Automation Won't Save You (But It Might Help)

Every few years, someone announces that technology is about to transform libraries. Self-check will eliminate circ desks. AI will replace catalogers. Chatbots will handle reference. The library of the future will run itself.

None of that has happened. What has happened is quieter and more useful: specific automation tools have saved real time on specific tasks, freeing staff to do work that matters more. The libraries that benefit most from automation aren\'t the ones chasing the latest trend. They\'re the ones that figured out which repetitive tasks to hand off to machines so humans could focus on the parts of the job that humans do better.

This article covers what\'s actually working in library automation, what isn\'t, and where to start if your library hasn't touched automation beyond the basics.

Part 1: What Is Actually Working

Self-Checkout

Over 70% of US public libraries now have self-check machines. Libraries with RFID-enabled self-check typically see 75-80% of all checkouts happen without staff interaction, up from 1-2% with barcode-only systems. Some libraries go higher: Aurora Public Library District in Illinois hits 88%, Redford Township District Library in Michigan reaches 94%.

The pandemic permanently accelerated this. 92% of consumers now expect self-service options from service providers. Patrons aren't reluctant. When the technology works well, they prefer it.

The staffing impact is reassignment, not elimination. Aurora PLD explicitly states that self-check freed staff for "research assistance, programming, and outreach." The circ desk didn't disappear. The work that happens there changed.

For small libraries that can't afford traditional kiosks ($1,500 to $20,000+ per unit), Meescan offers cloud-based self-checkout using patron smartphones or library-supplied iPads, with annual license fees that include support.

Automated Materials Handling

AMH systems use RFID to identify, sort, and route returned items automatically. Multnomah County Library in Portland opened an Operations Center with a 65-bin AMH system that lets staff check in returns by the crate instead of individually. The facility houses over 500,000 items with nearly three miles of shelving.

A King County Library System cost study found $135,000+ in annual staff time savings at a single branch comparing operations with AMH versus without: $47,322 in frontline staff time, $75,712 in backroom staff time, and $12,950 in reduced unavailable resources.

The catch: AMH systems are custom-built capital investments. They make sense for high-circulation libraries and multi-branch systems with complex sorting needs. For a small single-branch library, the math rarely works. Major vendors include Bibliotheca, Tech Logic, Lyngsoe Systems, and EnvisionWare.

Automated Notices

This is the simplest, most widely available automation that many libraries still aren't fully using. Modern ILS platforms support automated overdue notices via email, SMS, and phone. PINES (Georgia) runs courtesy notices 3 days before due date, first overdue notice at 1 day past due, second at 10 days, all without staff intervention.

If your library still generates overdue notices manually, or only uses one notification method when your ILS supports three, you're spending staff time on a problem your software already solved.

Scheduled Reports and Batch Processing

Most ILS platforms support scheduled report generation, batch weeding candidate lists, batch record editing, and automated statistical reports. CollectionHQ (Baker and Taylor) automates collection analysis reporting with data-driven weeding recommendations.

If you're running the same report manually every Monday morning, your ILS can almost certainly generate and email it to you automatically.

Part 2: What Is Not Working (Or Not Ready)

Chatbots

Library chatbots struggle with complex questions and nuanced inquiries. They frustrate patrons when they hit the "I didn't understand" wall repeatedly. Error rates are higher for non-native English speakers. Repeated failures erode patron trust and drive users away from the service entirely.

Chatbots can work as first-line triage: "What are your hours?" "Where is the nearest branch?" "How do I renew a book?" But they're not a replacement for reference services. A hybrid approach, where the chatbot handles simple questions and escalates complex ones to staff, is the only model that works consistently.

Automated Catalog Enrichment

Enriching catalog records automatically sounds great until you look at the error rates. A study of Google Books digitization found 36% of sampled books contained metadata errors. Crosswalking metadata between formats introduces additional errors. AI models trained on biased datasets propagate inaccuracies. The consistent finding is that automatically generated metadata "generally needs human intervention to be successful."

