From the Big Picture to Quick Wins: The Next Two Years of AI in Academic Libraries (Part 1)

A roundtable with ACRL's Artificial Intelligence Interest Group conveners

Posted on May 4, 2026 in Blog Posts

The simple truth is that it’s hard to keep up with AI. Even those passionate about the topic surely struggle to stay up to date with the latest applications, problems, policies, lawsuits, and discussions surrounding it. And while understanding the present landscape is challenging enough, we must nevertheless plan for the future to make sure we and our institutions are ready for it. 

The editorial team at Choice has been mulling over the question, What does AI in libraries look like two years from now? Two years ago, we were finalizing our white paper on defining AI literacy, which would come to form the backbone of ACRL’s AI Competencies for Library Workers. These were big moves, but only preliminary ones. Since then, libraries across the world have put together workshops and communities of practice, integrated chatbots into digital collections, evaluated research tools, and taught critical frameworks to students and faculty. And yet the question remains: now that we have a better understanding of AI and more confidence in dealing with it, where do we go from here?

To help us tackle this question, we invited the three conveners of ACRL’s Artificial Intelligence Interest Group, who have engaged with AI on the ground at their institutions and also have a unique bird’s-eye view of the profession from their perch in the interest group. In the interview below, Heather Sardis (MIT), Virginia (Ginny) Pannabecker (Virginia Tech), and Amanda Y. Makula (Univ. of San Diego) discuss the next two years of AI in libraries, quick wins for libraries with limited budgets, partnerships between libraries and other university bodies, and more. Below is the first part of our conversation.

💫 Learn more about the ACRL AI Interest Group and join on ALA Connect

---

When talking about emerging technology, even the short term feels long-term. “Where will AI be in two years?” is a difficult question because we simply don’t know. So, let me put it to you this way: in two years from now, where would you like academic libraries to be with regard to AI?

Heather: We do have some sense of where AI might be in two years! The transformer architecture that enabled the development of today’s large language models emerged in 2017, and while it was a legitimate breakthrough, it was still built upon decades of focused research in neural networks. Now, AI researchers are working on addressing known problems with today’s transformer-based architectures and building the next generation of large language models.

The work of Yann LeCun and others on “world models” aims to fix a core limitation of transformer-based AI: because current large language models are trained heavily on text and images, they have a limited understanding of our real world and struggle to ground their outputs in meaningful context. LeCun argues that more robust AI models that learn by building internal representations of how the physical world works will become prominent in the next few years. Similarly, researchers like MIT’s Yoon Kim are developing more efficient models that can maintain a more consistent understanding of the internal “state” of an interaction over time and are better able to track their own uncertainty and limitations. In his framing, scale and brute force alone won’t result in AI systems that can genuinely reason, and his research focuses on developing new model designs that learn in ways closer to how humans track and update information over time.

But, while the next steps in technological development are not entirely opaque, the question of who will build the guardrails for these systems remains unanswered. This is why the role of libraries remains critical, now and into the future. Our voices are needed to help govern how these technologies are implemented, with our professional values at the center. This work will be necessary and timeless regardless of the specific mechanics of the models themselves.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• ITHAKA’s Research on Generative AI and Its Implications for Librarianship
• Streamlining Your Literature Review Workflow with ResearchRabbit
• Balancing User Needs and Technology after System Migration

📣 Do you teach AI literacy? We want to hear from you. Check out our call for papers, “AI Literacy Across Disciplines.”

---

Ginny: In the context of “where would you like academic libraries to be with regard to AI?” while AI literacy and training programs and use of AI for operational, research, teaching, or other endeavors vary across libraries and within each library system, academic libraries demonstrate a range of engagement.

Libraries are supporting AI literacy and training initiatives through university-wide programs, news media, events, guest lectures in courses, and skills workshops for audiences ranging from undergraduate students to faculty and the public.

For example, academic libraries are providing workshops on AI and literature reviews covering such topics as AI features in library databases and online collections, AI tools that provide starting points for literature reviews, and uses for GPT chatbots in literature review or other processes. For each of these, benefits, limitations, and methods for evaluating AI results are included.

Academic libraries have also been engaged for years in using machine learning and deep learning methods, building large (and small) language models (LLMs, SLMs), utilizing specific AI tools, and increasingly, utilizing generative AI models and agents for operational needs to build and increase description of and access to collections and support or engage in research and teaching. The University Libraries at Virginia Tech is one academic library system engaging with AI in a variety of ways, such as: to combat AI-related mis/disinformation, for digital curation, and through the Center for Digital Research and Scholarship’s AI Initiative.

Broadly across libraries, there is a need for increased knowledge, skills, and experience in using AI, evaluating when not to use it, and evaluating AI results for use in library work or in support of growing AI use in a range of academic disciplines and other areas.

