Download

Mar 23, 2026

by Gregory M. Duhl*

 

I. Finding the Right Tool for the Job

    Teaching contract drafting in a first-year Contracts course has long presented significant challenges: limited class time; students’ unfamiliarity with transactional practice; specialized vocabulary; and minimal exposure in the first-year legal writing, research, and analysis course to contract drafting and revision. When I resolved to redesign the course around AI integration, I sought a more effective approach than requiring students to draft clauses from scratch. I needed a legal AI platform that students could partner¹ with or supervise to generate contracts, propose edits and explain them, evaluate risk and compliance, revise agreements, and negotiate terms. In other words, I was looking for a platform mirroring the workflow of contemporary contract practice.

    This initiative was grounded in current professional realities. AI adoption in contract work is accelerating dramatically. Among large law firms with five hundred or more attorneys, reported usage ranges from 48 to 100 percent, depending on survey methodology.² Individual attorney adoption has risen to between 55 and 81 percent, with in-house counsel at the forefront.³ The legal profession has progressed beyond debating whether to adopt AI for contract work; the focus now is on selecting appropriate tools and employing them effectively and efficiently. Because students will inevitably practice in a profession transformed by AI, they must learn to work with AI technologies in law school, a controlled setting where they can safely experiment, make mistakes, and refine their oversight and judgment.

    For these reasons, I evaluated several transactional AI legal tools, consulted colleagues teaching dealmaking and experimenting with AI in their courses, and carefully considered capabilities aligned with my pedagogical goals. Because the sample size of professors using professional contract drafting applications was small, I also searched social media for testimonials from professors, students, and lawyers experienced with these tools. My goal was to incorporate into my course a platform that would help students acquire essential transactional skills for practice in an AI-transformed profession.  

    This evaluation led me to Spellbook. I was especially drawn to this AI platform because it was specifically designed for transactional lawyers to draft, review, edit, and negotiate contracts, with seamless integration into Microsoft Word.⁴ Most compelling was its support for the full drafting and negotiation workflow, encompassing contract creation, compliance, and negotiation.⁵ Upon discovering its free educational partnerships for law schools, I contacted two professors who had incorporated Spellbook in their courses; their feedback was positive. In collaboration with my colleague Amanda Soderlind, Director of Instructional Design and Development at Mitchell Hamline School of Law (MHSL), who guided the AI integration throughout my course redesign, we conducted further research and were convinced to adopt the platform after a demonstration by the company’s partnerships manager, Dev Thain.

    This spring, I am teaching a fully redesigned, AI-embedded first-year Contracts course. The semester’s second half features Spellbook simulations that progressively build from basic contract review to complex negotiation and drafting. Building part of the Contracts course around Spellbook is not merely an isolated case study in AI use in legal education; it offers a transferable pedagogical model to other AI tools and courses. Before describing this pedagogy, however, clarity is needed on AI’s capabilities and limitations in contract drafting and transactional practice, and why these matter for legal education.

II. The State of AI in Contract Drafting

    The legal profession has widely adopted AI for contract work, yielding striking results. JPMorgan Chase’s Contract Intelligence program reduces annual review time for commercial loan agreements from 360,000 hours to mere seconds.⁶ A LawGeex study found that its AI tool achieved 94 percent accuracy in identifying risks in nondisclosure agreements, compared to 85 percent for experienced human lawyers.⁷ A comprehensive 2025 benchmark study comparing thirteen AI applications with human lawyers determined that Gemini 2.5 Pro, the top-performing AI platform, shows a 73.3 percent reliability rate in contract drafting, slightly outperforming the best human lawyer’s 70 percent.⁸ Most tellingly, specialized legal AI tools have detected material risks that human lawyers overlooked entirely. In tasks involving potentially unenforceable penalty clauses, AI flagged enforceability concerns, whereas human lawyers provided no risk assessment.⁹

    However, research in this field also reveals limitations: Although AI excels at pattern recognition, consistency checks, and identifying standard risks, it struggles with commercial judgment, interpreting nuanced client intent, and synthesizing multiple information sources in context-dependent scenarios. In a multisource drafting task, all AI platforms produced outputs with incomplete or inaccurate information, whereas human lawyers successfully extracted complete party information from a provided screenshot.¹⁰

    From these findings, we can conclude that AI tools are transforming contract practice by managing routine tasks with exceptional efficiency, but they cannot replace the strategic thinking, ethical judgment, and contextual understanding essential to competent lawyering. As Professor Gillian K. Hadfield anticipated in 2018, AI in contract work best augments practitioners by enabling them to redirect attention “from routine activities to much more high value work involved in shaping strategies and navigating complex legal problems.”¹¹

    This is precisely what I aim to teach my students: how to employ AI to enhance their capabilities while developing the practical judgment that AI cannot replicate. The core pedagogical objective is for students to learn how to partner with or supervise AI.

