Within the winter of 2022, because the tech world was changing into mesmerized by the sudden, explosive arrival of OpenAI’s ChatGPT, Benjamin Alarie confronted a pivotal alternative. His authorized tech startup, Blue J, had a good enterprise constructed on the AI of a bygone period, serving tons of of accounting companies with predictive fashions. But it surely had hit a ceiling.
Alarie, a tenured tax regulation professor on the College of Toronto, noticed the nascent, error-prone, but highly effective capabilities of huge language fashions not as a curiosity, however as the longer term. He made a high-stakes resolution: to pivot his whole firm, which had been painstakingly constructed over almost a decade, and rebuild it from the bottom up on this unproven know-how.
That wager has paid off handsomely. Blue J has since quietly secured a $122 million Collection D funding spherical co-led by Oak HC/FT and Sapphire Ventures, putting the corporate's valuation at over $300 million. The transfer remodeled Blue J from a distinct segment participant into one among Canada's fastest-growing authorized tech companies, multiplying its income roughly twelve-fold and attracting 10 to fifteen new prospects day-after-day.
The corporate now serves greater than 3,500 organizations, together with international accounting big KPMG and several other Fortune 500 firms. It’s tackling a vital bottleneck within the skilled providers trade: a extreme and worsening expertise scarcity. The U.S. has 340,000 fewer accountants than it did 5 years in the past, and with 75% of present CPAs anticipated to retire within the subsequent decade, companies are determined for instruments that may amplify the productiveness of their remaining consultants.
“What as soon as took tax professionals 15 hours of guide analysis to do can now be accomplished in about 15 seconds with Blue J,” Alarie, the corporate's CEO, mentioned in an unique interview with VentureBeat. "That worth proposition—we are able to take hours of labor and switch it into seconds of labor—that’s driving lots of this."
When the dean's biography was improper: the second that modified the whole lot
Alarie vividly remembers January 2023, when the dean of the regulation college stopped by his workplace for New Yr's greetings. He requested her about ChatGPT and prompted the AI to explain her. ChatGPT confidently generated a biography. Some particulars had been correct. Others had been fully fabricated.
"She was like, 'Okay, that is actually sort of scary. That is improper, and this has implications,'" Alarie mentioned. But that second of apparent failure didn't deter him. As an alternative, it crystallized his conviction.
The corporate's first iteration, launched in 2015, used supervised machine studying to construct predictive fashions that might forecast judicial outcomes on particular tax points. Whereas technically subtle, it had a elementary flaw: it couldn't reply each tax analysis query.
"The problem was it couldn't reply each tax analysis query, which was actually the holy grail," Alarie mentioned. Clients beloved the software when it utilized to their drawback, however would shortly abandon it when it didn't. Income plateaued round $2 million yearly.
Regardless of ChatGPT's infamous hallucinations, Alarie satisfied his board to make the pivot. "I had this conviction that if we continued down that path, we weren't going to have the ability to deal with our primary limitation," he mentioned. "Massive language fashions appeared like a really promising course."
He gave his group six months to ship a working product.
From 90-second responses to three million queries: How Blue J tamed AI hallucinations
By August 2023, Blue J was able to launch. What they launched was, in Alarie's candid evaluation, "tremendous janky." The system took 90 seconds to reply. About half the solutions had points. The Internet Promoter Rating registered at simply 20.
What remodeled that flawed product into as we speak's platform — with response occasions measured in seconds, a dissatisfaction price of only one in 700 queries, and an NPS rating within the mid-80s — was relentless concentrate on three strategic pillars.
First is proprietary content material at large scale. Blue J secured unique licensing with Tax Analysts (Tax Notes) and IBFD, the Amsterdam-based international tax authority masking 220+ jurisdictions. "We’re the one platform on earth that takes in the most effective U.S. tax info from Tax Notes and the most effective international tax info from IBFD," Alarie mentioned.
Second is deep human experience. Blue J employs tax consultants led by Susan Massey, who spent 13 years on the IRS Workplace of Chief Counsel as Department Chief for Company Tax. Her group always assessments the AI and refines its efficiency.
Third is an unprecedented suggestions flywheel. With over 3 million tax analysis queries processed in 2025, Blue J is amassing unparalleled knowledge. Every question generates suggestions that flows again into the system.
Weekly lively consumer charges hover between 75% and 85%, in comparison with 15% to 25% for conventional platforms. "A charitable ratio is like we're 5 occasions extra intensively used," Alarie famous.
Inside Blue J's early entry partnership with OpenAI
Blue J maintains an unusually shut relationship with OpenAI that has confirmed essential to its success. "We’ve got an excellent relationship with OpenAI, and we get early entry to their fashions,"Alarie mentioned. "It's fairly collaborative. We give them lots of actually top quality suggestions about how effectively completely different variations of forthcoming fashions are performing."
