Prompt Library
24 structured prompts for the full ID workflow. Each follows the five-element anatomy: role + audience + task + format + constraints. Copy the prompt, fill the [bracketed] fields with your project specifics, and use as a starting point.
Phase 1 — Needs Analysis & Discovery
01 — Needs analysis brief
I am an instructional designer starting a needs analysis for a [module topic] programme. The target audience is [role description] in [organisation type]. The performance gap is [describe the gap — what people are currently doing vs. what they should be doing].
Generate 8 interview questions I can use with subject matter experts to understand: the context in which people need this skill, the current obstacles to performance, and what good performance looks like from a manager's perspective. Questions should be open, not leading. Do not ask about training preferences — focus on the work itself.
Edit the questions to remove any that your SME would consider obvious or condescending given their seniority and domain expertise.
02 — Audience research synthesis
I have completed SME interviews for a [module topic] learning programme. Here are my interview notes: [paste notes].
Synthesise these notes into: (1) the three most important things this audience already knows, (2) the three most common misconceptions or gaps, and (3) the single most important observable behaviour change the learning needs to produce. Format as three short paragraphs, not bullet lists.
Review the synthesis carefully — AI may conflate distinct SME opinions or over-weight the most recent input in your notes.
03 — Job-to-task breakdown for AI-augmented roles
I am designing learning for [job title] roles who now use Microsoft Copilot as part of their daily work. Their primary job tasks are: [list 5–8 tasks].
For each task, describe: (1) which part of the task Copilot can assist with, (2) what the human must still do and cannot delegate to AI, and (3) what new skill the role now requires that did not exist before Copilot. Format as a table with columns: Task | AI-assisted element | Human-only element | New skill required.
Validate the table with an SME from the actual role. AI will produce plausible task breakdowns that may not match how the work is actually done in your organisation.
04 — Content audit for existing material
I have existing training material on [topic] that I need to update for an audience that now uses Copilot in this area. Here is the existing content: [paste content].
Review this content and identify: (1) sections that are still valid without change, (2) sections that need updating because the task is now partly AI-assisted, and (3) sections that should be removed because the AI now does this task. For sections in category 2, suggest what the updated learning objective should be.
Phase 2 — Objectives & Structure
05 — Learning objective writing (AI-augmented role)
Write three learning objectives for a module on [topic] for [audience]. The module is about [describe the work the learner needs to do, including where Copilot is involved].
Each objective must: use an observable, measurable verb; name the thing the learner will do it to; and specify the condition (including whether Copilot is the tool they will use or evaluate). Format: "Given [condition], the learner will [verb] [object]." Do not use the verbs: understand, know, appreciate, or be aware of.
Check each objective against the three-layer model (before AI, during, after). A set of three objectives should cover at least two layers.
06 — Module outline generation
Generate a [number]-section outline for a [duration]-minute module on [topic].
Audience: [describe audience — role, experience level, prior knowledge, device context].
Learning objective: [paste the objective].
Constraints: Include one worked example section and one short review. Do not include a general introduction to [topic area] — the audience already knows what [tool/concept] is. Each section should be described in a single sentence that names the learner's task in that section, not the content covered.
Reject any section described as "Introduction to [topic]" — this is a content description, not a learner task. Reprompt with the exclusion restated.
07 — Persona builder for AI-using learner
Create a learner persona for a [role title] who is about to start using Microsoft Copilot at work. Include: name and brief role summary, the specific tasks they do most frequently, their prior experience with AI tools (honest — likely low to moderate), their main source of anxiety about Copilot, the time constraints they operate under, and one thing they would stop doing at work if AI could reliably do it for them.
Do not make the persona optimistic about AI. Make them realistic — someone who will engage if the learning is relevant and quit if it is not. Do not include demographic information.
The anxiety element is the most instructionally useful part. Build at least one scenario that addresses this persona's specific anxiety directly.
08 — Learner journey map for Copilot onboarding
Map the learner journey for a [role title] going through Copilot onboarding in a [organisation type]. Include three phases: before the training (what they know, feel, and expect), during the training (the learning experience moments), and after the training (the first 30 days of use). For each phase, identify: the key touchpoint, the primary emotion, and the biggest risk to successful transfer. Format as a table.
