By Kristen Sonday
Starting an AI project may seem like an overwhelming task, especially in such a fast-paced space with limited resources. However, the recent AI for Legal Aid Summit proved that it’s not only feasible, but can yield transformative results.
At the Summit, we explored six innovative AI initiatives already driving real-world impact, from automating repetitive tasks to expanding access to legal information for underserved communities. Alongside these success stories, hands-on workshops — like crafting effective prompts and building custom GPT models — demystified the process, showing how any organization can start small and scale intelligently.
By following a clear, structured roadmap, launching your first AI project can be both accessible and impactful. Below is a 10-step guide to help you get started:
1. Identify the Problem
As Laura Safdie, Head of Innovation for Legal at Thomson Reuters, explained in her keynote, you can think of AI like an engine, and you use that engine to power a solution that’s well-equipped to solve specific problems (e.g. driving a car down a road, helping a plane ascend into the air). Similarly, impactful legal aid AI projects begin with a clear understanding of the challenge at hand. In our example, many legal aid organizations struggle with bottlenecks in their client intake processes, leading to delays in providing assistance. By identifying this pain point, you’ll create a focused target for your AI initiative.
2. Define the Use Case
Once the problem is clear, articulate a specific use case. Andrew Shulman from Thomson Reuters walked us through how he partners with LSOs to identify a good use to start with. He asks: what is a part of your work that is 1) resource intensive, 2) repetitive or common and causing bottlenecks, or 3) you’d prefer to automate? In this example, the use case could be: “Develop an AI-powered chatbot to handle basic client inquiries, collect initial information, and triage cases for further review.” A well-defined use case guides your project and helps set measurable goals.
3. Evaluate Resources
Now, you’ll need to assess the resources you’ll need to execute the project, and whether (or to what degree) you have them internally. Resources may include:
- Training Data: Gather historical intake forms or FAQ data to train the AI model. Potentially look at other intake forms and aggregate best practices.
- Budget: Assess your funding options, including grants or partnerships.
- Skills: Identify whether your team has the technical expertise internally, or if external partners will be necessary.
4. Engage Stakeholders
Successful projects involve collaboration. Engage attorneys, support staff, and leadership to ensure the solution aligns with organizational needs. Importantly, consult clients to understand their preferences and pain points. For example, clients may prefer text-based chatbots over voice-based systems for accessibility reasons.
5. Choose the Right Tools
Decide whether to build a custom AI solution, collaborate with an external developer, or leverage off-the-shelf tools. For instance, several platforms offer ready-made chatbot frameworks tailored for nonprofits. Prioritize tools that emphasize privacy, security, and ethical use of AI, ensuring compliance with legal and organizational standards.
6. Pilot the Project
Start with a small-scale pilot to test the chatbot, and keep humans in the loop to tweak responses as you go. For example, you might deploy the chatbot only for eviction cases initially. Collect feedback from users, both clients and staff, to refine the system. Early testing helps identify potential pitfalls at scale and ensures the technology meets real-world needs.
7. Train and Educate Staff
Introduce the chatbot to your team through training sessions. Staff should understand how it works, how it integrates into their workflows, and the impact it is intended to have, whether for clients, the organization, or the public. Emphasize that the chatbot is a tool to augment their work and increase capacity to serve client needs, not replace it.
8. Monitor and Evaluate
Use metrics to assess the chatbot’s effectiveness. Key performance indicators (KPIs) might include:
- Reduction in client wait times
- Increased number of cases triaged
- Client satisfaction scores
Establish a feedback loop to continually improve the system based on user experiences.
9. Scale Up
Once the pilot proves successful, expand the chatbot to cover additional case types or geographic areas. For example, after addressing eviction cases, you could extend it to handle family law or immigration queries.
10. Maintain and Update
AI systems require ongoing maintenance to remain effective. Budget for updates and monitor advancements in AI that could further enhance your chatbot’s capabilities. Additionally, ensure the system adapts to any changes in legal procedures or organizational priorities. Set periodic times (once a month, quarter, or biannually), to run internal tests and surface any issues, as well as ensure content is up-to-date.
Practical Example: AI-Powered Intake Chatbot
In one project, a legal aid organization faced challenges with overwhelmed intake staff and delayed responses to clients. Using historical client inquiry data, they developed a chatbot to:
1. Ask clients basic questions.
2. Triage cases based on urgency and eligibility.
3. Schedule appointments with the appropriate staff member.
The chatbot’s pilot focused on eviction cases in a single jurisdiction. Within three months, the organization saw a 40% reduction in intake times, freeing up staff to focus on complex cases. Clients appreciated the quick response and accessibility, especially during peak hours when human staff were unavailable.
AI has the potential to transform legal aid organizations by streamlining processes and increasing capacity to serve clients. The technology is in early stages though, so be patient with the process, be intentional about your design decisions, and accept that you will iterate. By starting with a well-scoped project, engaging stakeholders, and adopting an iterative approach, however, you can set your organization up for success in the AI era. Let’s harness this technology to bridge the justice gap together.
Kristen Sonday is the Co-Founder and CEO of Paladin, whose mission is to increase access to justice by helping legal teams run more efficient pro bono programs. She is a member of LSC’s Leaders Council and hosts a NextGen Justice Tech column for Thomson Reuters’ Legal Executive Institute.
AI for Legal Aid: A Step-by-Step Guide was originally published in Justice Rising on Medium, where people are continuing the conversation by highlighting and responding to this story.