In the winter of 2022, when the tech world was being mesmerized by the sudden, explosive arrival of OpenAI’s ChatGPT, Benjamin Alari Faced with an important choice. His legal tech startup, blue jayBygone era there was a respectable business built on AI, serving hundreds of accounting firms with predictive models. But it hit a limit.
Alari, A Tenured Tax Law Professor But University of TorontoViewed the nascent, error-prone, yet powerful capabilities of large language models not as a curiosity, but as the future. He made a hugely risky decision: he had to rebuild his entire company, which had been built painstakingly over almost a decade, on this unproven technology.
That bet has paid off handsomely. The Blue Jays have since quietly secured a $122 million Series D Co-led funding round Oak HC/FT And Sapphire VenturesKeeping the valuation of the company over $300 millionThis move transformed Blue Jay from a niche player into one of Canada’s fastest growing legal technology firms, increasing its revenues nearly twelvefold and attracting 10 to 15 new clients every day,
The company now serves more than 3,500 organizations including global accounting giants kpmg and many Fortune 500 companies. It is tackling a serious hurdle in the professional services industry: a severe and worsening talent shortage. There are 340,000 fewer accountants in the US than five years agoAnd with 75% of current CPAs expected to retire in the next decade, companies are desperate for tools that can increase the productivity of their remaining experts.
“The work that used to take tax professionals 15 hours of manual research can now be completed in about 15 seconds with Blue Jay,” company CEO Alari said in an exclusive interview with VentureBeat. "That value proposition – we can take hours of work and turn it into seconds of work – is what’s driving a lot of it."
When Dean’s Biography Was Wrong: The Moment That Changed Everything
Alari remembers well in January 2023, when the dean of the law school stopped by his office to wish him a happy new year. They asked him about ChatGPT and prompted the AI to describe it. Chatgpt confidently prepared a biography. Some details were accurate. Others were complete fabrications.
"She was like, ‘Okay, this is really scary. This is wrong and has implications." Alari said. Yet that moment of apparent failure did not deter him. Instead, it made his convictions more evident.
The first iteration of the company, launched in 2015, used supervised machine learning to create predictive models that could predict judicial outcomes on specific tax issues. Although technically sophisticated, it had a fundamental flaw: it could not answer every tax research question.
"The challenge was that it could not answer every tax research question, which was really the holy grail," Alari said. Customers liked the tool when it applied to their problem, but quickly abandoned it when it didn’t. Revenues hovered around $2 million annually.
Despite ChatGPT’s notorious hallucinations, Alari convinced his board to pivot. "I firmly believed that if we continued down that path, we would not be able to address our number one limitation," He said. "Large language models seemed like a very promising direction."
He gave his team six months to deliver a working product.
From 90-second reactions to 3 million questions: How Blue Jay mastered AI hallucinations
By August 2023, blue jay Was ready to launch. In a frank assessment of Alari, what he issued was, "Super Junkie." The system took 90 seconds to respond. About half the answers had problems. net Promoter Score Registered at just 20.
What transformed that flawed product into the platform it is today – response time measured in seconds, a dissatisfaction rate of only one in 700 queries, and an NPS score in the mid-80s – was a sustained focus on three strategic pillars.
The first is largely proprietary content. blue jay obtained special license with Tax Analyst (Tax Notes) And IBFDThe Amsterdam-based global tax authority covers 220+ jurisdictions. "We are the only platform on earth that has the best US tax information from TaxNotes and the best global tax information from IBFD." Alari said.
The second is deep human expertise. Blue Jay is staffed by tax experts susan masseywho spent 13 years IRS Office of Chief Counsel As branch head for corporate tax. Their team constantly tests the AI and improves its performance.
The third is an unprecedented feedback flywheel. With over 3 million tax research queries processed in 2025, Blue Jay is collecting unparalleled data. Each query generates feedback that flows back into the system.
Weekly active user rates range between 75% and 85%, compared to 15% to 25% for traditional platforms. "A charitable ratio is such that we use five times more intensively," Alari noted.
Inside Blue Jay’s early access partnership with OpenAI
blue jay maintains a Unusually close relationship with OpenAI Which has proved important for its success. "We have a great relationship with OpenAI, and we get early access to their models,"Alari said. "It is very helpful. We give them really high quality feedback on how well different versions of upcoming models are performing."
