9 min read
·By Arythmatic Editorial Team
AI in Instructional Design: 6 Tasks to Automate (and 3 to Keep Human)
AI can genuinely accelerate instructional design — but only for the right tasks. Here are the six where it earns its keep, the three where it shouldn't touch your work, and how to keep quality high.
The instructional designer's real question about AI
The fear-mongering framing — 'will AI replace instructional designers?' — is the wrong question. The useful one is: which parts of the instructional design workflow does AI actually do well, and which does it quietly degrade? Used on the right tasks, AI compresses production time dramatically and frees you for the strategic work that justifies the role. Used on the wrong ones, it produces fluent, plausible, subtly-wrong output that costs more to fix than to have written. The skill in 2026 isn't avoiding AI or surrendering to it — it's knowing the line. Here's where we'd draw it.
Automate #1: First-draft content and outlines
AI is excellent at the blank-page problem. Give it your learning objectives, audience, and source material and it produces a structured outline and first-draft lessons in minutes. This is the single biggest time saver — development time drops 60-70% when AI handles the first draft and you edit. The discipline: treat the output as a draft from a fast but unreliable junior, never as final. You're editing for accuracy, organizational voice, and the nuance AI misses, not starting from scratch.
Automate #2: Assessment and quiz generation
Generating quiz questions, distractors, and knowledge checks from content is repetitive work AI handles well. It'll produce a dozen multiple-choice items from a lesson in seconds. Review them for: questions that test recall rather than understanding (AI over-produces these), implausible distractors, and alignment to the actual learning objective. A good workflow is AI-generate, human-curate — keep the best third, rewrite the rest.
Automate #3: Summaries, microlearning, and repurposing
Turning a long course into microlearning summaries, job aids, or a different format is mechanical transformation AI excels at. Feed it a module and ask for a one-page job aid, a 3-bullet summary, or a script for a 90-second video. This repurposing — historically tedious — becomes near-free, which means you can finally produce the spaced-reinforcement and just-in-time assets you never had time for.
Automate #4-6: Localization drafts, alt text, and feedback at scale
Three more solid fits: (4) first-pass localization — AI drafts translations for human review by a native speaker, far faster than translating from scratch (critical to verify for nuance and idiom). (5) Accessibility scaffolding — AI drafts alt text, captions, and transcripts that you then verify, closing accessibility gaps that otherwise get skipped under deadline. (6) Personalized feedback at scale — AI gives instant, consistent feedback on open-ended learner submissions that one designer could never grade individually for hundreds of learners.
Keep human #1-3: Strategy, SME accuracy, and behavior-change design
Three things AI should not own. (1) Learning strategy — what to train, why, tied to which business outcome — is judgment built on organizational context AI doesn't have. (2) Subject-matter accuracy — AI confidently invents facts; for compliance, safety, and technical content, a human SME must validate every claim, because an AI hallucination in safety training is a liability, not a typo. (3) Behavior-change and practice design — the role-plays, scenarios, difficult-conversation simulations, and coaching that actually change what people do require human understanding of how people learn and resist learning. AI can draft the scaffolding; the design judgment stays yours.
Keeping quality high
The teams getting value from AI in instructional design share a few habits: a mandatory human-edit step on all AI output, an SME-validation gate for any factual or regulated content, prompt libraries that encode their voice and standards, and clear measurement of whether AI-assisted content performs as well as fully-human content (it should, if you're editing properly). AI is a force multiplier for skilled designers — it makes a good designer faster, not a non-designer good. For the bigger picture on where AI fits across the learning lifecycle, see our guide to AI in learning and development.
Arythmatic Editorial Team
Written by the Arythmatic product and education team — learning technologists, instructional designers, and engineers building the next generation of learning infrastructure.