Capstone & the Future Lawyer: Becoming the Human-in-the-Loop
Ship a verified AI-assisted work product, defend your process, and own your place in the loop.
The hook
“The model's confidence is not evidence. The lawyer who forgets that files fabricated case law; the lawyer who remembers it becomes irreplaceable. This module is where you prove which one you are.”
What you'll be able to do
- Synthesize the whole course into a single discipline: never ship an unverified AI answer — the model's confidence is not evidence.
- Present a capstone: a verified AI-assisted legal work product accompanied by a documented verification trail and a reflective account of process and ethics.
- Situate AI within professional identity and the future of legal work — the 'centaur' lawyer who pairs human judgment with machine speed.
- Demonstrate, by re-running the Module 1 demo, that the cohort's own behaviour has changed — an evidence-based, self-referential test of the duty to verify.
In short
The capstone module pulls the eight-module arc into one professional commitment: the human stays in the loop, and an AI answer is never shipped until it is verified to source. Students present a verified AI-assisted work product with a full verification trail and a reflective essay, then debate the future of the profession — Susskind's changing job and Fry's centaur — before a self-referential finale that re-runs Module 1's priming/decoding demo to show the room how its own judgment has changed.
The AI bridge
The whole course converges here: using AI effectively and responsibly means being a disciplined human-in-the-loop. The capstone is the proof — a real AI-assisted product shipped only after every citation and proposition is verified to source, with the trail and the reasoning made visible. Mastery is not faster output; it is the judgment to interrogate the model, the verification to catch its fabrications, and the ethical literacy to know where law forbids relying on it.
In this module
- 01
Capstone presentations: each student defends a verified AI-assisted legal work product (e.g. a research memo or contract), the verification trail that backs every citation and proposition, and a reflection on process and ethics — operationalizing the duty to check.
- 02
The course synthesis, stated as a spine: confidence-vs-correctness. A fluent, assured AI answer is a statement costumed as proof; the lawyer's job is to climb the Ladder of Misinference on it before relying on it.
- 03
The single non-negotiable takeaway: never ship an unverified AI answer — the model's confidence is not evidence, and the human-in-the-loop owns the integrity of the work product.
- 04
The future of the profession: Susskind on how the legal job is changing (Tomorrow's Lawyers; online courts and the future of justice) — routine work gets automated, judgment and verification become the premium skills.
- 05
The centaur lawyer (Fry's centaur chess): the strongest performance comes not from human alone or machine alone, but from a human pairing judgment with machine speed and refusing to abdicate the decision.
- 06
Professional identity and ethics, integrated: competence (including technological competence), candour to the tribunal, confidentiality (no privileged data into public LLMs), supervision, and disclosure of AI use are the working norms of the AI-era lawyer.
- 07
The self-referential closer: re-run the Module 1 priming/decoding demo and show the cohort's behaviour has changed — evidence, gathered on the room itself, that verification has become a habit rather than a slogan.
The interactive demos
Every idea is a Mirror Move
Run it on the room, show it inside the machine, prove it live on a real AI, then name the skill.
The self-referential finale — re-run Module 1's priming/decoding demo
On us
Re-run the Module 1 priming/anchoring demo on the same room: split the cohort and prime with a low vs. high anchor before a legal estimation, then reveal the two averages — testing whether the cohort still gets pulled by the anchor the way it did in Module 1.
In the machine
LLMs remain just as suggestible to a number or framing smuggled into a prompt — the same effect the room is now alert to in itself.
Live AI
Optionally show, once more, an AI answer delivered with full confidence, and have the trained room reflexively start climbing the Ladder / demanding the source rather than accepting it.
The skill
An evidence-based finale that mirrors the duty of verification: the cohort's changed behaviour is the proof that 'the model's confidence is not evidence — never ship an unverified AI answer' has become a habit, not a slogan.
The lab
Capstone — the verified AI-assisted work product
Each student presents and defends a capstone: a verified AI-assisted legal work product (e.g. a research memo or a contract) built with AI but shipped only after verification. The presentation must show the work product, the verification trail (every citation and proposition checked to source — which tool was used, what was checked, what was rejected), and a reflective account of the student's process and the ethics of their AI use. This is the 30% summative assessment and the synthesis of all eight modules.
Deliverable
A verified AI-assisted legal work product + a documented verification trail + a reflective essay on the student's process and the ethics of AI use, presented to the cohort.
Key sources & cases
Richard Susskind, Tomorrow's Lawyers
Anchors the 'future of the profession' content: how the legal job is changing as routine work is automated and judgment becomes the premium skill.
Richard Susskind, Online Courts and the Future of Justice
Extends the future-of-the-profession discussion to dispute resolution and access to justice; frames AI in the system, not just the firm.
Fry's centaur chess
Source of the 'centaur' lawyer image — human judgment paired with machine speed outperforms either alone; the model for the human-in-the-loop.
The course's own cases
Module 8 reuses the cautionary-case spine (e.g. Mata v. Avianca; Gummadi Usha Rani; the Bombay HC cost order) as the lived stakes behind the capstone's verification discipline — used exactly as the earlier modules describe them.
Readings
- Richard Susskind, Tomorrow's Lawyers
- Richard Susskind, Online Courts and the Future of Justice
- Hannah Fry, Hello World (2018) — the centaur idea
- Frank Pasquale, New Laws of Robotics (2020)
- The course's own cautionary-case spine (Mata v. Avianca; Gummadi Usha Rani; the Bombay HC ₹50,000 cost order)
The arc is complete
Modules 01–08
Back to all eight modules
Revisit the courseBring a credit-bearing AI course to your students
A 16-hour course that treats using AI well as a professional duty — one a council can approve, and a graduate can defend in court.