Phase 3 — Content & Scripting
09 — Screen text — trainer-talk paragraph
Write a [word count]-word trainer-talk explanation of [concept or principle]. This text will appear on a single screen in a workplace eLearning module.
Voice: Second person, active, warm and direct. No hedging ("it is worth noting that…"). No bullet points — flowing prose only. Audience: [describe audience]. Include one specific example from [industry/role context]. Maximum sentence length: 25 words. Do not open with a question.
After generating, read aloud. Any sentence you stumble over needs rewriting.
10 — Accordion item — body text
Write the body text for an accordion item titled "[accordion item title]" in a module about [topic]. The learner has just read [brief description of preceding content]. This accordion item should: add one specific, practical piece of information not covered by the title alone; use trainer-talk voice; and be under 80 words. No sub-headings inside the accordion body.
11 — Process step description
Write descriptions for each step in this process: [list the steps by name]. Each description should: explain why this step matters, not just what it is; be under 50 words; and name one thing that goes wrong when this step is skipped. Voice: direct, second person. Format each as: Step name / Description / Consequence of skipping.
12 — Worked example — Copilot scenario
Write a 200-word worked example showing a [role title] using Copilot to [task description]. Show: the prompt they wrote (specific, with context and constraints), the Copilot output they received (realistic — useful but not perfect), and the one revision they made to improve it. Do not frame this as a success story — frame it as a realistic working session. Voice: third person narrative. Context: [organisation type, industry].
Phase 4 — Knowledge Checks & Assessment
13 — Knowledge check — application level
Write a multiple-choice knowledge check question that tests this objective at application level: "[paste objective]".
Requirements: One clearly correct answer. Three plausible distractors — each should represent a common misconception or partial understanding, not an obviously wrong answer. Question stem: a scenario or applied situation, not a definition recall. Each option under 25 words. The correct answer must not be answerable by re-reading the immediately preceding screen.
Check: can the learner find the answer by re-reading the screen? If yes, rewrite as an application question.
14 — Knowledge check — evaluation level
Write a knowledge check at evaluation level for a module about [topic]. The learner should be asked to judge the quality of a piece of [AI output / prompt / learning objective / scenario] and identify what is wrong with it. Provide: the question stem (with the thing to evaluate presented in full), the correct evaluative judgment as the right answer, and three distractors representing plausible but wrong evaluations.
15 — Feedback text — correct and incorrect
Write feedback for a knowledge check question about [topic]. The correct answer is: [correct answer]. The most common wrong answer will be: [wrong answer].
Correct feedback (shown when right answer selected): 1–2 sentences. Explain WHY it is correct, not just that it is correct. Do not say "Well done" or "Correct!" — start with the explanation.
Incorrect feedback (shown when wrong answer selected): 1–2 sentences. Point toward the right answer without giving it away. Direct, not apologetic.
16 — Performance assessment — prompt submission rubric
Create a 4-criterion rubric for assessing a learner's Copilot prompt as a Level 2 performance assessment. The task was: [describe the task the learner was asked to complete with Copilot]. Criteria should cover: specificity of the outcome statement, quality of context provided, presence and appropriateness of constraints, and format instruction quality. For each criterion, provide: what a strong response looks like, what an adequate response looks like, and what an inadequate response looks like.
Phase 5 — Review, Launch & Evaluation
17 — Trainer-talk rewrite
Rewrite the following text in trainer-talk voice: [paste text].
Rules: Second person throughout. Active voice. Contractions where they sound natural. Maximum 25 words per sentence. Remove all hedging phrases ("it should be noted", "it is important to", "in many cases"). Remove all passive constructions. Keep all factual content intact — do not add or remove information, only change the voice and structure.
18 — Launch communications — manager briefing
Write a 200-word manager briefing email for a Copilot learning programme launching on [date]. The programme covers [brief description — what learners will be able to do after completing it]. The manager's role is: to encourage their team to complete it within [timeframe], to have a 10-minute conversation with each team member about how they will use Copilot in their specific role after completing the module.
Voice: direct, professional, no hype. Do not use the words "excited", "journey", or "transformational". Include one specific action the manager should take before the launch date.
