L
E
G
A
L
L
E
G
A
L
LEADERS EXPLORING GENERATIVE AI IN LAW
LEADERS EXPLORING GENERATIVE AI IN LAW
Case Study Canvas Annotations
ANNOTATION: The Case Study Canvas is our semi-structured follow-up instrument for exploring GenAI use cases in real depth—but only when depth is actually warranted. In every organization, use cases are numerous, uneven, and constantly evolving. Asking providers to fully document every use case upfront would create enormous burden while producing more detail than most clients could reasonably absorb. The Canvas solves this problem by creating a targeted, on-demand pathway for deeper exploration.
The idea is simple: our survey already captures high-level summaries of what providers are Doing, Planning, and Thinking across a wide range of GenAI applications with commercial impact. Those summaries give clients a quick, comparable view of activity—but they are not meant to be exhaustive narratives. If a client reviews those summaries and wants to understand a particular use case more fully, then the Case Study Canvas becomes the next step. It provides a structured way for the provider to unpack a specific use case—its purpose, workflow, safeguards, outcomes, constraints, and lessons learned—without requiring them to write essays about every experiment or deployment in their portfolio.
Put differently:
- If a client never engages with the high-level summaries, they were never going to read (or need) deep case studies in the first place.
- If a use case does spark genuine interest, the Canvas ensures the follow-up is focused, comparable, and meaningful—rather than ad hoc or overwhelming.
By reserving detail for where it matters, the Case Study Canvas accomplishes two things:
- It reduces burden on providers by eliminating the need for comprehensive write-ups on every use case.
- It enhances the quality of client–provider dialogue by channeling attention toward the use cases that actually activate client interest and merit deeper discussion.
The result is a smarter, lighter, more purposeful exchange: rich where it needs to be rich, efficient where it needs to be efficient, and well-structured enough to support clarity, alignment, and decision-making on both sides.
1. Use-case Name. What is the best available shorthand for referring to the use case?
ANNOTATION: This is about clarity and recall. The shorthand should be what a colleague or client would naturally use to identify the project—not internal jargon or an overly technical label. A crisp, intuitive name makes it easier to benchmark use cases across organizations and to reference them consistently in comparative outputs.
2. Deployment Date. When was the use case first rolled out?
ANNOTATION: This question situates the use case in time. We’re not looking for initial experimentation but rather when it became operationally available in practice, even in pilot form. The goal is to establish a temporal benchmark that allows us to map maturity and adoption trajectories across different organizations.
3. Updates. When was the last material upgrade or update?
ANNOTATION: This asks when the use case last advanced in a meaningful way—whether through new features, expanded scope, technical improvements, or significant process refinements. “Material” here distinguishes substantive evolution from routine maintenance. The goal is to capture how actively the use case is being iterated on.
4. Brief Description. Please provide a concise description (3-6 sentences) of the use case.
ANNOTATION: This is the narrative anchor of the canvas. A good description should be concrete and client-facing, explaining what the use case does, who it serves, and how it adds value. It should be understandable without specialized technical knowledge. Avoid promotional tone—focus on clarity, specificity, and enough detail to distinguish it from other use cases.
5. Operation. Where in the workflow (e.g., stage of matter) does the use case operate? Which concrete tasks are being automated or augmented?
ANNOTATION: This question locates the use case within the legal workflow. The emphasis is on specificity: which stage(s) of a matter it applies to and which tasks are directly impacted. Naming concrete activities reduces ambiguity (e.g., drafting NDAs, summarizing depositions, reviewing compliance documents).
Strategic Rationale & Impact
6. Objective(s). Which objectives does this use case advance? (check all that apply)
- New Offering(s) to Clients
- Output Quality
- Speed
- Cost Predictability
- Compliance
- Risk Mitigation
- Other:
- Outcome Predictability
- Volume/Scalability
- Client Experience
- Cost Reduction
- Multilingual Capacity
- Knowledge Capture & Reuse
ANNOTATION: This question surfaces the strategic “why” behind the use case. It’s a structured way of mapping GenAI activity to broader business or client-facing goals, from efficiency and cost control to quality and risk mitigation. Checking multiple boxes is common—most use cases serve several objectives. What matters is being candid about which outcomes the initiative is intended to advance, not overstating impact.
7. Impact on Objective(s). Please explain how the use case advances the objective(s) checked directly above. Where possible, include specific examples or directional metrics (e.g., throughput, turnaround time, predictability scores).
ANNOTATION: This moves from intent to evidence. We’re asking you to connect the dots between objectives and outcomes, ideally with directional measures (e.g., “reduced review time by ~30%,” “expanded coverage across X matters”). Precision is welcome but not required—directional examples are often the most helpful. The goal is to understand how organizations are translating GenAI into observable results.
8. Client Value. What specific and visible value does this initiative deliver to clients? What about this use case should influence their commercial decisions (e.g., work allocation, fee structures) or deliver demonstrable benefits (e.g., measurable savings, tangible improvements)?