Over-Automated Phone Systems

51% of customers have abandoned a business entirely after reaching an automated phone menu. 61% say interactive voice response (IVR) systems create a poor experience. Complex or lengthy menus confuse callers and increase call duration. When callers finally reach a human, the agent often lacks context, forcing patrons to repeat themselves.

If your library installed a phone tree that makes patrons press four buttons before reaching a person, you\'ve optimized for the library\'s convenience at the patron's expense.

Automated Social Media

Scheduling tools work well for planning posts. But purely automated social media feels robotic, because it is robotic. The New Jersey State Library\'s guidance is practical: batch-create content and schedule it, but log in regularly to engage with comments. If you respond to comments promptly, it shows followers "you're not just churning out pre-scheduled posts."

Part 3: AI in Libraries (What Is Real vs. What Is Hype)

What Is Real Right Now

OCLC AI Cataloging Suggestions (December 2025): OCLC launched AI-generated suggestions for Dewey Decimal Classification, Library of Congress Classification, and LCSH subject headings in WorldShare Record Manager and Connexion. The system draws on WorldCat's hundreds of millions of records. Catalogers maintain full control and can accept or ignore suggestions. Each library can customize which suggestions appear.

It's a practical, usable tool. It helps catalogers work faster on routine records while keeping human judgment in the loop.

Recommendation Engines: NoveList (EBSCO) has been providing read-alike recommendations for 30 years using a combination of human-written reviews and algorithms. BiblioCommons and LibraryThing generate recommendations from hundreds of millions of data points in readership patterns. These aren't new, and they work well.

What Is Not Ready

AI Cataloging Without Humans: The Library of Congress tested five open-source machine learning models on approximately 23,000 ebooks. Results: none reached the 95% accuracy threshold except for identifying LC Control Numbers. LLMs scored only 26% accuracy on predicting LCSH subject headings. The Library of Congress is now developing human-in-the-loop workflows where AI suggests and catalogers review. Phase three kicked off in August 2024.

26% accuracy on subject headings means AI gets it wrong three out of four times. Catalogers aren\'t being replaced. They\'re getting a tool that helps with the easy records and still needs them for everything else.

AI-Driven Collection Development: AI algorithms can analyze usage patterns, but real-world implementation in public libraries is still early-stage. The data exists. The tools to act on it reliably don't.

AI Chatbots for Reference: As covered above, current technology struggles with the complexity and nuance of real reference questions. AI chatbots work for factual lookups. They don\'t work for "my grandmother used to tell me about a book with a blue cover about a family in the 1800s." That\'s a reference interview, and it requires a human.

Part 4: The Staffing Question

The fear that automation will eliminate library jobs comes up in every conversation about technology. The data tells a different story.

The Numbers

In 2024, approximately 289,400 professional and technical workers were employed in US libraries: 186,500 librarians, 37,400 library technicians, and 65,500 library assistants. Library employment peaked at nearly 395,000 in 2006 and has declined since.

The Bureau of Labor Statistics projects librarian positions to grow 2% from 2024 to 2034, with about 13,500 annual openings. Median annual wage: $64,320. Library technician and assistant roles are projected to decline 7%, but about 25,800 openings per year are still projected due to retirements and turnover.

The pattern is clear: the human-centered professional roles are the ones being preserved and growing. The roles most susceptible to automation (routine processing, shelving, check-out) are the ones declining.

Reassignment, Not Replacement

Research consistently shows automation replaces specific routine tasks, not entire positions. Paraprofessionals have taken on work previously done by librarians, while new technology-related tasks have been created. The current library workforce challenge is staffing shortages, not automation-driven layoffs. Libraries are struggling to hire and retain staff, not trying to cut headcount.

About 23% of librarians are union members. Some union contracts include protections against technology-related job losses. AFSCME and AFT represent library workers and advocate for fair working conditions and appropriate staffing levels.

The Real Equation

A Simmons University study found that the impact of technology depends on how and why it\'s used, not on the technology itself. Management\'s implementation approach matters more than the technology. Automation imposed on staff without input or training creates resentment. But automation developed with staff input to eliminate their least favorite tasks? That creates relief.