19 — Level 1 survey — confidence and intention
Write a 4-question Level 1 evaluation survey for a module on [topic]. Requirements: Question 1 should measure confidence on a 1–5 scale for a specific skill (not general satisfaction). Question 2 should measure transfer intention — what will they do differently. Question 3 should ask for one thing that felt confusing, irrelevant, or wrong (open text). Question 4 should ask for one specific action they will take in the next 5 working days. Total completion time: under 3 minutes. No rating scales except Question 1.
20 — 30-day follow-up — learner check-in
Write a 3-question 30-day follow-up message to send to a learner who completed a Copilot module 30 days ago. Their stated transfer intention at the time was: "[paste their answer to the transfer intention question]".
Questions should: reference their specific intention, ask what happened (not a scale — open text), and ask what would help them use Copilot more effectively. Tone: conversational, not bureaucratic. Under 120 words total including intro.
21 — Post-launch survey — 30-day manager version
Write a 3-question survey to send to managers 30 days after their team completed a Copilot learning programme. The programme aimed to change these behaviours: [list 2–3 intended behaviour changes]. Questions should ask whether the manager has observed each behaviour change, with space for a brief example. Tone: direct, respectful of their time. Under 100 words per question. Make it completable in under 4 minutes.
22 — Lessons learned debrief
I have just completed delivery of a [type] learning programme on [topic]. Here is what I know about how it went: [paste notes on completion rates, survey results, manager feedback, production issues].
Synthesise these notes into: three things that worked well and should be repeated, three things that should be changed in the next iteration, and one structural change to the ID process that would have improved this project. Do not include generic ID advice — focus only on what the data suggests about this specific project.
23 — Accessibility review prompt
Review the following storyboard section for accessibility issues: [paste section].
Check for: missing or inadequate alt text descriptions for images, captions or transcripts for audio/video assets specified but not written, use of device-specific language (click, tap, swipe — normalise to "select"), colour-only information (identify any place where meaning is conveyed only by colour), and reading level (flag any sentence over 25 words). List each issue with a specific fix suggestion.
24 — Bloom's taxonomy sense-check
Review these learning objectives and classify each one by Bloom's taxonomy level: [paste objectives]. For each, state: the Bloom's level (Remember, Understand, Apply, Analyse, Evaluate, Create), whether the verb is observable, and whether the objective is at an appropriate level for the intended audience and module type. Flag any module where all objectives sit at Remember or Understand level only — these need at least one Apply-level objective.
Cowork Skills for ID Practice
Five reference cards describing how to apply the Cowork delegation model to specific instructional design tasks. Each card names the skill, the ID application, and the prompt pattern to use.
Cowork Skill 01
Brief Before You Prompt
Before opening Copilot, write the outcome, context, and constraints for the task on paper or in a notes file. The brief should take 2–3 minutes to write. If you cannot describe the output precisely, you are not ready to prompt.
ID application: Use before every outline, script, or knowledge check generation. The brief becomes the quality check criteria once the output arrives.
Applies to: all prompt typesCowork Skill 02
Context-Load the Prompt
The more specific context you provide — audience, role, prior knowledge, device context, industry — the more specific the output. Context is not padding. Every context element closes a gap that Copilot would otherwise fill with a generic average.
ID application: In every script-generation prompt, include the audience description in full — even if you have used it before in the session.
Applies to: scripting, scenarios, objectivesCowork Skill 03
Constrain the Output
Name what Copilot should not produce. Common exclusions for ID: no bullet lists in prose sections, no general AI introduction, no hedging language, no product feature names that change frequently, no passive voice.
ID application: Add a constraints line to every prompt. Review the output against the constraints and reprompt if they are violated.
Applies to: all prompt typesCowork Skill 04
Evaluate Against the Brief
Read every Copilot output against the brief you wrote before prompting, not just as a standalone piece of text. The three questions: does it teach the objective, is the example specific, does the knowledge check require retrieval?
ID application: Complete the three-question review on every screen before it moves from the storyboard to the build stage.
Applies to: all AI-generated contentCowork Skill 05
Iterate With Precision
When the output is close but not right, do not start over. Write a corrective instruction that names exactly what is wrong and what the replacement should be. "Make this more specific" is not a corrective instruction. "Replace the generic example with one from a legal services context, showing a contract review task" is.
