The OBA AI Academy: Training with Tomorrow’s Tools

The OBA AI Academy is free for all Ontario Bar Association members, and, through a partnership with the Canadian Bar Association, available at no cost to CBA members across Canada until January 31, 2026. Artificial intelligence has a way of making lawyers nervous: for some, it conjures headlines about job loss and disruption; for others, it sparks curiosity about a future just beyond the edges of familiar practice. But AI is no longer a distant prospect. It is already embedded in the software we use, in the services our clients expect, and in the rules of professional conduct that tie competence to technological awareness.

Developed by Colin Lachance, the OBA’s Innovator-in-Residence, the Academy is designed to be both educational and practical. It teaches lawyers when and how to use AI responsibly, the kinds of prompts that work, and the situations where these tools can add real value. Lessons are delivered through self-paced modules, reinforced by a secure sandbox that lets members practice in a low-stakes environment.

Built into the program is LawQi, an AI assistant developed by PGYA Consulting and Praxis AI for the Academy. LawQi is designed as a mentor for lawyers, prioritizing the Academy’s own course materials over generic internet answers. The Academy describes it as a kind of “digital guide,” and I found that apt: it works like an Ariadne’s thread through unfamiliar terrain. I decided to see how well that thread would hold when tested against the daunting labyrinth of the Income Tax Act.

Learning by Doing

Before facing the threat of the labyrinth, I had to prepare. So, I worked through all the Academy modules available at the time, which took me about two and a half hours. For context, I had used AI before, thus some concepts were familiar, but the Academy still sharpened my approach.

The modules walk you through the common pitfalls of using AI, like hallucinations and over-reliance, while also showing how to get the best results through prompt techniques and creative applications. Take hallucinations: when AI makes up a case citation or statutory section, it is not “lying” so much as doing what it was designed to do, predict the most plausible next words. If no real case exists, it may “hallucinate” one that sounds convincing. As lawyers, we often assume there must be a case for every concept, only to be disappointed when research shows otherwise. AI seems to share that same disappointment, except it sometimes goes a step further, inventing authority to fill the gap. Hopefully, none of us will ever feel so desperate in the labyrinth that we start hallucinating precedent ourselves.

On the creative side, the Academy highlights prompts that go beyond simple Q&A. You can ask AI to build comparison tables, turn dense text into point-form clarity, or reframe material in ways that make complex issues easier to grasp. Even more practical is the ability to upload documents, like a contract, and have the AI flag clauses that are particularly strong, weak, or unusual. Used responsibly, these tools can become a real aid to legal practice, helping lawyers spot issues more quickly and present information in client-friendly ways.

The lessons I found most useful were those on legal prompt engineering (Track 1.3), which really open your mind to what can be asked of AI; the overview of commercial AI models and legal use cases (Track 2.3), especially section 2.2.2 on ChatGPT for legal drafting; and the module on professional guidance (Track 4.1), which walks through the Law Society’s direction on responsible AI use. Taken together, these sessions do a good job of grounding you in both the practical risks and the opportunities of using AI in legal practice.

The Academy also runs on a point system, awarding points for every module completed. There is even a leaderboard, purely for fun. At the time of writing, I sit eighth with 5,100 points. Seven others remain ahead of me, something that continues to irk my competitive spirit. If you manage to pass my score, dinner is on me, just let me know how you earned those extra points.

With those preparations complete, I was ready to test whether the Academy’s thread could guide me through the Income Tax Act.

Testing in a Fiscal Context

I decided to put LawQi to the test by uploading some of the densest provisions in the Income Tax Act, sections 18.2 and 18.21, namely EIFEL. My Word document ran to 49 pages, so good luck buddy. When you upload a file, LawQi immediately starts analyzing. Oddly, even though my document was entirely in English, it launched into French. Maybe it was thrown off by the word “Eifel” in the file name, or maybe it just sensed I was francophone through the screen. Either way, not quite the Eiffel Tower I was trying to climb.