Part 5: Getting Started (Free or Nearly Free)

Before you buy anything new, audit what you already have. Most libraries are underusing the automation features built into their existing systems.

ILS Features You Probably Have But Are Not Using

• Automated notices: Overdue, courtesy, hold availability via email, SMS, and phone. Most ILS platforms support all three. Many libraries only use one or two.
• Scheduled reports: Configure your most-used reports to generate and email automatically on a schedule.
• Batch operations: Batch weeding, batch record editing, batch deletion. If you're processing records one at a time, check whether your ILS supports batch mode.
• API integrations: Koha supports REST APIs and plugins for connecting to third-party services. Other modern ILS platforms offer similar capabilities.

Free Workflow Automation Tools

• IFTTT: Free plan allows 2 custom automations with unlimited executions. Over 900 app integrations. Good for simple "if this, then that" workflows like auto-sharing blog posts to social media.
• Zapier: Free plan includes 100 tasks per month. About 3,000 app integrations. Good for connecting library tools that don't natively talk to each other.
• Microsoft Power Automate: Free plan with 500 actions per day. Good option for libraries already in the Microsoft 365 ecosystem.
• Google Apps Script: Free for Google Workspace users. Automates tasks across Gmail, Sheets, and Docs.

Email Newsletter Automation

If you're still manually emailing patron newsletters:

• MailerLite: Free for up to 1,000 subscribers and 12,000 monthly emails. Includes automations, signup forms, and landing pages.
• Mailchimp: Free for 500 emails to 250 contacts per month.

Where to Start This Week

1. Audit your ILS. Make a list of every automated feature it supports. Check which ones you\'ve configured and which you haven\'t. Focus on automated notices and scheduled reports first.
2. Identify your most repetitive task. What do you or your staff do every day or every week that\'s the same every time? That\'s your automation candidate.
3. Try one free tool. Set up a single IFTTT or Zapier automation. Start small: auto-share new blog posts to social media, or auto-generate a weekly circulation report email.
4. Talk to your staff. Ask what tasks they find most tedious and repetitive. The best automation targets come from the people doing the work, not from vendor sales pitches.

Part 6: What Should Never Be Automated

Automation is a tool. It's good at some things and bad at others. Knowing where the line is matters more than knowing the latest product.

Reference and Readers' Advisory

Readers" advisory has been described as "one of the most intimate library services." It\'s most satisfying when patrons know their advisor responds from personal reading experience. During the pandemic, 83% of users preferred detailed, thoughtful responses over quick turnaround, confirming the value of human exchange in reference work.

Some patrons can't be served by online forms or chatbots at all: those with limited digital access, limited English proficiency, or complex information needs that require a conversation to untangle. These are often the patrons who need the library most.

Programming and Community Engagement

Story time, author events, job search workshops, ESL classes, summer reading programs. These require human judgment, relationship-building, and cultural sensitivity. You can automate the registration form. You can't automate the event.

The Patron Who Just Needs a Person

Some people come to the library because it\'s the only place left where a real person will help them with a problem. Navigating government websites. Filling out forms. Understanding a letter from a landlord. You won\'t find these in any job description, but they're library services in practice. A chatbot will never be enough for the patron who needs someone to sit down with them and say "let me help you figure this out."

Humans First, Machines Second

Automation won\'t save your library. It won\'t replace your staff or solve your budget problems or your staffing shortages. Anyone who tells you otherwise is selling something.

What automation can do is take specific, repetitive, high-volume tasks off your staff\'s plate so they can spend more time on work that requires human judgment, human empathy, and human connection. That\'s not a transformation. It's a trade. A useful one, but only if you choose the right tasks to automate and keep humans at the center of everything else.

If you do nothing else this week, try these:

1. Audit your ILS automation features. You\'re almost certainly paying for capabilities you're not using. Automated notices and scheduled reports are the lowest-hanging fruit.
2. Ask your staff what they want automated. Not what a vendor thinks should be automated. What the people doing the work every day would gladly hand off to a machine.
3. Protect the human stuff. Reference, readers" advisory, programming, and the patron who just needs someone to sit with them. That\'s not a cost center. That's why the library exists.