ANNOTATION: This question tests whether the initiative produces client-facing value, not just internal efficiencies. It prompts respondents to articulate tangible benefits clients can perceive, measure, or act upon in commercial terms. The goal is to capture how GenAI initiatives translate into differentiators that matter in client decision-making.
9. Commercial Impact. What has been the commercial impact of this use case? Please describe how it has changed your client relationships (e.g., generated new work, reduced costs, altered fee models). Where possible, include specific examples or directional metrics (e.g., % cost reduction, new matter types, volume of work retained).
ANNOTATION: This is the heart of our exercise. We want to understand how the use case has shifted the economics of service delivery—who does the work, how clients are billed, whether new revenue or pricing models have emerged. Even modest or directional examples are valuable: a fee model adjustment, a client retention story, or a new matter category created. The emphasis is not on internal efficiency alone, but on observable changes in commercial relationships.
Users & Adoption
10. Matters. Which matter types does this use case primarily apply to (e.g., general M&A, litigation, compliance, IP)? Please specify practice areas or categories, not client names.
ANNOTATION: This question asks where the use case “lands” in practice. The goal is to identify which kinds of legal work it most directly applies to, using categories or practice areas rather than client examples. Precision matters less than clarity—naming the matter types helps benchmark adoption patterns across the industry.
11. Users. Who are the primary users (e.g., associates, partners, paralegals, project managers, technical staff, clients)? Please specify roles or functions, not individuals.
ANNOTATION: This captures who is actually interacting with the tool or output. We’re interested in functions and roles, not specific names. Responses might include different levels of lawyers, allied professionals, clients, or technical staff. The focus is on identifying the primary user base so we can better understand adoption dynamics and workflow impact.
12. Adoption Metrics. What are your best available adoption indicators? Please include quantitative metrics where possible (e.g., % of matters covered, # of active users, frequency of use) and qualitative signals (e.g., user feedback, repeat usage).
ANNOTATION: This moves beyond intent to evidence of uptake. We’re asking how you measure adoption, whether through numbers (e.g., percentage of matters, frequency of use) or qualitative indicators (e.g., strong repeat usage, positive feedback). Even directional or partial metrics are useful.
13. Target. What would an optimal adoption footprint look like (e.g., proportion of matters, user penetration, business units)? What specific actions are you taking to drive adoption toward that goal?
ANNOTATION: This question highlights both aspiration and tactics. We want to know what “good” looks like for you (e.g., adoption across all compliance matters, or penetration across X% of associates) and the concrete steps you’re taking to get there.
Tooling & Evaluation
14. Models. Which GenAI models does this use case rely on (e.g., GPT-5, Gemini, Claude, Llama)? Please specify version(s) if relevant.
ANNOTATION: This question highlights both aspiration and tactics. We want to know what “good” looks like for you (e.g., adoption across all compliance matters, or penetration across X% of associates) and the concrete steps you’re taking to get there.
15. Buy. Which commercially available tools are part of the stack (e.g., Legora, Harvey, CoCounsel, MS Copilot)? Please list primary tools and their function in this use case.
ANNOTATION: This focuses on the off-the-shelf tools you’ve layered into the use case. Naming the tool and its role provides important context. We’re not looking for an exhaustive list of every integration—just the key tools that materially enable the use case.
16. Build. Which components have you built, configured, or customized (e.g., connectors, fine-tuning, UI)? Why was this approach chosen instead of using off-the-shelf solutions?
ANNOTATION: This question explores where you’ve gone beyond plug-and-play. Did you develop connectors, fine-tune models, or build interfaces? And why? The goal is to surface the balance between buying vs. building, and the rationale behind it—cost, control, data sensitivity.
17. Data Sources. Which data sources (e.g., DMS, matter management, precedent libraries) are required for this use case? What steps were taken to prepare or clean data to achieve AI readiness?
ANNOTATION: This probes the data foundation. We’re interested both in the sources being tapped (DMS, precedent sets, case law, etc.) and in what was required to make that data usable (cleaning, tagging, structuring). The emphasis is on readiness and data governance, since these are common adoption bottlenecks.
18. Selection. How did you evaluate and select models, tools, and integrations for this use case? What criteria (e.g., accuracy, cost, security, vendor reputation) were most decisive?
ANNOTATION: This is about decision rationale. We want to understand the criteria that mattered most—accuracy, cost, security, client comfort, or vendor reputation. Responses help benchmark how organizations are weighing trade-offs when making GenAI investment choices.
19. Evaluation. How did you test full-stack performance before rollout (e.g., pilots, benchmarks, feedback)? What factors (e.g., accuracy, efficiency, user adoption) ultimately drove your decision to proceed?
ANNOTATION: This captures the “last mile” before deployment. We’re interested in how you validated the solution (pilots, benchmarks, controlled rollouts) and the factors that ultimately tipped the decision to proceed. The emphasis is on the evaluation process, not just technical performance.
Notice, Risk, Controls, & Ethical Use
20. Opt-in/Opt-out. Do you offer clients the ability to opt-in or opt-out of these GenAI-enabled services? If yes, please describe when and how you present these options.
