Why everything just changed
Instructional design has always been a writing job with a design problem inside it. You research, you plan, you write, you review, you build. That shape has not changed. What has changed is that every stage of it now has an AI tool that can generate a plausible first draft in seconds.
That is not the same as good design. But it is a new raw material — and learning to use it, and to judge it, is now a core ID skill.
By the end of this unit- Describe how Copilot changes each stage of the five-stage ID workflow.
- Identify the three categories of ID task where AI adds the most value.
- Apply a quality check to AI-generated learning content.
The AI shift is a tool shift, not a role shift
The arrival of Copilot and large language models in the workplace is the same kind of shift as the arrival of word processors, then authoring tools, then LMS platforms. Each one accelerated production. None of them replaced the human judgment that sits underneath the work.
What this series covers
This is a series of eight courses on designing learning in an environment where your learners use AI tools daily — and where you, as a designer, have AI assistance at every stage. The two things are related. You cannot design for Copilot-using learners without understanding how Copilot works in practice.
Design with AI
Use Copilot to accelerate needs analysis, storyboarding, knowledge check generation, and scenario writing.
Design for AI-using learners
Write objectives, scenarios, and evaluation for people whose jobs now involve Copilot, Cowork, and agents.
Maintain design judgment
Know what to accept, what to fix, and what to discard from every AI output — and why.
The AI-augmented workflow
The five-stage ID workflow does not disappear when you add AI. Plan, write, check, build, sign off. But the time spent at each stage, and where your attention should land, shifts significantly.
Where AI does its best work
AI is fastest and most valuable in the middle of the workflow — where you need a large volume of words to react to. It is least helpful at the beginning, where you need information about a specific organisational context that no AI has seen.
Plan and storyboard — AI generates the outline
Write — AI produces the first draft
Proofread — AI assists, human decides
Build — AI does not replace authoring skills
QA and sign-off — the unchanged stage
In your current practice, which stage takes the most time? Is that the same stage where AI would most help you — or a different one?
What Copilot can and cannot do
Being clear about Copilot's capabilities prevents two opposite mistakes: over-reliance, which produces generic content; and under-use, which leaves real productivity gains on the table.
The capability map
- Generating multiple versions of the same knowledge check question
- Expanding bullet points into trainer-talk paragraphs
- Writing first-draft scenarios from a role description and task
- Reformatting content from long-form to screen-sized chunks
- Suggesting plausible distractors for multiple-choice questions
- Generating alt text and audio script stubs from image descriptions
- Organisation-specific context — AI cannot know your client's processes
- Regulatory accuracy — AI may state out-of-date rules confidently
- Learner persona nuance — AI generates demographic stereotypes, not real constraints
- Instructional sequence — AI will generate logical-looking sequences that are pedagogically wrong
- Whether the objective is the right one — only a needs analysis can tell you this
- Whether the scenario is realistic — only someone who does the job can confirm
- Whether the knowledge check tests the right thing — requires reading the whole unit
- Whether the tone is right for this audience — requires knowing the audience
Every AI-generated passage should be treated as a first draft written by a confident generalist who has never met your learner. Your job is to rewrite it as if you have.
Quality control in AI-assisted design
The risk of AI-assisted ID is not obvious errors — it is confident-sounding content that misses the point. The quality check process needs to catch this before the build stage.
The three-question review
Apply these three questions to every AI-generated piece of content before it goes into a storyboard:
Does this teach the objective — or just describe the topic?
AI tends to summarise subject matter rather than build toward a performance outcome. If the content describes what Copilot does rather than what the learner will do with it, the objective is not being served.
Is the worked example specific to this learner's context?
Generic examples ("a project manager at a large organisation") do not build the transfer that makes learning stick. Replace them with examples from the actual role and workplace of your target learner.
Does the knowledge check require recall — or can it be answered by re-reading?
AI-generated knowledge checks often test whether the learner read the screen. Real retrieval requires the learner to apply knowledge they have already moved to memory. If the answer is visible on the page, the check is a reading comprehension exercise, not a learning check.
Which of these is the strongest reason to edit AI-generated instructional content?
End of Unit 1
You should now be able to:
- Describe how AI tools change each stage of the five-stage ID workflow.
- Identify where Copilot adds the most value and where human judgment is irreplaceable.
- Apply a three-question quality check to AI-generated content.
Continue to the unit review.
Unit review
Four questions. Choose the best answer in each.
Which stage of the ID workflow is AI currently least able to replace?
What is the main risk of AI-assisted instructional writing?
Which of these is AI best suited to help with during the storyboard stage?
The three-question review checks for: does it teach the objective, is the example specific, and…?
End of module
You have completed Course 01: AI as Your ID Partner. Next: Writing Outcomes for AI-Augmented Roles — how to write learning objectives for people whose jobs now involve Copilot and agents.