III. What Spellbook Does Well

    Spellbook integrates directly into Microsoft Word, a feature of greater significance than it may appear.¹² With this integration, students do not need to learn a separate interface or toggle between applications; they operate within a familiar environment. The tool provides several key capabilities aligning closely with professional workflows¹³:

•  
• •  Contract Review and Analysis: Spellbook analyzes documents against established standards, identifies potential issues, and generates redlined versions.¹⁴ It highlights specific clauses requiring attention, such as indemnification, renewal, liability limitations, and choice-of-law provisions, and proactively suggests the potential consequences of specific clauses.¹⁵ For example, it may highlight a one-sided indemnification clause and explain associated risk exposure.¹⁶ The tool also benchmarks terms against industry standards, helping users identify atypical provisions.¹⁷
•  
• •  Drafting and Revision: Spellbook enables students to draft new clauses or entire contracts from scratch.¹⁸ It includes a clause library with standard boilerplate language, and users can upload custom libraries to incorporate organizational language.¹⁹ This functionality is particularly valuable pedagogically: Students observe AI’s handling of routine drafting and then practice refining and customizing its output. Spellbook also suggests revisions to existing drafts and generates summary emails outlining key changes for clients.²⁰
•  
• •  Negotiation Support: The tool offers features tailored to contract negotiation, including searches for compliance issues, preparation of suggested changes in emails to opposing counsel, and support for iterative negotiations within the platform.²¹ Students can use the “Ask” feature to obtain real-time strategic guidance on specific provisions during negotiations.²²
•  
• •  Security: Spellbook is SOC 2 Type II-compliant with encrypted data,²³ a critical consideration for law schools concerned about student privacy and data security in AI platforms.
•  

For teaching purposes, these capabilities enable students to engage with the authentic workflow of modern transactional practice. Rather than merely reading about contracts and disputes or drafting hypothetical clauses, they actively draft, revise, and negotiate alongside AI as a collaborative partner, mirroring their future professional experience.

IV. Building Pedagogy Around Professional Tools

    Spellbook provides foundational assignments and onboarding training videos tailored for law school use.²⁴ However, transforming a professional AI tool into effective pedagogy requires dedicated instructional design expertise, which most law schools and AI providers lack.

    MHSL successfully obtained American Bar Association (ABA) approval for and launched the nation’s first hybrid JD program at an ABA-accredited law school in 2015, largely because of MHSL’s deliberate investment in in-house instructional design expertise.²⁵ This foundational lesson—that pedagogical experts who understand the institutional mission must drive major curriculum shifts—has proven equally essential to the Contracts-redesign initiative.

    My partnership with Amanda Soderlind anchors the transformation of Spellbook from a professional tool into a cornerstone of law school pedagogy. Her role extends beyond technical or logistical support; it is fundamentally pedagogical. As Amanda explains:

 

Instructional designers can serve as the bridge between technology and pedagogy ... helping shift the mindset from “How can I use AI in my course?” to more focused questions such as “What new legal reasoning skills or ethical perspectives can students develop because of these tools?” That shift requires viewing AI through the lens of educational outcomes and considering how it can foster critical thinking and reflection rather than basic convenience or automation.²⁶

 

In our Spellbook integration, Amanda ensured we introduced students not to a shiny, new tool that performs their work for them, but to one with which they can partner in a way that pedagogically mirrors professional practice.

    Amanda insists on using Spellbook with rigor and sound pedagogy. We are not adopting the tool merely for automation or efficiency, although those are beneficial byproducts. Instead, every Spellbook-based activity or assignment is designed to promote critical thinking and reflection, preserving the human values central to transactional lawyering (that is, judgment, strategy, ethics, and client interests) and to legal practice more broadly.

    When redesigning the course, particularly the second half, in which students develop proficiency in transactional lawyering, we employed an outcomes-based course design, commonly known as “backward design.” We began by identifying desired learning outcomes, determining how to assess whether students achieved those outcomes, and designing aligned learning activities. This method, systematically articulated by education scholars Grant Wiggins and Jay McTighe in their influential work Understanding by Design, reverses traditional planning and ensures coherence and intentionality throughout the course.²⁷ A critical course outcome is AI literacy: Students must demonstrate their ability to understand, effectively use, and critically evaluate Spellbook as a professional tool.

    Within this framework, we developed a comprehensive Spellbook-centered curriculum for the second half of the course, comprising four major assignments:

•  
• •  Week 8: AI Review and Redlining: Students employ Spellbook’s “Benchmark” feature to review a nondisclosure agreement against market standards to identify risks and atypical provisions. Spellbook provided the foundational concept for this assignment, which we transformed into a fully developed pedagogical exercise with explicit learning outcomes and assessment rubrics. The task emphasizes issue-spotting by requiring students to critique the application’s audit, explain their understanding of its suggestions, and articulate why certain provisions pose risks.
•  
• •  Week 10: AI-Assisted Negotiation: Students represent a client negotiating a Master Services Agreement against opposing counsel. They use Spellbook’s “Negotiate” feature to audit the counterparty’s draft for risks and compliance concerns, then conduct the negotiation with real-time strategic support from the platform. Spellbook provided the initial assignment concept, which we expanded with structured negotiation parameters, assessment criteria distinguishing effective tool use from independent strategic judgment, and reflective components requiring students to articulate their justification for accepting or rejecting Spellbook’s suggestions. This routine simulates professional practice while enabling students to test negotiation strategies in a low-stakes environment.
•  
• •  Week 12: Complex Contract Drafting: Students draft a comprehensive employment agreement based on detailed client requirements involving multiple, sometimes conflicting, objectives. They employ Spellbook to generate initial drafts, then demonstrate improvements over the AI output using “Track Changes.” The assessment emphasizes students’ revisions instead of AI’s initial draft, reflecting their judgment about where AI was inaccurate or incomplete. This assignment structure, developed collaboratively with Amanda, ensures students learn the supervisory skills essential for AI-augmented practice.
•  
• •  Week 14: Summative Negotiation Project: Students work in pairs to negotiate a resolution to a commercial landlord-tenant dispute, with each pair member representing opposing sides. They draft, revise, and finalize an agreement, which serves as their capstone demonstration of strategic AI use alongside their independent professional judgment. This summative project, created in collaboration with Amanda, synthesizes all skills developed throughout the Spellbook sequence.
•  