ANNOTATION: This question explores whether, when, and how clients are given formal choice or control over GenAI use in their matters. “Opt-in” typically means explicit client approval is required before GenAI is used; “opt-out” means GenAI is the default, but clients are notified and may decline. The follow-up prompt asks how these choices are presented (e.g., standard engagement terms, individual matter scoping, policy statements).
21. Accuracy. What accuracy threshold do you consider acceptable for this use case? How do you measure it (e.g., benchmarks, user validation, error rates), and what safeguards (e.g., human-in-the-loop, escalation protocols) ensure it is consistently achieved?
ANNOTATION: This question targets how you define and maintain quality. We’re interested in what “good enough” looks like for the use case—how accuracy is measured, and what controls are in place to maintain it. The emphasis is on thresholds and safeguards rather than perfection. We know human oversight is often part of the answer; the key is clarity on how accuracy is tracked and enforced.
22. Ethical Use.What practices or guidelines did you adopt to ensure ethical and responsible use of AI in this project (e.g., bias testing, transparency measures, restrictions on sensitive data use)?
ANNOTATION: This asks how you’ve framed the use case in ethical terms. That may include fairness testing, transparency practices, or restrictions on sensitive data. The goal is to surface principles and practices—not just technical controls—that help ensure the project aligns with professional standards, client trust, and responsible innovation.
23. Controls. What technical or organizational controls are in place to address risks such as data segregation, model privacy, auditability, and bias mitigation?
ANNOTATION: This narrows in on safeguards beyond accuracy—focusing on governance and compliance. Examples include data segregation, audit trails, access permissions, or bias-mitigation protocols. We’re asking how you’ve operationalized risk management, both technically and organizationally, to make the use case sustainable and defensible.
Looking Back & Ahead
24. Road Map. What enhancements or next steps are planned for this use case, and on what timeline (e.g., 3-6 months, 12 months, 24+ months)? Please include both technical improvements and adoption/rollout milestones.
ANNOTATION: This question looks forward. We’re interested in how you expect the use case to evolve—both on the technical side (e.g., model upgrades, added features) and the adoption side (e.g., broader rollout, more users, new matter types). Timelines help benchmark maturity: are you in a 6-month iteration cycle, a 12-month scale plan, or a longer 24+-month roadmap? The intent is not to lock you in but to better understand the planned trajectory.
25. Lessons Learned. What are the most important lessons learned from this use case so far—whether technical, strategic, cultural, or commercial?
ANNOTATION: This asks you to step back and distill experience into insight. The lesson could be positive (unexpected value, client enthusiasm) or cautionary (cultural resistance, integration challenges). We’re not looking for a long list—just the most salient takeaway that would help others understand what it really takes to implement a GenAI use case in practice.
26. [optional] Anything else? This optional catch-all question leaves space for—but does not require—information, observations, or opinions not elicited above that you consider important to share with respect to your organization’s usage of GenAI to deliver legal services.
ANNOTATION: This question provides open space for input that doesn’t neatly fit earlier categories but may be valuable to the broader dialogue. It’s entirely optional. Even brief observations can highlight areas meriting deeper exploration in the future.
Tags (check all that apply): Area(s) of Law
- General Purpose (not specific to any area of law)
- Agriculture
- Cannabis
- Contract
- Education
- Finance
- Health
- IP
- Municipal
- Public & Admin
- Securities
- Transportation
- Banking
- Commercial & Trade
- Corporate
- Energy
- Food & Drug
- Information Security
- Labor & Employment
- Personal & Family
- Real Property
- Tax
- Other
- Bankruptcy/Restructuring
- Constitutional & Civil Rights
- Criminal
- Environmental
- Gaming
- Insurance
- Law of Obligations
- Personal Injury & Tort
- Religious
- Telecoms, Media, & Entertainment
ANNOTATION: This question asks you to categorize the use case by the area(s) of law it most directly applies to. The intent is not to force artificial precision—many use cases are cross-practice—but to provide a directional benchmark of where GenAI is being deployed. If it’s general-purpose, you can indicate that. If it spans multiple categories, check all that reasonably apply.
Tags (check all that apply): Area(s) of Application
- Billing & Time Tracking
- Chronology Creation
- Coding/Software Development
- Depositions
- Document & Knowledge Management
- Drafting (Contracts/Corporate)
- Due Diligence
- Generating Images/Multimedia
- Non-Legal Research
- Pricing
- Regulatory Compliance
- Search & Retrieval
- Written Discovery
- Other:
- Case/Matter/Project Management
- Client Communications
- Contract Review & Redlining
- Document/Data Review & Analysis
- Drafting (Litigation)
- Drafting (Other)
- eDiscovery
- Legal Research
- Predictive Analytics
- Proofreading/Editing
- Regulatory Tracking
- Summarization
ANNOTATION: This identifies the functional layer of work being augmented—contract review, drafting, research, compliance, etc. The goal is to capture the “what” of the work, not the “why.” This helps build a taxonomy of GenAI’s actual touchpoints in legal service delivery. As with law areas, multiple selections are expected and valuable.