Since its inception in 2015, Blue J has evolved from a directed research project at the University of Toronto into the definitive AI-first operating layer for tax and legal professionals. Today, the company employs approximately 130 people and serves over 5,000 law and accounting firms. By moving beyond simple wrappers and building an architecture that respects the hierarchical logic of the tax code, Blue J has achieved a 80%+ monthly active usage rate – a stark contrast to the 15-25% adoption typical of legacy tax research tools.
In this interview, CEO and co-founder Ben Alarie discusses the technical and philosophical journey from the classroom to the cutting edge of legal tech.
Q: How did you make the transition from academics to startup founder? What made you do it? And, are you still teaching?
Alarie: My trajectory was very academic early on. I moved quickly from studying economics as an undergraduate to law school, then to graduate work in economics and then to graduate work in law at Yale, and finally to a Supreme Court clerkship in Canada. In 2004, at age 26, I became the youngest ever tenure-track law professor at the University of Toronto. Over the next decade, I wrote dozens of peer-reviewed articles and co-authored one of the leading textbooks for teaching tax law.
The turning point for the Blue J story was when I led an effort to change the curriculum – something that hadn’t been done in more than 40 years. While senior colleagues cautioned me against it, I drew the opposite conclusion. It got me thinking: What will change over the next 40 years? That’s when I started focusing on machine learning and AI, and what they specifically meant for the future of tax law.
It felt like being struck by lightning. I was in my mid-30s and realized I could easily be at the law school for another 30+ years, but I’d be increasingly cornered by technology. I was afraid I’d be standing at the front of a classroom with chalk in my hand in 2040, wondering why I didn't do something earlier.
I still have a full-time, tenured appointment with an academic chair at the University of Toronto’s law school, although I am currently on leave. Given how quickly the business is growing, the right place for me at this moment is not at the front of a classroom, but in a seat with my team, working to keep everything pointed in the right direction.
Q: What do you mean by "AI coming for the law?”
Alarie: In 2023, I published a coauthored book called The Legal Singularity about how AI can make law radically better. What lawyers do is fundamentally make predictions about what the law requires and how judges will adjudicate cases; advocacy is essentially a bundle of judgments and predictions. Because machine learning and AI are algorithmic prediction machines, they are perfectly suited for this work.
As algorithms recreate what is central to the practice of law, there is an interesting back-and-forth regarding the boundaries of relative competence between algorithms and human professionals. I concluded the right thing to do was to start a company. We incorporated Blue J Legal Inc. on April 21, 2015, and set out to leverage AI to make tax law radically more transparent. Accelerating tax research was our ambition from the very outset.
Q: Was there a specific moment when you knew this needed to become a product?
Alarie: Late in 2014, I approached the Computer Science department at the University of Toronto to create a cross-listed directed research course, which we established in January 2015. We had eight students and met weekly in an innovation lab to whiteboard our approach. We were two-thirds of the way through the semester when we hit on a strategy that allowed us to predict, with 90% accuracy, how certain cases would be resolved based on a description of the facts. That was the moment I realized, "Okay, this actually works." Then I had to figure out how to actually start a company, because at that point, I was just a tax law professor.
Q: How does your background in law and economics shape your development velocity? Academia usually has much longer timelines than a startup.
Alarie: I wasn't your prototypical law professor. I was always ambitious with timelines, and I was wired for "team production" through my experience co-authoring books. Law and economics is essentially about decision-making under uncertainty, which is a perfect mindset for a startup environment where you are constantly making informed bets.
I use a two-pedaled mind approach: convergent thinking (logic, data, and structured reasoning) and divergent thinking (creativity and imagining the "art of the possible"). You need both pedals to move the bicycle forward.
Q: How did you come up with the name "Blue J"?
Alarie: In legal writing, a "J" behind a surname signifies a "Judge." I would be Alarie J. if I were writing a judgment. So, "Blue J" is essentially "Judge Blue" or "Justice Blue." We chose the color blue because we were based at the University of Toronto (the Varsity Blues) and we initially thought we’d be using IBM’s technology (Big Blue). It was a way to fit those connections and a legal tech identity into just five characters. Plus, I could get the domain name.
Q: Did Blue J move beyond standard LLM wrappers to build an architecture that understands the hierarchical logic of the tax code?
Alarie: We actually started out not using LLMs at all. We used supervised machine learning models because LLMs weren't effective at the time. We only started leveraging transformers and GPT-3 when they became proficient.
Most AI tools in this space just point foundational models at legal text, which results in a sophisticated summarization experience rather than a true tax research platform. We built something different because we are tax experts. We understand the specific hierarchy of authority: you start with the Internal Revenue Code, then regulations, then case law, and finally IRS guidance. Our architecture mirrors that hierarchy.
When you ask Blue J a question, it reasons through that source structure. Behind the scenes, it is a series of sophisticated agentic workflows that operate depending on the kinds of tax questions asked. It’s not just running a single prompt against a model; it’s orchestrating multiple reasoning steps, choosing the right sources, and assembling a response that is attentive to the user’s goal. We also have content exclusivity with Tax Notes and IBFD, and we plug directly into the internal libraries of major firms.
Q: Is it true that tax can be recursive and prone to hallucination risks? How do you handle "gray areas"?
Alarie: Every LLM has hallucination risk, but we treat it as an engineering problem. We mitigate it by grounding the model in authoritative primary sources. Hallucinations often occur when there is a gap in the material; with complete coverage, we reduce the model's need to "make something up."
Regarding gray areas, we handle them the way a human expert would. If you ask an ambiguous question, the system will usually explain that it is a fraught area that could go either way and detail what the outcome depends on. We don’t force a black-and-white resolution if the law is truly gray. Our goal is not to replace professionals but to provide the scaffolding that amplifies their judgment.
Q: How did the AI revolution impact your business?
Alarie: The AI revolution, led by ChatGPT, was a massive tailwind. It validated the technology for our customers; their first exposure might have been writing a Mother's Day poem, which made them less afraid to try it for work. It softened the market and allowed our product improvement flywheel to move much faster.
We doubled our ARR and customer count in the first half of 2025, and in March, we signed up our 5,000th law or accounting firm. We started with zero in mid-2023 when we stood up this new LLM-based version of Blue J. Our "disagree rate" has fallen from 1% to 0.1% over the last couple of years. We remain model-agnostic, leaning on Google, OpenAI, or Anthropic depending on which is strongest for a given query, allowing us to produce an answer that is stronger than any single model could on its own.
Blue J's story is a reminder that the most durable AI companies aren't built by technologists who discovered a problem, but rather by domain experts who discovered a solution. For Ben, the decade spent in classrooms and courtrooms wasn't a detour from the startup world. It was the foundation for it. As AI continues to reshape the legal and tax landscape, Blue J's early bet on depth over shortcuts looks less like a strategic choice and more like an inevitability.