A library director in Atlantic Canada asked us for something after a consultation this spring. Not the strategy deck. Not the debate about what AI means for the profession. The plain version: what does a frontline colleague actually do with this on a Tuesday, which patron conversations does it show up in, and how does someone who has never opened one of these tools take a first step without feeling foolish?

Here it is. If you lead a library and want the everyday answer for your staff, this is the piece to forward.

 The useful way to think about a generative AI tool at a service desk is as a very fast, sometimes confident assistant who has read a great deal and remembers none of your context. That colleague is excellent at first drafts, outlines, plain-language rewrites and “explain this to me quickly,” and unreliable about specific facts, titles, names and numbers. 

So the pattern that works is not a new service. It is the work you already do, with a drafting step added in front of your own judgment. If a tool is used, feel free to tell the patron however, make sure to emphasize that you added the final judgment layer over it. Transparency maintains trust you already have.

Two habits carry the whole thing. In our GenAI Foundations course we teach an approach called READ for working with these tools, and a quick check called SCAN for judging what comes back. You do not need the course to start. You need the instinct behind them: be specific about what you want, and never let an unverified fact leave your hands.

Each of these fits in the gaps between patrons. Each uses a prompt shaped by CRAFT, the prompting method we teach: give the tool Context, a Role, a clear Action, the Format you want, and then Try again with what you learned.

1. Turning a programme idea into a plan. “Imagine you are a public library programmer. You run a 45-minute smartphone basics class for seniors, most of whom have an Android phone and have never used the camera. Give me a session outline with three hands-on activities and one thing to send home. Use plain language.” You will change half of it because this tool does not know your community like you do. You will also have a plan in five minutes instead of forty, which is especially useful when you are building entirely from scratch.

2. Accessible to everyone. Signs, shelf talkers, the paragraph for the newsletter, the social post about the closure on Monday. Paste your draft and ask for it at a grade six reading level, or in a warmer tone, or in 40 words. The tool is good at this because there is nothing to get wrong except the facts, and the facts are yours.

3. Readers’ advisory prep. “A patron loved The Thursday Murder Club. Suggest eight read-alikes published since 2015 with one line each on why.” Then check every single title in your catalogue before you say a word, because these tools invent books that do not exist, with confident authors and plausible covers. Remember this is an overconfident (and sometimes wrong) assistant. This is the SCAN moment: the list is a starting point, the catalogue is the truth.

4. Getting through the long document before the meeting. A new provincial policy, a vendor contract, a forty-page report. “Summarise this in ten bullets, then list anything that creates a new obligation for a public library.” Read the summary, then read the parts it flagged. You still read the document. You read it knowing where to look.

5. Drafting the FAQ for the thing that changed. New printing system, new hold shelf, new hours. Describe the change and ask for the ten questions patrons will ask and a two-sentence answer to each. Fix the answers. Post them. Nobody on the desk answers the same question thirty times that week.

 AnothYou You do not have to bring AI to the desk. Patrons are bringing it. These are some of the five conversations we hear about most often, and a reasonable response to each.

“Can you help me use ChatGPT for my résumé?” Yes, the same way you help with any tool on a public computer. Sit with them. Show them how to paste in the job posting and their old résumé and ask for a tailored version. Then remind them to read every line, because it will invent a qualification if it thinks the job wants one. And give the patron some additional context — everyone will also use ChatGPT to create a résumé so its even more important now to think deeply about what skills the job is asking for and how your experience highlights it. Patrons now have a machine that can sharpen how they communicate their qualifications.  

“I asked the AI and it said this. Is it true?” This is the best question a library can be asked, because answering it is the oldest thing libraries do. Fact-check it together. Show them where you looked. You have just taught information literacy without calling it that.

“Can you help me with this government form?” Help, yes. Paste their personal details into a public tool, no. Use the tool to explain what the form is asking in plain language, generically, and let the patron fill in their own information themselves. The Government of Canada’s own guidance on generative AI says the same about personal and sensitive information, and it is a good line to be able to quote. See here for this guidance: Guide on the use of generative artificial intelligence — Canada.ca

The teenager with homework. Most schools now have a position on AI, and the library’s job isn’t to enforce it. A useful stance is simply the honest one: AI can be a good study partner, explaining a concept in plain language, quizzing you on it, helping you get unstuck, but if it writes the assignment, that’s no different from handing in someone else’s work. Most teens already know this; they don’t need a lecture, just a straight answer. The more interesting thing to offer, if the conversation allows, is a reason that has nothing to do with rules. School is one of the few times in life when someone is paid to read what you think and help you say it better. An AI answer is an average of how everyone else on the internet writes, which is exactly why it sounds like nobody in particular. Your own way of putting things, awkward spots and all, is the thing worth developing, because it’s the only part nobody else can produce.

The person worried about a scam. Voice clones, fake letters, messages that sound exactly like a grandchild. They are right to worry, and the library is the right place to bring it. Walk through what to check, and point them to your next digital literacy session. If you do not have one on this yet, you have found your next programme.

1. Nothing personal about a patron goes into a public tool. Not a name, not a card number, not the question they asked at the desk. Describe the situation generically if you need help with it.
2. You are the author. Every name, date, title, number and quotation gets checked before it leaves your hands. The tool drafts; you publish.
3. Say so when it matters. If a notice was translated by a tool, add a line saying a human check is welcome. Patrons trust libraries because libraries tell them things like that.

Week one. Pick one task from the list above and one tool you already have. If your organisation runs Google Workspace or Microsoft 365, there is very likely a tool already inside it. Spend fifteen minutes a day on that one task. Use CRAFT. Keep a note of the prompt that worked.

Week two. Bring your one example to a team meeting. Ask who else has one. Start a shared document of prompts that worked, sorted by task. That document is the beginning of your library’s AI practice, and it cost nothing.

If you would like company for the first step, consider starting a learning circle in your own library: a few staff working through an introductory AI course together, an hour a week. It’s a good format precisely because nobody has to be the expert. If you have questions about a good introductory course, please reach out to us at info@engrsolutions.ca, we’re happy to point you in the right direction!

In our Mount AI model we use with library leaders, all of this is the first two stages, Explore and Experiment. You do not need a policy to explore. You do need one before you run a pilot that touches patron services, and by the time you have two weeks of examples in a shared document, you will know what that policy needs to say.

We are more than consultants, we are your AI transformation partners.

• Strategy and policy guidance
• AI literacy and workforce training
• Pilot design and technical implementation
• Change management and digital leadership

Whether your team is opening a tool for the first time or planning its first pilot, we will help you climb Mount AI while staying true to your mission and the communities you serve.

Let's put AI to work. Visit engr.solutions or write to info@engr.solutions.

1. Government of Canada, Guide on the use of generative AI
2. Toronto Public Library, AI Learning Circles
3. ENGR Solutions, A call to action for Canada's public libraries: leading the nation in AI literacy
4. ENGR Solutions, Mount AI