AI is the Future of Instructional Design
Background
I’m a Senior Instructional Designer with an MA in Technology and Learning Design. I work in the utilities industry. Over the last year I’ve been helping my department integrate AI tools into our workflows. After a brief stint with a localized Amazon product (essentially a third-hand version of ChatGPT), we moved to Microsoft Copilot. So far, so good. At home, for my own projects, I use Claude, Grok, and Gemini.
I’m unapologetically all-in on AI (I even had Grok review this before posting it), and I believe AI has made instructional designers more important, not less. I’m also worried that the profession will start losing valuable people, either to overzealous management or to practitioners who are intimidated by the tools.
What I'm Seeing
Most companies that roll out AI tools simply hand them out. Training, when it exists, is mostly a long list of things not to do. The reason is straightforward: relatively few people inside the organization have a real grasp of what these tools are actually good for. AI is still new and moving fast.
The result is that many employees who have had access for months are only now realizing they can use it to summarize email chains, draft messages, or sort their inboxes. Based on common usage, you’d think AI was primarily an email tool.
There are also people who would still be using rotary phones if they could. Some on principle, others because their first encounter produced garbage output and they wrote the technology off entirely. Both groups are missing the capability curve, or they’re approaching it so slowly they risk being left behind. I keep coming back to a line I saw recently that feels exactly right: “AI will not take your job, but someone who knows how to use AI will.”
How AI Should Be Used
If you don’t know what RAG is, you should learn. Understanding retrieval-augmented generation, along with solid prompting, will change both the speed and the accuracy of your work.
In oversimplified terms: imagine you have a brilliant new assistant fresh out of a master’s program. You ask her to write a knowledge check on the company’s logout procedures. You can either rely on what she remembers from school and general industry knowledge, or you can hand her the actual company documentation and have her work from that. The second option is obviously better. That is essentially what RAG does. You give the model the real source material, then direct it. Here’s how I typically start:
Prompt: "You are a highly skilled instructional designer. I need you to help me create a knowledge check for our company’s logout procedure. I’m uploading SME-vetted documentation on this topic with this prompt. Analyze the contents. Ask me any clarifying questions you have. Then create a set of learning objectives. Prefer Mager-style objectives where feasible and Bloom’s taxonomy for everything else. I will use these objectives to build the knowledge check."
Once the clarifying questions are answered and the objectives are reviewed (preferably by an SME), I move to the next prompt:
Prompt: "Using the learning objectives and the documentation you already have, construct a set of knowledge check questions and activities. For multiple-choice items: make the distractors similar in length to the correct answer, ensure they are plausible but clearly wrong, randomize the options, and provide four choices for each question."
You still read every item carefully. The point is not to hand over judgment. It is to let the model handle the heavy lifting of analyzing large amounts of source material quickly and accurately. That is one of its most powerful functions for instructional designers.
A Pre-AI Project
Just before I had reliable AI tools, I built a full training program from roughly 140 pages of source material spread across seven formats: Excel, Word, PowerPoint, scans, text documents, handwritten notes, and transcribed interviews. Sorting, updating, segmenting, and sequencing the material alone took weeks. Developing objectives, assessments, learning activities, ILT/VILT materials, and eLearning took about three months in total.
Once I had AI tools, I fed the entire set of source documents and the finished program into the model and asked it to evaluate the design. The analysis took less than thirty minutes. The program largely held up, but both the SMEs and I had missed several issues. I was able to regenerate and refine everything in a few hours, not months.
My Advice
Learn to use AI to accomplish the tasks of the job. The job itself has never been the tasks. The job is judgment, and judgment is more important than ever. AI will create whatever it is asked to create. The amateur will create training where none is needed. The professional decides what is worth building, what the source material actually means, and whether the result is accurate, safe, and useful in a high-stakes environment.
I actually wrote a book on this topic, but I DO NOT suggest you buy it.
In some respects, it's already out of date.
Perhaps the most useful part of the book are the appendices, and you can access them here for free at the following website: www.indieprocess.com
Once again, that's a free resource. I'm not flogging my book.