This suggestions proves invaluable as a result of Blue J has developed what Alarie calls "ecologically legitimate" take a look at questions — drawn from precise tax skilled queries, with appropriate solutions decided by Blue J's skilled group. This helps OpenAI enhance efficiency on advanced reasoning duties.
The corporate assessments fashions from all main suppliers — OpenAI, Anthropic, Google's Gemini, and open-source options — repeatedly evaluating which performs finest. "We're not essentially 100% dedicated to any explicit supplier," he defined. "We're testing on a regular basis."
This method helps Blue J navigate a difficult enterprise mannequin: charging roughly $1,500 per seat yearly for limitless queries whereas absorbing variable compute prices. "We've pre-committed to delivering them a extremely good consumer expertise, limitless tax analysis solutions at a set value," Alarie mentioned. "We're absorbing lots of that danger."
Competitors amongst basis mannequin suppliers creates downward stress on API pricing, whereas Blue J's conservative utilization modeling has confirmed correct. Gross income retention exceeds 99%, whereas internet income retention reaches 130% — thought-about best-in-class for SaaS companies.
Taking up Thomson Reuters and LexisNexis with 75% weekly engagement
Blue J faces competitors from established publishers like Thomson Reuters, LexisNexis, and Bloomberg, all of which introduced AI capabilities all through 2023 and 2024. But Blue J's engagement metrics counsel it has captured important momentum, rising from simply 200 prospects in 2021 to over 3,500 organizations as we speak.
The every day updates show essential. Whereas the tax code itself modifications solely when Congress acts, the ecosystem evolves always via IRS laws, new rulings, and courtroom instances. All 50 states modify their tax codes often.
"Issues are altering actually day-after-day," Alarie mentioned. "Day by day we're updating the supplies, and that's simply the U.S. We cowl Canada, we cowl the UK. The aspirations are really international for this factor."
Alarie's ambitions prolong past constructing a profitable startup. As writer of the award-winning e-book "The Authorized Singularity" and college affiliate on the Vector Institute for Synthetic Intelligence, he has spent years considering AI's long-term affect on regulation.
In tutorial papers revealed in Tax Notes all through 2023 and 2024, he chronicled generative AI's rise, predicting that "shoppers will turn out to be considerably extra subtle" and that AI would push human consultants towards higher-value strategic roles fairly than routine analysis.
Blue J's $122 million plan: From tax analysis to 'international tax cognition'
The Collection D funding, which introduced complete capital raised to over $133 million, will gas aggressive geographic and product growth. Blue J already operates within the U.S., Canada, and the U.Okay., with plans to finally cowl 220+ jurisdictions via its IBFD partnership.
Future capabilities might embody automated memo era, tax type completion, doc drafting, and conversational historical past sustaining context throughout classes—remodeling Blue J from a analysis software into what Alarie describes as "the working layer for international tax cognition."
For all its success, Blue J operates in a site the place errors carry severe penalties. The hallucination drawback hasn't been eradicated — it's been minimized via cautious engineering, content material curation, and human oversight. Blue J has skilled its fashions to acknowledge after they can’t reply a query fairly than fabricate info.
The enterprise additionally faces financial dangers if compute prices spiral or utilization patterns exceed projections. And subtler questions loom about skilled judgment: as AI techniques turn out to be extra succesful, will customers defer to outputs with out enough vital analysis?
From 15 hours to fifteen seconds: What Blue J's AI pivot teaches each trade
Blue J's transformation provides classes past tax software program. The corporate's willingness to desert eight years of proprietary know-how and rebuild on an initially unreliable basis required each braveness and calculated risk-taking.
The choice paid off not as a result of generative AI was inherently superior to supervised machine studying in all dimensions, however as a result of it addressed the appropriate drawback: comprehensiveness fairly than precision in slender domains. Tax professionals didn't want 95% accuracy on 5% of questions. They wanted good-enough accuracy on 100% of questions.
The advance from an NPS of 20 to 84 in simply over two years displays relentless iteration knowledgeable by large knowledge assortment. The content material partnerships created differentiation that pure know-how couldn't replicate. The group of tax consultants offered area data vital to make sure reliability.
Most essentially, Blue J acknowledged that the true competitors wasn't different AI startups and even established publishers. It was the previous manner of doing issues — the 15 hours of guide analysis, the institutional data locked in retiring professionals' heads.
"Individuals are like, 'What does Blue J do? They supply higher tax solutions. Okay, I believe we want that,'" Alarie mirrored.
As AI transforms career after career, that readability of function might matter greater than technological sophistication. The long run belongs to not those that construct probably the most superior AI, however to those that most successfully harness it to unravel issues people even have.
For a tax regulation professor who began with frustration about inefficient analysis strategies, constructing a $300 million firm marks an audacious endpoint. For the 1000’s of execs now answering advanced questions in 15 seconds as an alternative of 15 hours, it represents the way forward for their career, arriving sooner than most anticipated.
The wager on ChatGPT when it was nonetheless hallucinating biographies has turn out to be a validation that generally the riskiest transfer is to not transfer in any respect.