This feedback proves valuable as Blue Jay develops what it calls Alari "ecologically sound" Test Questions – drawn from real tax professional questions, the correct answers of which have been determined by Blue Jay’s expert team. This helps OpenAI improve performance on complex reasoning tasks.
The company tests models from all major providers – OpenAI, anthropic, Google’s Geminiand open-source alternatives – while constantly evaluating which ones perform best. "We are not necessarily 100% committed to any particular provider," he explained. "We are testing all the time."
This approach helps blue jay Navigate a challenging business model: charging approximately $1,500 annually per seat for unlimited queries while absorbing variable compute costs. "We’ve already committed to giving them a really good user experience, unlimited tax research answers at a fixed price." Alari said. "We are bearing the lion’s share of that risk."
Competition among foundation model providers creates downward pressure on API pricing, while Blue Jay’s conservative usage modeling has proven accurate. Gross revenue retention exceeds 99%, while net revenue retention reaches 130% – considered best-in-class for SaaS businesses.
Thomson rivals Reuters and LexisNexis with 75% weekly engagement
blue jay face competition from established publishers such as Thomson Reuters, LexisNexisAnd bloombergAll of which have announced AI capabilities during 2023 and 2024. Yet Blue Jay’s engagement metrics show it has gained significant momentum, growing from just 200 customers in 2021 to more than 3,500 organizations today.
Daily updates prove important. While the tax code only changes when Congress acts, the ecosystem constantly evolves through IRS rulings, new rulings, and court cases. All 50 states regularly revise their tax codes.
"Things are literally changing every day," Alari said. "Every day we’re updating the content, and it just covers the US, we cover Canada, we cover the UK. The aspirations for this thing are truly global."
Alari’s ambitions extend far beyond building a successful startup. As the author of an award-winning book "legal singularity" and faculty affiliates Vector Institute for Artificial IntelligenceHe has spent years considering the long-term impact of AI on law.
In academic papers published throughout Tax Notes 2023 And 2024He described the rise of generative AI by predicting "Customers will become substantially more sophisticated" And that AI will push human experts toward high-value strategic roles rather than routine research.
Blue Jay’s $122 million plan: From tax research to ‘global tax cognizance’
Series D fundingBringing the total capital raised to more than $133 million, it will drive aggressive geographic and product expansion. Blue Jay already operates in the US, Canada and the UK, with plans to eventually cover 220+ jurisdictions through its IBFD partnership.
Future capabilities could include automated memo generation, tax form completion, document formatting, and conversation history that maintains context across sessions – transforming Blue Jay from a research tool to a library. "Operational layer for global tax awareness."
For all its success, Blue Jay operates in a domain where errors have serious consequences. The problem of hallucinations has not been eliminated – it has been minimized through careful engineering, content curation, and human oversight. Blue Jay has trained its models to admit when they can’t answer a question rather than fabricate information.
The business also faces economic risks if computation costs increase or usage patterns exceed expectations. And subtle questions about professional judgment are emerging: As AI systems become more capable, will users disregard outputs without adequate critical evaluation?
From 15 hours to 15 seconds: What Blue Jay’s AI pivot teaches every industry
The Blue Jay’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 courage and calculated risk-taking.
The decision paid off not because generative AI was inherently superior to supervised machine learning in all dimensions, but because it addressed the right problem: comprehensiveness rather than accuracy in narrow domains. Tax professionals did not require 95% accuracy on 5% of the questions. They required adequate accuracy on 100% questions.
The improvement from NPS 20 to 84 in just two years reflects sustained iteration informed by large-scale data collection. Content partnerships created differentiation that pure technology could not replicate. The team of tax experts provided the necessary domain knowledge to ensure credibility.
Most fundamentally, Blue Jay recognized that the real competition was not other AI startups or even established publishers. It was the old way of doing things – 15 hours of manual research, institutional knowledge locked in the minds of retired professionals.
"People say, ‘What does a Blue Jay do?’ They provide better tax answers. Well, I think we need it,’" Alari reflected.
As AI continues to transform one profession after another, clarity of purpose may matter more than technical sophistication. The future does not belong to those who build the most advanced AI, but to those who use it most effectively to actually solve humans’ problems.
For a tax law professor who started out with frustrations about inefficient research methods, building a $300 million company marks a courageous endpoint. For thousands of professionals who are now answering complex questions in 15 seconds instead of 15 hours, this represents the future of their profession, arriving faster than most expected.
Bet on chatgpt when it was still hallucinating biographies, it has become a recognition that sometimes the riskiest move is not to move at all.