Setting the language mishap aside, I asked the very easy task to explain EIFEL in three lines. LawQi came back with this:

“(1) EIFEL limits how much interest and financing expenses a taxpayer can deduct, to prevent base erosion; (2) the cap is calculated by formula using adjusted taxable income and a ratio of permissible expenses; and (3) the rules include anti-avoidance measures, allow transfers of unused capacity, and provide group allocation mechanisms.”

Fair enough, that was a great summary.

Next, I uploaded the thin capitalization rules, sections 18(4) to 18(8), about 11 pages of material. LawQi’s three-line summary was similarly crisp:

“(1) thin cap prevents Canadian corporations and trusts from deducting interest paid to specified non-resident shareholders/creditors when the company's debt-to-equity ratio exceeds 1.5; (2) it defines “specified non-residents” and calculates equity differently for corporations versus trusts; and (3) it includes anti-avoidance rules for back-to-back loans, with an exception for controlled foreign affiliates.”

Here I noticed a gap: thin cap also applies to partnerships, albeit indirectly through subsection 18(7). I pointed this out, and with a little prodding LawQi corrected itself and explained the deeming mechanism that attributes a partner’s share of partnership debt for thin cap purposes.

When I asked it to boil the regimes down into a single 3×5 table of key distinctions, it produced the following:

It does bear noting that a limitation of LawQi’s response was its omission of key complexities and disctintions. One important distinction it overlooked is that thin-cap–denied interest is recharacterized as a dividend. Likewise, the final heading, “Group Coordination”, struck me as imprecise. In tax law, “group” is not a term I would use to describe thin cap’s partnership attribution rule.

The more substantive debate came when I asked about carryforwards. Having written on EIFEL in my LL.M., I found LawQi’s claim of “unlimited carryforward of unused capacity” misleading. At the very least, the term “unlimited” was ambiguous, did it mean unlimited in time, in amount, or in both? When I flagged this, LawQi first pointed me to subsection 18.2(6), but on inspection that provision clearly deals with filing obligations for capacity transfers, not carryforwards. This misstep forced us to dig deeper.

Working through the mechanics, I realized LawQi was conflating two concepts. The first is unused excess capacity: deduction room under the 30% rule that can be carried forward for three years. The second is denied interest when a taxpayer exceeds the limit: those amounts are not lost but instead become Restricted Interest and Financing Expenses (RIFE), which can be carried forward indefinitely under section 111(1)(a.1).

Initially, LawQi told me denied interest was gone forever, unlike EIFEL’s three-year carryforward of unused capacity. Knowing that was wrong, I pressed it until it corrected itself. By the end of the exchange, the distinction was clear: thin cap permanently denies interest and recharacterizes it as a dividend, while EIFEL creates a two-track system, (1) excess capacity, carried forward for three years, and (2) denied interest, transformed into RIFE with indefinite carryforward.

Overall, having some familiarity with these provisions before probing with LawQi does help ensure the summary is precise. Still, even without that background, LawQi gives a solid representation of the purpose and basic mechanics, distilling them into just three sentences. The Ariadne thread may need a bit of tugging now and then, but it remains a highly useful guide.

Looking Ahead

AI is not meant to replace lawyers, but to equip them. The future of practice will not be defined by whether AI exists, but by how thoughtfully we weave it into our work. Ariadne’s thread could not defeat the Minotaur on its own; it mattered because Theseus knew how to use it. AI is the same: a tool whose effectiveness depends on the judgment and skill of its user.

For me, working through the Academy was worthwhile precisely because it offered a structured way to engage with AI: to see its promise, test its limits, and learn about the professional questions it raises. It is an invitation to experiment, to question, and to reflect on where the profession is heading.

One final note: LawQi itself always ends its answers with a quotation, and in that spirit, I’ll do the same here.

“The future is not set. There is no fate but what we make for ourselves.” - Terminator 2

Any article or other information or content expressed or made available in this Section is that of the respective author(s) and not of the OBA.