relevant to their practice.[2] Its guidance on why a firm needs a Gen AI use policy sets out what a written policy should cover.[3]

Between them, these sources identify the relevant duties, the risks, and the questions a firm should work through. The gap I work in sits downstream of that: turning them into something a two-person firm can actually run on a Tuesday afternoon.

What I Do

I help small Canadian law firms convert responsible-AI principles into procedures people can follow. That means examining current use, including the use nobody formally approved; assessing specific workflows rather than tools in the abstract; defining what information may be processed where; assigning review to named people rather than to the firm generally; and keeping the records that show, afterward, what was decided and why.

Where legal, privacy, cybersecurity, contractual, or technical questions arise, I organize the facts and identify their operational implications so the appropriate professional can advise. I do not replace those professionals, and I will tell you when you need one.

What Would Make Me Tell You Not to Proceed

I published a piece on how to vet an AI strategy consultant — How Should a Canadian Law Firm Vet an AI Strategy Consultant?[4] One of the questions I said you should ask is what would make the consultant recommend against proceeding. It would be poor form not to answer it myself.

The problem is a process problem. If intake is inconsistent or precedents are scattered, AI will produce inconsistent results faster. Standardize first. Automate second.

Nobody can own the review step. If everyone competent to verify the output is already at capacity, adding a tool that generates more material to verify makes the firm less safe, not more productive.

The matter is too sensitive for the available controls. Some files should not touch a general-purpose tool, and prompting discipline does not change that.

The firm wants speed without changing the process. That is the request I decline outright. It is also the most common one.

I would rather lose the engagement than help a firm build a defensibility problem.

I Apply This to My Own Work

I separate AI-assisted drafting from source verification and final editorial judgment, and I do them in that order, as distinct steps. Every case, statute, and regulatory reference in my published work is checked against a primary source — CanLII, Justice Laws, the King's Printer, or the issuing law society — before publication. Where I cannot confirm something, I say so or I cut it.

Draft in seconds. Verify against the source. Approve with judgment.

The goal was never to write faster. It was to make the work easier to check.

Why I Do This Work

New tools make professional work faster without making it more dependable. The distance between those two things is where the risk sits, and it is not visible in the output.

Small firms can close that distance. It takes deciding on purpose, before the tools are already in the workflow — which is cheaper than deciding afterward, and considerably less unpleasant.

References

1. Court of Appeal of Alberta, Court of King's Bench of Alberta, and Alberta Court of Justice. Notice to the Public and Legal Profession: Ensuring the Integrity of Court Submissions When Using Large Language Models, 6 October 2023.
   

2. Law Society of Alberta. The Generative AI Playbook: How Lawyers Can Safely Take Advantage of the Opportunities Offered by Generative AI, January 2024 (last updated February 2026).
   

3. Law Society of Alberta. Why a Generative AI Use Policy? 
   

4. Klein, Charles Austin. How Should a Canadian Law Firm Vet an AI Strategy Consultant? 18 July 2026.
   

Charles Austin Klein is an AI integration strategist working with solo and small-firm lawyers in Alberta and British Columbia. The Canadian AI Operations Toolkit puts the questions, records, and review structures described here into a form a small practice can run on its own.