Harvard Law School shield AI & the Legal ProfessionCenter on the Legal ProfessionHarvard Law School
Read the series
The seminar Harvard Law School · Fall 2026 · 2 credits

AI, Complex Decision-Making and the Future of the Legal Profession

A seminar taught by David B. Wilkins and Anthea Roberts, running 14 September to 1 October 2026. It asks what AI is doing to legal work, legal organizations and the formation of lawyers — and how anyone could know whether the leading claims are true. This series is the research program the seminar sits inside.

Catalog entry & enrollment↗ Read the series→
The seminar seated around a long conference table at OpenAI's New York headquarters, with further participants joining on the wall screen by Zoom.
17 Sep. Session 4 at OpenAI, New York — in the room and by Zoom.
Eight people, lanyards on, in front of a backdrop reading “OpenAI for Law Summit”.
17 Sep. The OpenAI for Law Summit, later that afternoon.
Twelve sessions · three weeks

The course map

Week oneThe structure and the mechanism
01
Mon 14 Sep
Introduction: Two Predictions and a Third
Course · field
02
Tue 15 Sep
Mapping the AI and Legal Landscape
Field · history
03
Wed 16 Sep
Hands-On with AI Tools and the Corpus
Method · practice
04
Thu 17 Sep
OpenAI, New York: The Model Layer Enters Law
Technology · field
Guests Jason Boehmig, head of the legal vertical, OpenAI; Stacey Wang.
Week twoOrganizations competing to deliver legal work
05
Mon 21 Sep
Application Companies: Harvey, Legora and the Vertical Layer
Organization
Guests Winston Weinberg, co-founder and CEO, Harvey; Tracey Brady Yurko, Chief Legal Officer, Bridgewater Associates; Isabel Yishu Yang, micro1.
06
Tue 22 Sep
GC 4.0: The Client as System Architect
Client · organization
Guest Bjarne P. Tellmann, formerly General Counsel, Haleon; author of Law in the Era of AI.
07
Wed 23 Sep
Incumbent Law Firms: Strategy, Economics and the Pyramid
Organization
Guest Mitch Zuklie, Chairman and Chief Executive Officer, Orrick.
08
Thu 24 Sep
AI-Native Law Firms: Selling the Work, Not the Tool
Organization · governance
Guests Zach Posner, co-founder and managing partner, The LegalTech Fund; Hariolf Wenzler, Bucerius Center on the Legal Profession, on managed services organizations.
Two people seated at a seminar-room table beside a window, with Winston Weinberg of Harvey filling the wall screen on a video call.
21 Sep. Session 5: Tracey Brady Yurko and David Wilkins in the room, Winston Weinberg of Harvey joining by video.
Bjarne Tellmann speaking beside a projected slide headed “Dragonfly lawyering”, illustrated with a dragonfly's compound eyes.
22 Sep. Bjarne Tellmann on GC 4.0, and what he calls dragonfly lawyering.
Week threeDistribution, judgment, formation and futures
09
Mon 28 Sep
Professional Judgment and the Diamond-Complete Lawyer
Individual
10
Tue 29 Sep
Where Do We Go from Here? Actors, Systems and Scenarios
Synthesis · strategy
Guest Ryan Daniels, Crosby.
11
Wed 30 Sep
Distribution: Entry-Level Opportunity and Access to Justice
Distribution · evidence
Guests Damien Charlotin, Sciences Po Law School, on hallucinated citations in court filings; Gerald Abila, founder and chief executive, BarefootLaw.
12
Thu 1 Oct
The Formation Problem: Apprenticeship and Legal Education
Formation
Guest Andrew Perlman, Dean, Suffolk University Law School.
The evidence base

Students work the same record this site is built on

The seminar is taught against a live corpus rather than a reading list frozen when the syllabus was written.

Students receive access to Second Chair, the living research corpus behind this series — scholarship, primary legal materials, industry evidence and practitioner interviews, created by Anthea Roberts and David B. Wilkins as part of this project, structured and analyzed by Insight Bridge. It can be searched, compared and traced from a synthesis back to the underlying source, and connected to a compatible AI tool through its read-only MCP connector.

The course also draws on the Dragonfly field analysis — the actor map, the drivers and systems map, and the scenarios. Those are analytical interpretations of the evidence base rather than additional evidence, and students are asked to test their assumptions and causal claims against the underlying sources. Any proposition used in assessed work must still be checked against the underlying document: the corpus is research infrastructure, not an authority in its own right.

Second ChairThe corpus: 1,501 sources, 369 topic clusters, 1949–2026. The explorerEvery cluster, what it argues, and the sources behind it. The lensesThe topic map and the actor map — the field analysis. MCP connectorConnect your own agent to the corpus, read-only.
Learning outcomes

What students should be able to do

The seminar is career-facing as well as analytical: by the end, students should be equipped to evaluate not only a technology but a prospective employer’s strategy, training system, economic model and approach to professional responsibility.

  • distinguish task-level capability and productivity evidence from evidence about jobs, organizations and markets;
  • explain why generation and verification may move at different rates, and apply the verification-frontier framework to a legal matter;
  • compare application companies, AI-native law firms, incumbent firms and AI-enabled legal departments by what they sell, what capital they require and who remains answerable;
  • evaluate how AI may redistribute access to entry-level professional opportunities and access to legal assistance, without collapsing those distinct questions;
  • weigh claims from scholarship, company announcements, founder interviews, benchmarks, professional rules and labor-market data according to what each source can establish;
  • describe the domain depth, breadth, practical AI fluency and orchestration involved in the diamond-complete ideal of professional excellence;
  • distinguish preserving junior tasks from preserving the developmental functions of apprenticeship;
  • use actor maps, system-dynamics maps and scenarios to identify feedback effects, signposts and robust institutional responses; and
  • develop an original, falsifiable argument using AI transparently and with source-level verification.
Enrollment & requirements

This page is the seminar as it sits inside the research program. Registration, prerequisites, credit and the formal requirements are held by the Law School: the Harvard Law School catalog entry ↗. The program it belongs to is the AI and the Legal Profession project ↗ at the Center on the Legal Profession.