In the winter of 2022, the tech world was becoming fascinated by the sudden and explosive arrival of OpenAI’s ChatGPT. Benjamin Alarie I was faced with a very important choice. his legal tech startup, Blue Jhad a respectable business built on a bygone era of AI, providing predictive models to hundreds of accounting firms. But that was a plateau.
Alarie, Lifetime professor of tax law in University of Torontosaw the early, error-prone but powerful capabilities of large-scale language models as the future rather than a curiosity. He made a high-stakes decision. It meant pivoting the entire company they had painstakingly built over nearly a decade and rebuilding it from the ground up around this unproven technology.
The gamble paid off in spades. Since then, Blue J has been quiet. $122 million Series D Funding round co-led by Oak HC/FT and Sapphire Venturesevaluate the company Over $300 million. This move transformed Blue J from a niche company to one of Canada’s fastest-growing legal tech companies, increasing revenue by approximately 12x and acquiring 10 to 15 new customers every day.
The company currently serves more than 3,500 organizations, including global accounting giants. KPMG This includes several Fortune 500 companies. The company is addressing the severe and worsening talent shortage, a critical bottleneck in the professional services industry. There are 340,000 fewer accountants in the U.S. than there were five years ago.And with 75% of today’s CPAs expected to retire within the next 10 years, companies are desperate for tools to increase the productivity of their remaining professionals.
“What used to take tax professionals 15 hours to manually research can now be done in about 15 seconds with Blue J,” CEO Alarie said in an exclusive interview with VentureBeat. "Much of this is driven by the value proposition of being able to turn tasks that take hours into seconds."
When the dean’s biography went wrong: The moment that changed everything
Alarie vividly remembers the day in January 2023, when the dean of her law school stopped by her office to say New Year’s greetings. He asks her about ChatGPT, prompting the AI to describe her. ChatGPT has created a biography with confidence. Some details were accurate. Others are completely fabricated.
"She said, “Okay, this is really scary.” This is wrong and this has consequences. ”" Alarie said. But even in moments of apparent failure, he remained undaunted. Rather, it embodied his beliefs.
The company’s first iteration, launched in 2015, used supervised machine learning to build predictive models that could predict judicial outcomes on specific tax issues. Although technically sophisticated, it had a fundamental flaw in its inability to answer all tax audit questions.
"The challenge was not being able to answer all the tax audit questions, but this was the holy grail." Alarie said. Customers loved tools that applied to their problems, but were quick to abandon them when they didn’t. Revenue peaked at about $2 million a year.
Despite ChatGPT’s infamous hallucinations, Alarie convinced the board to change course. "I was convinced that if I continued down this path, I would never be able to deal with my biggest limitations." he said. "Large-scale language models seemed like a very promising direction."
He gave his team six months to deliver a working product.
From 90 second responses to 3 million queries: How Blue J overcame the AI illusion
By August 2023, Blue J It was ready to fire. According to Alarie’s honest assessment, what they announced was: "Super junkie." It took 90 seconds for the system to respond. Approximately half of the responses had problems. of net promoter score I was registered when I was just 20 years old.
What turned this flawed product into the platform it is today (with response times in seconds, a complaint rate of only 1 in 700 queries, and an NPS score in the mid-80s) was a relentless focus on three strategic pillars.
The first is original content at scale. Blue J Secure exclusive license Tax Analyst (Tax Note) and IBFDa global tax authority based in Amsterdam and covering more than 220 jurisdictions. "We are the only platform on the planet that brings in the best US tax information from Tax Notes and the best global tax information from IBFD." Alarie said.
The second is deep human expertise. Blue J employs the following tax professionals: susan masseyspent 13 years in IRS Office of the Chief Counsel Appointed as Director of the Corporate Tax Branch. Her team is constantly testing the AI and improving its performance.
Third, there is an unprecedented feedback flywheel. With over 3 million tax audit queries processed by 2025, Blue J has amassed unparalleled data. Each query generates feedback back to the system.
Weekly active user rates hover between 75% and 85%, compared to 15% to 25% on traditional platforms. "The ratio of philanthropy is such that we are 5 times more intensive." Alarie pointed out.
Inside Blue J and OpenAI’s early access partnership
Blue J maintains: Unusually close relationship with OpenAI It has proven essential to success. "We have a very good relationship with OpenAI and have early access to their models."Alarie said. "It’s quite cooperative. We have provided a lot of very high quality feedback on the performance of various versions of the upcoming model."