In the coming two years, there are many ways academic libraries can learn from each other, share what they’re trying—what’s working and what isn’t—and support our professional community writ large, including higher education as a whole. Many are making strides already and sharing their experiences, including in the areas described below, through symposia, lecture series, conferences, professional groups like the ACRL AI Interest group, and elsewhere. Some markers of where it would be great to see academic libraries within two years include:

• Developing and sharing library organizational structures and frameworks that work well for coordinating AI use and initiatives across the library and within the library’s role as a core university partner. As part of this: integrating with institutional infrastructure, guidelines, and best practices, and through the institution or library’s unique AI initiatives; connecting to broader professional community best practices; demonstrating effective coordination of training, support, and collaboration across organizations, with flexibility for exploration by individual library workers and areas.
  
  
• Building guidance for library workers to develop AI skills, such as the AI Competencies for Academic Library Workers from ACRL, including:
  
  
  • Reliably strong awareness from library employees of tools and methods that fall under the AI umbrella and an understanding of their differences
    
    
  • Increased confidence in operationalizing ethical and responsible evaluation and use of AI in services, operations, and AI-related purchases or funding
    
    
• Examples of libraries’ success in adopting and adapting (or even developing and sharing) open-source AI models and tools that meet common higher education data privacy and security needs, with resources developing in the community for others to learn from and make progress in this area, including how these fit in higher ed IT infrastructure
  
  
• Examples of how to maximize always limited monetary investments in AI access, including equipment, licenses, AI feature add-ons, within existing subscription tools or resources, for the best support for our communities
  
  
• Increased success in messaging and communicating libraries’ and information professionals’ roles in AI, with demonstrable examples of contributions across use, development of, research and teaching with, and administrative or operational use of AI
  
  
• Increased outreach to public communities, from local to international, about digital privacy rights and AI policy elements to spread awareness of useful approaches that allow for creativity and innovation while also protecting creators, copyright holders, users, and developers from plagiarism, fraud, data theft, or other concerns
  
  
• Spaces for frank discussions about AI opportunities, along with comfort in discussing AI concerns and issues to support the use of AI in specific contexts and gain facility in using AI relevant to one’s own context, along with effective ways to evaluate results, address caveats, considerations, and known or potential downsides and impacts.

What are some “quick wins” that libraries can pursue to reach this goal, particularly if they have limited budgets?

Ginny: Start or increase time investment in cross-library communication and coordination around exploring, understanding, using, and providing support and guidance in the use of AI by constituent communities. For example, does your library have an AI community of practice, strategy group, training initiative, or something similar?

If considering this option,

• Ensure at least one person is responsible for a regular schedule and can support the group in developing a charge or goals as high-priority interests and needs develop.
  
  
• Start informally, such as quarterly get-togethers, where all employees are invited to share examples of AI exploration or how they are using AI, or discuss training, resource, or administrative needs.
  
  
• Build as you go: increase meeting frequency; develop a structure where alternating meetings include completing a training together, trying a particular tool, sharing questions, or discussing concerns and benefits in real time. Other meetings can focus on more long-term planning and proposing or describing needs for future success, including funding for priority tools or equipment and organizational structure needs (AI-focused positions, responsibilities, standing committees or units, etc.).
  
  
• Consider including asynchronous methods (messaging app channels, discussion boards) for ongoing communication.
  
  
• Aim to avoid chastising approaches if people aren’t aware of institutional guidelines or requirements for approving tools, etc. Instead, share it as something to know about and discuss in real time how to shift to compliance, or tag it as something you or someone else will follow up on later.
  
  
• Be open to those who are choosing not to use AI tools and to learning more about why. As a group, look into the concerns they have and see what can be addressed in your environment, or how your library can contribute information about the challenges AI brings and strategies for addressing them.
  
  
• As you grow your group, you may identify needs for specific working groups, subcommittees, or designated responsibilities for coordination.

Explore and develop your skills and support for AI features that can be added to your current products (literature databases, ebook collections, discovery search platform, etc.), especially those that don’t require additional funds.

Talk with others at your institution about resources your institution as a whole is supporting and investigating. Are there tools approved by your institution for secure use and/or everyday work that you can use without additional cost to the library, or that you can fund access to for some library employees to begin using to reach a higher level of confidence? (Review example tools to consider supporting or providing access to, such as those shared in the Ithka S+R Generative AI Product Tracker, and help others learn by submitting new tools to the Tracker.)

Explore possibilities with library consortium partners for AI tool trials and agreements.

Work together across the library and with institutional partners to develop institutional knowledge and connections by organizing AI events, speaker series, or journal clubs. (Join us at the ACRL Artificial Intelligence Interest Group!)

Explore partially open and open-source AI models and tools, or consider how you can adapt or customize tools or create agents within institutionally approved tools. Start an in-house group to investigate these options and evaluate whether any may work within your institutional context for adopting new technology or integrating with existing IT infrastructure, permissions, or compliance.

➡️ Check out part 2 of this conversation

➡️ Join the ACRL AI Interest Group on ALA Connect

Related Posts