Each assignment includes detailed rubrics explicitly distinguishing tool proficiency from legal reasoning. Students are assessed on their ability to determine when AI suggestions are appropriate, require modification, or should be rejected entirely. The rubrics emphasize that grading targets the student’s judgment regarding the AI output, not the output itself.

V. The Broader Lesson: AI Tools Need Pedagogical Frameworks

    Our work at MHSL highlights a decisive advantage: in-house instructional design expertise in both pedagogy and legal education. Instead of simply licensing an application and hoping for effective use, we built a deliberate pedagogical framework around the tool, designing learning experiences that employ AI to develop skills it cannot replicate.

    This experience reveals a significant gap in legal education’s AI adoption: The AI transformation is overtaking law schools because it has already reshaped the legal profession. Law school may be the only stage in a lawyer’s career when they can learn to use AI platforms appropriately in a supervised environment. In this environment, mistakes yield learning rather than pose malpractice risks. Nevertheless, licensing AI applications without robust pedagogical frameworks is unproductive. Tools without pedagogy produce graduates who can operate software but cannot exercise professional judgment.

    Most law schools lack dedicated instructional design teams. Most AI vendors do not provide pedagogical frameworks because they are product providers rather than educators. This discrepancy creates a significant disconnect: Schools seek AI integration to align with professional demands and vendors aim to sell licenses and introduce students to their platforms, yet few are constructing the vital bridge between professional tools and educational outcomes.

    The solution requires collaboration between legal educators and instructional designers or other pedagogical experts to accomplish the following:

•  
• •  Create scaffolded learning progressions integrating AI platforms with doctrinal learning.
• •  Develop assessment rubrics separating tool proficiency from analytical reasoning.
• •  Design assignments requiring human judgment, even as AI manages routine tasks.
• •  Establish explicit learning outcomes linked to both doctrine and professional competencies.
•  

    These imperatives extend beyond Spellbook. Any AI application integrated into legal education requires similar pedagogical scaffolding. Whether a contract-drafting platform, legal research assistant, document review tool, or litigation support system, the principle remains constant: The platform provides capabilities, but educators must craft learning experiences that convert those capabilities into enduring competencies.

VI. Why This Matters

    I am enthusiastic about incorporating Spellbook into my redesigned first-year Contracts course. Its capabilities align well with contemporary contract practice, enabling students to learn actual professional workflows. The free educational partnership eliminates financial barriers for institutions like MHSL, which serve diverse student populations.²⁸ Similarly, seamless integration with Microsoft Word allows students to operate in a familiar environment rather than learning yet another specialized interface.²⁹

    Most importantly, Spellbook provides a strong foundation for developing the practice-ready skills students will require. They will master the following:

•  
• •  Using AI tools efficiently while developing the judgment to supervise them effectively;
• •  Understanding AI’s strengths and limitations; and
• •  Engaging in the iterative processes central to transactional work: drafting, reviewing, and negotiating.
•  

    I have launched this redesigned course in spring 2026 and will simultaneously conduct a research study to assess student learning outcomes. I will report findings candidly, whether they confirm or contradict my hypotheses. I also welcome opportunities to share my Spellbook assignments with colleagues interested in adapting them.

    Although Spellbook is highly effective, its educational value depends entirely on the pedagogical framework built around it. This framework provides the template for successful AI integration: powerful tools combined with robust instructional design, yielding learning experiences in which AI enhances, rather than replaces, the development of legal judgment and critical thinking. This is the future of legal education: educators, instructional designers, and academic technologists collaborating to build a pedagogy that prepares students for an AI-transformed profession.

 

 

* Professor of Law, Mitchell Hamline School of Law (MHSL). I teach contracts, sales, secured transactions, business organizations, and related courses, and I am currently researching the pedagogical impact of AI integration in required doctrinal courses. I thank MHSL instructional designer Amanda Soderlind for collaborating on the Spellbook pedagogy described here and for exemplifying the value of strong professor-instructional designer partnership. I also thank former colleagues Mary Pat Byrn and Simon Canick and MHSL Professor Emerita Nancy Ver Steegh for helping me appreciate that technological innovation demands strong pedagogy and that instruction designers are critical in this regard.