Blue J is doing what Alerie calls development, so this feedback will prove valuable. "ecologically valid" Test questions — drawn from real tax professional questions, with correct answers determined by Blue J’s team of experts. This allows OpenAI to improve performance on complex inference tasks.
The company tests models from all major providers – OpenAI, human, google geminiand open source alternatives – continually evaluate which performs best. "We are not necessarily 100% committed to any particular provider." he explained. "We are constantly testing."
This approach will help Blue J Overcome a challenging business model of charging approximately $1,500 per seat per year for unlimited queries while absorbing fluctuating computing costs. "We are upfront committed to providing you with an exceptional user experience, unlimited tax answers for a flat fee." Alarie said. "We absorb much of that risk."
Competition among underlying model providers creates downward pressure on API prices, but Blue J’s conservative usage modeling has proven accurate. The gross revenue retention rate is over 99% and the net revenue retention rate is 130%, considered best-in-class for a SaaS business.
Competes with Thomson Reuters and LexisNexis with 75% weekly engagement rate
Blue J facing competition from established publishers such as Thomson Reuters, lexisnexisand bloombergHowever, Blue J’s engagement metrics suggest that the company has gained significant momentum, growing from just 200 customers in 2021 to over 3,500 organizations today.
You will find that daily updates are very important. While the tax code itself only changes when Congress acts, the ecosystem constantly evolves through IRS regulations, new rulings, and court cases. All 50 states change their tax laws periodically.
"The situation is changing literally every day." Alarie said. "We update our material daily, but only in the US, and also cover Canada and the UK. The desire for this is truly global."
Alarie’s ambitions extend beyond building successful startups. As the author of an award-winning book "legal singularity" and faculty affiliates. Vector Artificial Intelligence LaboratoryHe has spent years pondering the long-term impact of AI on law.
Across academic papers published in Tax Notes 2023 and 2024he chronicled the rise of generative AI and predicted: "Clients will be significantly more sophisticated" And AI will propel human experts away from mundane research and into more valuable strategic roles.
Blue J’s $122 million plan: From tax audits to ‘global tax awareness’
of Series D fundingTotal funding exceeds $133 million and will fuel aggressive geographic and product expansion. Blue J already operates in the United States, Canada and the United Kingdom, and plans to eventually cover more than 220 jurisdictions through its IBFD partnership.
Future features could include automatic note generation, tax form filling, document creation, conversation history that maintains context between sessions, and more. This transforms Blue J from an investigative tool to the tool Alarie describes. "Global tax awareness operating layer."
Although Blue J has been successful, it operates in an area where errors can have significant consequences. The problem of hallucinations has not yet been resolved. Minimized through careful engineering, content curation, and human oversight. Rather than fabricating information, Blue J trained its model to recognize when it can’t answer a question.
Businesses also face financial risk if computing costs rise or usage patterns exceed expectations. And delicate questions arise regarding professional judgment. As AI systems become more capable, will users withhold output without sufficient critical evaluation?
From 15 hours to 15 seconds: What Blue J’s AI pivot can teach every industry
Blue J’s transformation offers lessons beyond tax software. The company’s willingness to abandon eight years of proprietary technology and rebuild on an initially unreliable foundation required both courage and calculated risk-taking.
This decision worked, not because generative AI was inherently better than supervised machine learning in every aspect, but because it addressed the right problem of comprehensiveness rather than narrow accuracy. Tax professionals didn’t need 95% accuracy on 5% of the questions. Sufficient accuracy was required for 100% of the questions.
The improvement in NPS from 20 to 84 in just over two years reflects constant iteration through massive data collection. Content partnerships have created a differentiation that cannot be replicated through pure technology. A team of tax experts provided the necessary expertise to ensure credibility.
Most fundamentally, Blue J recognized that its real competition wasn’t other AI startups or established publishers. That was the old way. 15 hours of manual research and organizational knowledge locked away in the retiring professional’s head.
"People are like, “What is Blue J doing?” They provide better answers regarding tax matters. Okay, I think that’s what you need.”" Alarie reflected.
As AI transforms various professions, clarity of purpose may become more important than technological sophistication. The future does not belong to those who build cutting-edge AI, but to those who use it most effectively to solve real human problems.
For a tax law professor who started out of frustration with inefficient research methods, starting a $300 million company represents an audacious end goal. For thousands of professionals, being able to answer complex questions in 15 seconds instead of 15 hours represents the future of their profession, arriving sooner than most expected.
Back when ChatGPT was still a fantasy biography, the bet proved that sometimes the riskiest thing to do is not move at all.
