How Vertical AI Achieves Defensible Accuracy - with Steve Hasker of Thomson Reuters
The future of AI in high-stakes professions isn't about replacing humans—it's about building machines that can only operate under strict human oversight. Steve Hasker, CEO of Thomson Reuters, argues that general-purpose AI is dangerously inadequate for legal, tax, and audit work, where even a 5% error rate can lead to disbarment, fines, or jail time. What sets 'vertical AI' apart is not just industry-specific training, but a foundation of centuries-old curated legal content, 4,500 full-time domain experts who train AI agents step-by-step, and end-to-end audit trails that allow professionals to validate every output. Hasker reveals that AI tools like Westlaw Advantage and Ready to Review aren't just drafting assistants—they're 'driverless cars' for complex workflows, automating tedious data collection and first-draft generation while forcing human experts to focus on judgment, not grunt work. Yet, he draws a firm line: machines can never assume the legal or ethical responsibility of a licensed professional. The real breakthrough isn't automation—it's creating AI that behaves like a world-class expert, trained by one, and validated by another. The episode exposes a critical divide: AI for marketing or internal comms can afford to be 'mostly right,' but in fiduciary roles, accuracy must be 100%. This demands a new model of AI development—one where content, expertise, and traceability are non-negotiable.
Professional-grade AI in legal, tax, and audit work requires 100% accuracy, which only vertical AI built on centuries of curated content and expert training can deliver.
Thomson Reuters' AI agents are trained by 4,500 full-time domain experts who replicate real-world expert workflows step-by-step, ensuring outputs are defensible and traceable.
AI tools like Ready to Review automate the entire tax return process from data ingestion to first draft, reducing reliance on scarce human talent in a shrinking profession.
Every AI output must include an end-to-end audit trail showing source materials, reasoning steps, and validation paths—essential for legal and regulatory compliance.
Machines cannot assume the legal responsibility of a licensed professional; human judgment, framing, and accountability must remain non-negotiable in fiduciary workflows.
…and 3 more takeaways available in PodZeus
Introduction to Vertical AI and Fiduciary AI
Host Daniel Fagella introduces Steve Hasker, CEO of Thomson Reuters, to discuss the critical distinction between general-purpose AI and vertical AI—especially in high-stakes professions like law, tax, and audit where accuracy is non-negotiable.
Defining Vertical AI: Content, Expertise, and Professional Standards
“And so I think that's the definition in our view of vertical AI and that sort of nature of industry specific. I think there's an additional layer on top of that, which is for want of a better term, professional grade AI for fiduciary professions.”
The Power of Proprietary Data and Expert Trained Agents
“Our experts in that particular form of the law and that particular type of transaction will train the agent to behave like a world-class expert at every one of the 30 steps.”
Why 95% Accuracy Isn't Good Enough in Legal and Tax Work
“The bad news is you're going to jail. Because, you know, and it's sort of a somewhat humorous example, I suppose. But I used the model that was 95% correct. And the 5% included some very audacious, you know, sort of amortization assumptions.”
Case Study: Westlaw Advantage and Co-Counsel as a Legal Partner
Hasker explains how Westlaw Advantage and Co-Counsel act as AI co-counsel, simulating the iterative process of a senior litigator reviewing arguments with partners—reducing hours of manual work while maintaining 100% traceability.
“Well, the bad news is you're going to jail. Because, you know, and it's sort of a somewhat humorous example, I suppose. But I used the model that was 95 correct. And the 5 included some very audacious, you know, sort of amortization assumptions.”
“In our view, a machine cannot assume the rights and responsibilities of a fully qualified practicing attorney or a fully qualified practicing CPA. They cannot, a machine cannot and should not assume.”
“We think will fill that gap in a really powerful way by combining the tax calculation engine, the content that we have and the constant updating of the content with the AI agents to basically pull together the entire system not unlike a driverless car.”
Host
Guest
Thomson Reuters
organization
Steve Hasker
person
Westlaw
product
Ready to Review
product
Co-Counsel
product
Emerge
organization
Noetica
organization
IRS
organization
SEC
organization
AI Models as a Commodity and Why Data Foundations Decide Who Wins - with Guillermo B. Vazquez of SAP
21m • 6/9/2026
Fixing the Decision Speed Gap in Modern Supply Chains - with Joris Wijpkema of Optilogic
27m • 6/15/2026
The Pricing Shift Reshaping Enterprise AI Spend - with Adam Mansfield of UpperEdge
37m • 6/1/2026
Human-Centered AI Development Strategies for CPG Leaders - with Shaje Ganny of Procter & Gamble
29m • 6/2/2026
How Industrial Service Leaders Are Closing the Knowledge Gap Before It's Too Late with Mike Hughes of Peak International Group
26m • 6/4/2026
Start discovering podcast insights today
Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.
No credit card required • 7-day trial • Cancel anytime