This session will feature two presentations centered around AI Agents: 

• (L)SaaS: Librarian Software as a Service
• Building a Reliable AI Agent: What Context Fixed, What Drifted, and Why Librarians Already Have the Core Skill

(L)SaaS: Librarian Software as a Service

This program explores how law librarians can act as “software as a service” providers by combining legal research expertise with practical knowledge of AI tools and vibe coding. Drawing on two examples from the Jameson Law Library—a vibe-coded web scraper created to assist a faculty member and an internal AI agent developed to analyze faculty publication patterns—the program will show how librarians can design practical, targeted technology solutions for their patrons. These examples serve as test cases for the broader potential of (L)SaaS as a model for expanding the role of law librarians in research support and legal technology innovation.

TARGET AUDIENCE:

All types of law librarians.

TAKEAWAYS:

1. Gain an understanding on how AI agents and vibe coding can be used to create useful tools for creating practical Librarian Software.
2. Identify opportunities within their own institutions where librarians can develop targeted technology solutions to support patrons.

SPEAKERS:

• Cody James, Director of the Law Library and Assistant Professor, William J. Jameson Law Library (University of Montana)

AALL BODY OF KNOWLEDGE DOMAINS: Research + Analysis, Information Management

Building a Reliable AI Agent: What Context Fixed, What Drifted, and Why Librarians Already Have the Core Skill

You’ve built your first agent. But how do you know you’ve done it well? This case study walks through best practices in building and testing agents, using a research support agent designed to help researchers respond to law firm research requests. Attendees see the agent in action, the scoring rubric used to evaluate it, and the test results across four versions, from a baseline that invented tools and misjudged how long requests would take to a version that passed every accuracy check.

Context engineering (the context you provide to the AI; in this case, organized knowledge about the department, its tools, and its limitations) fixed the dangerous failures, including fabricated tools, wrong tool routing, and scope creep. Formatting and style proved harder. Even with a formatting rule written into the agent’s context, style drifted between runs because written guidance is not the same as structural enforcement. An agent can hold context and still drift from it, which is why humans stay in the loop.

The deeper lesson for our profession is self-advocacy. The work that made this agent reliable was classification, metadata, information architecture, and taxonomy. Law librarians already have the core skill. It translates directly to building and maintaining AI systems.

TARGET AUDIENCE:

Law librarians and legal information professionals who are evaluating, piloting, or already using AI tools. Especially relevant for directors and managers deciding whether an AI agent is reliable enough for real work, and for anyone who suspects their information organization skills matter more in the AI era, not less. No technical background required.

TAKEAWAYS:

1. Distinguish context engineering from prompt engineering, and map the information organization skills librarians already use (classification, metadata, information architecture) to the work of making AI agents reliable.
2. Build a simple test rubric that shows whether an AI agent actually works, and reveals which failures context resolves and where a human still needs to stay in the loop.
3. Apply a progressive trust model to decide what an AI agent can see and do at each stage of adoption.

SPEAKERS:

• Courtney Toiaivao, Managing Director & Founder, Grey Letter Advisory

AALL BODY OF KNOWLEDGE DOMAINS:  Information Management, Management + Business Acumen

All AALL webinars include live captioning services.

The opinions shared during this program represent the views of the speakers and do not necessarily reflect those of the American Association of Law Libraries (AALL). Recording, capturing, or using AI tools to duplicate, transcribe, or otherwise reproduce an AALL program in any form is strictly prohibited without prior written consent from AALL. This includes, but is not limited to audio, video, or any other content shared. By attending an AALL program or accessing an AALL recording, you agree to adhere to this policy.