If generative AI catapulted to the top of the list of technologies catching the attention of librarians, the arrival of agentic AI is likely to be even more eye-catching. Looking forward to the new year, many pundits in the IT space proclaimed agentic AI to be the most important technology for 2025. Considering that GenAI hit those "technologies to watch" only in late 2022, it's astounding to realize that in two very short years, newer versions and implementations have supplanted GenAI on those lists and GenAI itself is now considered almost an accepted part of daily life, embedded to the point of being invisible.
In the library world, the advent of GenAI and its rapid integration into the research and writing process of academia at all levels, from primary grades up through doctoral and post-doctoral studies, has been truly breathtaking. Information professionals struggle to keep up with changes in AI capabilities, potential new uses, and possible misuse.
It's not only academic libraries that are intrigued by and coping with GenAI. Corporate, government, and public libraries also confront the implications of the technology within their particular situations.
This latest innovation, agentic AI, presents new challenges and opportunities for librarians, library clientele, and library parent institutions.
What is agentic AI?
A natural outgrowth of intelligent agents, agentic AI adds autonomy and reasoning power to existing agent technologies. Intelligent agents have been around almost since the term "artificial intelligence" was coined in the 1950s. Writing in ONLINE magazine in July 1994 ("Intelligent Agents: Software Servants For An Electronic Information World [and More!]"), Marina Roesler and Donald T. Hawkins, then at Bell Laboratories, defined intelligent agents as "autonomous and adaptive computer programs operating within software environments such as operating systems, databases or computer networks." The new generation of software, they wrote, combined AI with system development techniques. They identified typical tasks as "filtering electronic mail, scheduling appointments, locating information, alerting to investment opportunities and making travel arrangements."
Although most of the products mentioned in the article no longer exist (NewsEDGE, anyone?), Roesler and Hawkins' speculations about the future direction for intelligent agents sound eerily like what is now being touted as agentic AI. Even 30 years ago, they predicted that agents would have autonomous agency, adaptive behaviour, mobility, cooperative behaviour, reasoning capability, and anthropomorphic interfaces. Those predictions were spot on.
Agentic AI leverages advances in AI technology, particularly GenAI, to make possible what was only speculation. While intelligent agents operated within pre-defined boundaries, these AI-powered agents are capable of independent decision making. No longer simply reacting to commands or events, agentic AI can proactively identify and pursue goals and frequently uses APIs to execute tasks. Although learning was implicit in early intelligent agents, agentic AI expands its learning via a feedback loop to quickly adapt to new situations, thus improving performance over time and adding to their autonomy.
Another attribute of agentic AI is the ability to ingest enormous amounts of information and data from multiple sources, allowing it to develop strategies and execute tasks using a multi-step approach. Think of a traditional online search strategy in library subscription databases. A human inputs a Boolean search, then modifies the strategy based on results obtained from the original search statement. Agentic AI does much the same thing, but uses Large Language Models (LLMs) rather than one database and relies on Retrieval-Augmented Generation (RAG) to reduce hallucinations and ground its reasoning powers in vetted, often proprietary, data.
Outside of libraries, AI agents are being used in manufacturing, data analysis, medical diagnoses and customer service, among other implementations. Mark Purdy, writing in the Harvard Business Review in December 2024, asked " What Is Agentic AI, and How Will It Change Work?" He notes that AI agents are focused on making decisions rather than creating content and, in the business world, are set to optimize a company's objectives of "maximizing sales, customer satisfaction scores, or efficiency in supply chain processes." He cites the discovery of novel organic compounds as a result of agentic AI involvement.
How agentic AI could affect librarians
A basic idea behind agentic AI is mirroring human behaviour so that it resembles human employees. Think of them as interns or paraprofessionals. Within libraries, agentic AI could be instrumental in anticipating the research needs of library users, suggesting relevant resources and identifying any research gaps it perceives. For collection development, an AI agent could not only suggest materials to purchase but also initiate purchase orders independently. It might also take on weeding duties.
In an article in worklife, Tony Case claims that a "hot new job" is AI librarian that "bridges the gap between human expertise and machine learning". However, this is not a job operating in an actual library, although professional librarians are cited as a source of talent. Instead, Case sees them more as knowledge managers in companies. What's needed is experts to curate data in knowledge hubs and translate data into actionable insights. AI librarians will ensure that the potential of AI technologies aligns with information accuracy. That's the job of an AI librarian in business environments.
The University of Chicago library system, in early December, posted a job opening for an actual Artificial (AI) librarian, a new position reporting to the Director of Digital Scholarship. Although the job description includes responsibilities familiar to professional librarianship—engaging with campus partners, raising awareness of AI technologies, contributing to strategy development, providing training and outreach to researchers and promoting digital literacy—a degree in library/information science is not required. Instead, the university wants someone with a Masters degree in computer science, data science, information systems, or AI/ML.
Agentic AI, drawing as it does on GenAI and expanding the possibilities offered by intelligent agents, is a new competency for the library community. It is not, however, a totally new skill. The empathy central to the work of human librarians is often missing from AI agents. Thus, the combination of emotion and technology fits perfectly into the role of AI librarian, whether it's called by that title or not and whether the job is within an entity called a library or situated elsewhere within an organization. Agentic AI is here to stay and it's worth the time and effort of librarians and information professionals to ascertain its relevance to the profession and to exploit the technology so that it serves the best interests of library clientele and librarians themselves. Looking ahead, it's not at all unrealistic to predict more AI librarian jobs on the horizon. Even without the title of AI librarian, many current jobs are set to incorporate agentic AI in their responsibilities. Will agentic AI open up new opportunities? Here's the simple answer: Yes.