Can AI Do Instructional Design? – Strengths and Limits
You've probably seen the demo. Someone types "create a course on workplace safety" into a chat window, and thirty seconds later there's an outline, objectives, a script and a quiz. It's impressive, and it raises a fair question for anyone who designs training for a living: if the machine can do all that, what's left for you? Can AI really do instructional design, or only some of it?
AI can do a large share of the production side of instructional design, but it can't do the whole job. AI for instructional design is excellent at drafting objectives, outlining content, writing scenarios, generating assessments and reorganizing source material. It struggles with the parts that depend on context it can't see: diagnosing the real performance problem, deciding whether training is even the answer, reading stakeholders and knowing your learners. Think of it as a fast junior designer with a great vocabulary and no job experience. With a skilled designer directing it, the combination is faster and often better than either one alone.
What Parts of Instructional Design Can AI Handle Well?
AI handles the structured, text-heavy parts of design well, especially when you give it good source material and clear instructions, because these are tasks where a pattern exists and the model has seen thousands of examples of it.
Turning Source Material into Structure
Give a model a dense procedure, a policy manual or a transcript of an SME interview, and ask it to pull out the tasks, decisions and common errors. The analysis is quick and usually thorough, and this step alone can save hours of highlighting and sticky notes.
Writing Objectives
Ask for measurable objectives at a specific level of Bloom's taxonomy and you'll get usable drafts. The model knows to avoid vague verbs like "understand" and to describe observable behavior, so you still need to check that each objective matches the real job, but the wording is rarely the problem.
Scenarios and Practice Activities
Scenario writing is where AI often surprises people. Ask for a branching scenario where a new supervisor handles an employee who keeps arriving late, with three realistic choices at each point and consequences for each, and you'll get a solid first pass. It can also produce variations, so learners don't see the same situation twice.
Assessments and Feedback
Quiz questions, distractors and feedback messages are a strong fit, since the model is good at writing plausible wrong answers, which is one of the more tedious parts of assessment writing. Your job is to make sure the "right" answer is right in your organization, not just right in general.
Where Does AI for Instructional Design Fall Short?
AI falls short wherever good design depends on information that isn't in the prompt, which is a bigger category than most demos suggest, and it's where experienced designers earn their keep.
- Performance consulting. A request for "a course on customer service" might really be a staffing problem, a broken process or a bad incentive. AI will happily build the course. A designer asks why scores dropped in the first place.
- Stakeholder management. Knowing that the regional VP cares about one metric and legal cares about another, and finding a design that satisfies both, is human work.
- Learner empathy. The model doesn't know that your warehouse crew takes training on a shared tablet during a 15-minute window, or that the last rollout left people annoyed.
- Accuracy in your context. It knows general best practice. It doesn't know your company's specific rules unless you give them to it, and it may fill gaps with confident guesses.
- Deciding what to leave out. Strong training is lean. AI tends to include everything it can think of, and cutting is a judgment call.
Cathy Moore's action mapping approach is a useful lens here, because it starts with a business goal and asks what people need to do, not what they need to know. AI can help at every step of that process, but someone has to insist on starting with the goal. A model asked for "a course about X" will give you a course about X, information dump included.
Where Does AI Help Across the Design Process?
AI fits into every phase of the ADDIE model (Analysis, Design, Development, Implementation, Evaluation), but its role changes from phase to phase, as the practical split in the table below shows.
| Phase | Where AI helps | Where the designer leads |
|---|---|---|
| Analysis | Summarizing documents, drafting interview questions, listing tasks | Defining the real problem and whether training fits |
| Design | Objectives, outlines, scenario ideas, assessment plans | Choosing the approach, sequence and what to cut |
| Development | Scripts, quiz items, voiceover, images, code snippets | Quality, accuracy, brand voice and usability |
| Implementation | Launch emails, job aids, FAQs, manager guides | Timing, rollout plan and change management |
| Evaluation | Survey questions, summarizing open comments, spotting patterns | Deciding what the data means and what to change |
The pattern is easy to see: AI drafts and the designer decides. Development is where the time savings pile up, which is why so many teams start there. If you want to estimate what that looks like on your own projects, our guide to AI course development time savings walks through the numbers phase by phase.
