Benjamin Alarie joins hosts Nathan Kiker and George Bogdan to examine how rapidly improving AI systems can strengthen the administration of law—starting with tax, where complexity, volume, and inconsistency have long constrained access to justice. Alarie argues that legal AI is transitioning from promise to performance, producing measurably more accurate and comprehensive research while enabling the kind of population-level analysis that policymakers and practitioners have struggled to execute at scale. The conversation moves beyond abstract fairness debates to operational impact. Drawing on a decade of building tools at BlueJ Legal, Alarie underscores a sharp decline in user disagreement with AI-generated answers as models and workflows have matured. He details how adversarial research techniques—prompting strongest counterarguments and “braiding” outputs from multiple frontier models—paired with strict verification against primary sources, drive better outcomes. The result is faster, deeper, and more consistent analysis across statutes, cases, and administrative guidance. A core strategic theme is harmonization. Legal AI can help reconcile fragmented regimes—tax versus employment standards, antitrust versus trade policy, or Quebec’s civil law alongside common-law provinces—by mapping doctrine to practical outcomes and integrating empirical evidence. That capability enables more consistent enforcement, clearer incentives for firms, and a stronger foundation for preventative governance rather than purely reactive adjudication. Alarie is clear on limits: AI augments, not replaces, professional judgment. Accountability remains with human experts, and progress in tools must be matched by attention to the digital divide—where paper filings, limited connectivity, and uneven digital literacy still impede equitable access. For policymakers and leaders, the stakes are systemic: modernization of legal research and administration can reduce friction in markets, enhance compliance, and promote rule-of-law consistency across jurisdictions. Allery’s new book, Superjustice: The Future of Law, outlines these pathways in depth. Key Topics Discussed - Performance over theory: why disagreement with AI-generated legal answers is falling and what that signals about maturity in real-world research workflows. - Harmonizing fragmented regimes: using AI to surface cross-jurisdictional patterns, reconcile conflicting doctrines, and align legal incentives with policy goals. - Adversarial AI research: stress-testing conclusions by eliciting strongest counterarguments, braiding multiple model outputs, and verifying against primary sources. - From cases to populations: enabling empirical, population-level analyses of legal outcomes that inform better policy design and preventative governance. - Access and equity: the digital divide, persistent reliance on paper processes, and why infrastructure and adoption dynamics will shape AI’s impact on justice. About the Guest Benjamin Alarie is the CEO of BlueJ Legal and co-author of Superjustice: The Future of Law. He has spent the past decade building AI tools for tax and related legal research, bringing a practitioner’s mindset to model design, verification, and workflow integration. His experience positions him to assess where legal AI delivers real value, how to measure it, and what it will take to translate technical progress into fairer, more consistent legal outcomes. Timestamps - 00:00 — Framing the opportunity: why tax is a proving ground for legal AI - 06:20 — Measuring performance: accuracy, coverage, and declining user disagreement - 14:05 — Harmonization use cases across jurisdictions and practice areas - 22:30 — Adversarial techniques and multi-model “braiding” for resilient answers - 30:15 — Empirical synthesis: population-level insights for policy and compliance - 38:10 — Professional accountability and verification against primary sources - 44:55 — Adoption dynamics: legislative, judicial, and organizational factors - 51:40 — Digital divide, paper processes, and access-to-justice implications - 57:10 — Closing reflections and resources (Superjustice at superjustice.com) Support the show