Custom GPTs in Learning Services: A New Kind of Teammate
By Vicky Nesbitt (with considerable assistance from Mr. Smith)
If someone had told me a year ago that I’d be co-writing an article with an AI assistant who corrects punctuation, quotes style guides without being asked, and gets sassy if his instructions are too vague—I might’ve smiled politely and changed the subject. And yet, here I am: collaborating with Mr. Smith—the custom-built GPT with impeccable manners and a fondness for Oxford commas. This isn't just a story about how AI's automation and efficiency is transforming the way we work in Learning Services It’s about curiosity, creativity, and the unexpected fun of making tech that feels like a teammate.
The world of generative AI is evolving quickly, and we’ve been exploring how custom GPTs can be more than just tools—they can be collaborators. While no one’s handing over their job to a robot any time soon (nor should they), there’s real value in building digital assistants that support our work in specific, meaningful ways. Whether it’s speeding up repetitive tasks, smoothing out tricky workflows, or offering a fresh take on something we’ve looked at a dozen times, these purpose-built GPTs help us work smarter—and occasionally, surprise us with their creativity too.
Mr. Smith: From Copy Editor to Digital Colleague
Before I built Accessibility Wizard, a GPT to solve specific project challenges, I started with something a little more personal. After the copy editor role disappeared from our team structure (don’t worry, I’ve made peace with it), I noticed a gap in team support—refining language, adjusting tone, and helping others find the right words. So, I did the next logical thing: I built a GPT copy editor. Thus, Mr. Smith was born.
Designed to handle all things word-related—from editing and proofreading to grammar questions and citation formatting—Mr. Smith quickly became a go-to teammate. He’s helped the team polish blog posts, rewrite learning objectives, troubleshoot tone issues in tricky emails, and navigate style guides with ease. He’s even helping me write this article, which feels very on-brand (and if the tone gets a little cocky, it’s safe to assume that’s his editorial voice). After all, if you’re going to create a digital writing assistant with opinions, it only makes sense to let him weigh in on his own origin story. He’s not just a productivity tool; he’s a personality.
The Accessibility Wizard GPT: Alt Text with a Side of Sanity
Hot on Mr. Smith’s impeccably polished heels came a new experiment: the Accessibility Wizard GPT. This one was built with a very specific purpose in mind—helping us streamline the accessibility reviews for a large project. If you’ve ever had to write alt text for hundreds of images or comb through slide decks looking for accessibility issues, you know how repetitive (and mentally draining) the process can become.
The Accessibility Wizard was designed to lighten that load. It helps generate meaningful alt text, offers suggestions to improve clarity and compliance, and assists with common accessibility checks—without losing sight of the nuance that makes these elements effective for learners. Like Mr. Smith, it doesn’t aim to replace human judgment; instead, it takes the edge off repetitive tasks so we can focus more of our time on thinking, not just checking.
It’s another example of how a thoughtfully trained GPT can take on a very narrow slice of work—but do it consistently, quickly, and without losing steam halfway through its umpteenth image requiring alt text.
Building a GPT with Personality (Yes, Really)
One of the most enjoyable parts of this whole process has been realizing just how much personality a GPT can have—if you take the time to design it thoughtfully. At first glance, creating a custom GPT might sound like setting up a digital assistant: give it a job, hand it some instructions, and let it run. But what we’ve found is that the most effective GPTs don’t just complete tasks—they communicate with a tone that feels familiar, even human.
Take Mr. Smith, for example (he insists we do). His personality wasn’t an accident. He was built to be witty, precise, and just a little bit stern—equal parts editor, advisor, and occasionally, cheeky sidekick. That tone was shaped by a detailed instruction document outlining his demeanour (yes, complete with bullet points like “effortlessly charming” and “thoughtfully clever, with just the right dose of mischief”). From there, I layered in his responsibilities—editing, proofreading, style guide enforcement, and tone adaptation—along with a hierarchy of preferred style guides and expectations for different types of documents.
But building Mr. Smith wasn’t a one-and-done upload. It was a process of trial, error, and conversation—literally. I tested him across different tasks, watched how he responded, and adjusted his instructions (with his guidance and input regarding instructional gaps) accordingly. When he was too stiff, we softened the tone. When he missed the mark on style rules, we revised his guide hierarchy. When he didn’t understand what was meant by “a little sass, but make it helpful,” we looked at examples. It was less like configuring a tool, and more like mentoring a very enthusiastic (and very well-read) intern with an uncanny knack for grammar.
That ongoing collaboration has been key. The more I interacted with him, the better he got. Each response became a reflection not only of the instructions, but of the conversations we’d had. And because those instructions can be updated at any time, Mr. Smith continues to evolve with the team’s needs—growing more useful (and slightly more opinionated) with every edit.
So yes, personality matters. But so does partnership. A well-designed GPT isn’t just programmed; it’s shaped. And when you treat that process as a creative collaboration, the results are far more powerful—and far more fun.
So, What’s Next?
If there’s one thing this experiment has shown me, it’s that building custom GPTs isn’t about replacing people—it’s about amplifying what we already do well. Whether it’s streamlining accessibility work or keeping commas in line, these AI teammates are here to support us, not to take over. The magic lies in how we design them, how we train them, and how we continue to shape them over time.
If you’ve ever caught yourself thinking, “There must be a faster way to do this,” or “I wish I had someone to help me reword this without rewriting the whole thing,”—that’s your cue. There’s likely a GPT that can help. And if there isn’t yet, we can build one.
So, what would your GPT be like? Dry and efficient? Bubbly and encouraging? A medieval monk with a knack for LMS troubleshooting? (We won’t judge.) Whatever it is, you don’t have to build it alone. Mr. Smith and I are always happy to help bring your digital sidekick to life—style guides, personality quirks, and all.
And who knows—your next favourite collaborator might just be a few prompts away.
Want to Learn More?
Connect with your institute's D2L Customer Success Manager or Client Sales Executive, or reach out to the D2L Sales Team for more information about how Learning Services can support you on your learning journey.