What Does a Good Designer-Plus-AI Workflow Look Like?
A good workflow puts the designer at the start and the end of every step, with AI doing the bulk drafting in the middle. Here is one I like for a typical compliance or process course:
- Meet with the stakeholder and agree on the business goal and the measurable behavior change. No AI yet. This conversation sets everything else up.
- Feed the AI your source documents and SME notes. Ask it to list the tasks, decisions and common mistakes, and to flag anything that seems unclear or contradictory.
- Review that list with your SME. Correct it, cut the nice-to-know items and mark the high-risk tasks.
- Ask the AI for objectives and a module outline based on the corrected list. Request two or three structure options.
- Pick a structure and ask for practice activities first, before any content screens. This keeps the course focused on doing.
- Draft content screens only for what learners need to complete the practice.
- Run a full accuracy and tone review, then an SME sign-off.
Notice the order: practice comes before content. That's a design decision AI won't make on its own unless you tell it to. Once you do, it follows instructions well.
Where AI still needs you: deciding what problem you're solving, what learners must do on the job, and what can safely be cut. Every other step gets faster with a good draft in hand.
How Do You Prompt AI for Better Design Work?
You get better design work from AI by giving it the same briefing you'd give a new team member: who the learners are, what they need to do, what constraints you're under and what good looks like. Short prompts get generic answers, while specific prompts get drafts you can actually use.
Here's a before-and-after comparison. A weak prompt reads: "Write learning objectives for a course on handling customer complaints." The result will be fine and forgettable, full of phrases like "understand the importance of empathy."
A stronger prompt reads: "You are helping an instructional designer. Audience: call center reps in their first 90 days, mostly part-time, taking training on a laptop between shifts. Business goal: reduce escalations to supervisors. Source notes are pasted below. Write four measurable objectives using observable action verbs. Avoid 'understand,' 'know' and 'learn.' For each objective, name the on-the-job behavior a supervisor could observe."
The second version gives the model a reason, an audience, a constraint and a format. The output usually needs light editing instead of a rewrite, and it also tends to surface behaviors you can turn straight into practice activities, which is where a course earns its value.
A few other habits make a real difference:
- Ask for options, not answers. "Give me three ways to sequence this module and the trade-off of each" gets you thinking, not just copying.
- Ask the model to critique its own draft against a standard you name, such as Gagné's events of instruction or your team's quality checklist.
- Paste in an example you like and ask it to match the style.
- Keep a shared prompt library so the whole team benefits from what works.
How Is AI Changing the Instructional Designer's Job?
AI is unlikely to replace instructional designers, but it is changing what the job looks like: the parts that are mostly typing are shrinking, and the parts that are mostly thinking are growing.
Designers who spent most of their week writing screen text will spend more of it on analysis, stakeholder work, testing with learners and measuring results. That's arguably the more interesting half of the job anyway, and it's also the half that clients and managers value most, because it's what turns training into better performance.
The skills worth building now are clear: writing precise prompts, giving models good examples, checking output with a critical eye and explaining to stakeholders why a draft isn't a finished course. Those sit on top of the design skills you already have. They don't replace them.
There's also a new skill that's easy to overlook, which is knowing when not to use AI. A sensitive topic, a tiny audience or content that changes daily may be faster to handle the old way. Good designers will make that call quickly, case by case.
The Designer Still Holds the Pen: What AI for Instructional Design Really Means
So, can AI do instructional design? It can do much of the drafting, and it does it fast. What it can't do is decide what the training is for, who it's really for and what success looks like, and that is still your job.
The practical takeaway is to treat AI for instructional design as a capable assistant you direct, not a replacement you hand the project to. Start every project with the business goal, give the model your real source material and keep a named review step at the end of each phase.
Try it on your next small project by using AI for the task list and the first scenario draft, then comparing the result to your usual process. You'll quickly see where it earns a permanent spot in your workflow.