Within the winter of 2022, because the tech world was turning into mesmerized by the sudden, explosive arrival of OpenAI’s ChatGPT, Benjamin Alarie confronted a pivotal selection. His authorized tech startup, Blue J, had a decent enterprise constructed on the AI of a bygone period, serving tons of of accounting corporations with predictive fashions. But it surely had hit a ceiling.
Alarie, a tenured tax legislation 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 long run. He made a high-stakes choice: to pivot his complete firm, which had been painstakingly constructed over almost a decade, and rebuild it from the bottom up on this unproven know-how.
That guess has paid off handsomely. Blue J has since quietly secured a $122 million Sequence 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 corporations, multiplying its income roughly twelve-fold and attracting 10 to fifteen new clients each day.
The corporate now serves greater than 3,500 organizations, together with international accounting big KPMG and a number of other Fortune 500 corporations. It’s tackling a important bottleneck within the skilled companies business: 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, corporations are determined for instruments that may amplify the productiveness of their remaining specialists.
“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 will take hours of labor and switch it into seconds of labor—that’s driving a number of this."
When the dean's biography was mistaken: the second that modified all the things
Alarie vividly remembers January 2023, when the dean of the legislation college stopped by his workplace for New 12 months's greetings. He requested her about ChatGPT and prompted the AI to explain her. ChatGPT confidently generated a biography. Some particulars have been correct. Others have been utterly fabricated.
"She was like, 'Okay, that is actually sort of scary. That is mistaken, and this has implications,'" Alarie mentioned. But that second of apparent failure didn't deter him. As a substitute, it crystallized his conviction.
The corporate's first iteration, launched in 2015, used supervised machine studying to construct predictive fashions that would forecast judicial outcomes on particular tax points. Whereas technically subtle, it had a basic 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. Prospects cherished the instrument when it utilized to their downside, however would rapidly 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. "Giant language fashions appeared like a really promising course."
He gave his workforce 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 at present's platform — with response instances measured in seconds, a dissatisfaction charge of only one in 700 queries, and an NPS rating within the mid-80s — was relentless deal with 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 perfect U.S. tax info from Tax Notes and the perfect international tax info from IBFD," Alarie mentioned.
Second is deep human experience. Blue J employs tax specialists led by Susan Massey, who spent 13 years on the IRS Workplace of Chief Counsel as Department Chief for Company Tax. Her workforce continuously exams 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 energetic person charges hover between 75% and 85%, in comparison with 15% to 25% for conventional platforms. "A charitable ratio is like we're 5 instances 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 have now an excellent relationship with OpenAI, and we get early entry to their fashions,"Alarie mentioned. "It's fairly collaborative. We give them a number of actually top quality suggestions about how effectively totally different variations of forthcoming fashions are performing."
This suggestions proves precious as a result of Blue J has developed what Alarie calls "ecologically legitimate" check questions — drawn from precise tax skilled queries, with appropriate solutions decided by Blue J's professional workforce. This helps OpenAI enhance efficiency on complicated reasoning duties.
The corporate exams fashions from all main suppliers — OpenAI, Anthropic, Google's Gemini, and open-source options — constantly evaluating which performs greatest. "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 person expertise, limitless tax analysis solutions at a hard and fast worth," Alarie mentioned. "We're absorbing a number of that threat."
Competitors amongst basis mannequin suppliers creates downward strain on API pricing, whereas Blue J's conservative utilization modeling has confirmed correct. Gross income retention exceeds 99%, whereas web 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 recommend it has captured important momentum, rising from simply 200 clients in 2021 to over 3,500 organizations at present.
The day by day updates show essential. Whereas the tax code itself adjustments solely when Congress acts, the ecosystem evolves continuously by way of IRS rules, new rulings, and court docket circumstances. All 50 states modify their tax codes repeatedly.
"Issues are altering actually each day," Alarie mentioned. "Every single 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 lengthen past constructing a profitable startup. As writer of the award-winning e book "The Authorized Singularity" and school affiliate on the Vector Institute for Synthetic Intelligence, he has spent years considering AI's long-term impression on legislation.
In educational papers printed in Tax Notes all through 2023 and 2024, he chronicled generative AI's rise, predicting that "shoppers will change into considerably extra subtle" and that AI would push human specialists towards higher-value strategic roles quite than routine analysis.
Blue J's $122 million plan: From tax analysis to 'international tax cognition'
The Sequence D funding, which introduced whole capital raised to over $133 million, will gasoline aggressive geographic and product growth. Blue J already operates within the U.S., Canada, and the U.Ok., with plans to finally cowl 220+ jurisdictions by way of its IBFD partnership.
Future capabilities may embody automated memo technology, tax kind completion, doc drafting, and conversational historical past sustaining context throughout classes—reworking Blue J from a analysis instrument 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 critical penalties. The hallucination downside hasn't been eradicated — it's been minimized by way of cautious engineering, content material curation, and human oversight. Blue J has skilled its fashions to acknowledge once they can’t reply a query quite 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 programs change into extra succesful, will customers defer to outputs with out ample important analysis?
From 15 hours to fifteen seconds: What Blue J's AI pivot teaches each business
Blue J's transformation gives 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 best downside: comprehensiveness quite than precision in slim 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 workforce of tax specialists supplied area data needed to make sure reliability.
Most basically, Blue J acknowledged that the true competitors wasn't different AI startups and even established publishers. It was the previous approach 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 goal could matter greater than technological sophistication. The long run belongs to not those that construct essentially the most superior AI, however to those that most successfully harness it to resolve issues people even have.
For a tax legislation professor who began with frustration about inefficient analysis strategies, constructing a $300 million firm marks an audacious endpoint. For the hundreds of pros now answering complicated questions in 15 seconds as a substitute of 15 hours, it represents the way forward for their career, arriving sooner than most anticipated.
The guess on ChatGPT when it was nonetheless hallucinating biographies has change into a validation that generally the riskiest transfer is to not transfer in any respect.
[/gpt3]