ID application: For every AI output that needs editing, decide first whether to correct in the platform (small text edits) or to iterate in Copilot (structural changes, voice changes, example replacements). Structural changes are faster to fix with a targeted reprompt than to edit by hand.
Applies to: all prompt types — especially scripting and scenario generationAgent Blueprints for ID Automation
Three Copilot Studio agent blueprints for automating high-volume, repeatable ID tasks. Each describes the agent's function, the inputs it needs, the outputs it produces, and the human review points.
Blueprint 01 — Level 1 Evaluation Agent
Agent name: ID-Eval-L1
Trigger: LMS completion event for [module name]
Actions:
1. Send Level 1 survey (4 questions) to learner email immediately on completion
2. Send 30-day follow-up to learner with their transfer intention pre-populated
3. Send 30-day manager check-in with learner's stated intentions
Inputs required:
- Learner email and name (from LMS)
- Manager email (from HR data connector)
- Transfer intention text (from survey response, stored on first send)
Outputs:
- Survey response data → SharePoint list
- Weekly digest of responses → ID team mailbox
- Flagged low-confidence responses → ID team for manual follow-up
Human review point: Review flagged low-confidence responses weekly. Check that transfer intention questions are producing specific, actionable answers — if they are vague, revise the question.
Requires: Microsoft Forms or equivalent, Power Automate connector to LMS, SharePoint list for response storage.
Blueprint 02 — Prompt Submission Review Agent
Agent name: ID-Assess-L2
Trigger: Learner submits prompt text via [submission method — form, Teams message, or SharePoint upload]
Actions:
1. Receive prompt submission
2. Evaluate against 4-criterion rubric (specificity, context, constraints, format instruction)
3. Score each criterion: Strong / Adequate / Needs work
4. Generate structured feedback referencing the rubric criteria
5. Return feedback to learner within [timeframe]
6. Flag borderline submissions (2 or more "Needs work" criteria) for human review
Inputs required:
- Learner's prompt text
- Rubric criteria (configured at agent setup)
- Context of the assessment task
Outputs:
- Scored rubric feedback → learner
- Borderline flagged submissions → ID assessor queue
- Aggregate rubric scores → SharePoint dashboard
Human review point: Review all borderline flagged submissions. Review aggregate scores weekly — if a criterion consistently scores "Needs work", the module content for that criterion needs strengthening.
Requires: Copilot Studio, SharePoint list for submissions and scoring, Power Automate for routing.
Blueprint 03 — Storyboard First-Draft Agent
Agent name: ID-Story-Draft
Trigger: Designer submits project brief via [Teams message or form]
Actions:
1. Receive project brief (objective, audience, duration, module type)
2. Generate section-by-section outline (learner-task-led, not topic-led)
3. For each section, generate: section objective, block types required, estimated screen count
4. Flag any section that appears to be a topic overview rather than a learner task
5. Produce output as a structured Word document or SharePoint page
Brief inputs required:
- Learning objective (verbatim, observable)
- Audience description (role, experience, device context)
- Total duration
- Structural requirements (must include worked example, review, etc.)
- Exclusions (what to leave out)
Outputs:
- Draft storyboard outline document
- Flagged sections list for designer review
Human review point: Designer reviews every flagged section and all section objectives before approving the outline. Do not proceed to scripting until the outline has been validated by the designer and SME.
The agent accelerates first-draft outline production. The designer's instructional judgment on whether the outline serves the objective remains non-delegable.
Observable Verb Library for AI-Augmented Roles
ESCO-grounded observable verbs for writing learning objectives in AI-augmented role contexts. Organised by cognitive level (Bloom's) and by the three phases of AI-augmented task performance.
By Bloom's level — with AI-context applications
| Level | Verbs | AI-context application |
|---|---|---|
| Remember | ListNameRecallIdentifyState | List the five elements of a structured Copilot prompt. Name the three phases of AI-augmented task performance. |
| Understand | DescribeExplainSummariseClassifyDistinguish | Describe the difference between a Copilot interaction and an agent action. Explain why context matters in a Copilot prompt. |
| Apply | ConstructDraftInstructBriefSpecifyConfigure | Construct a five-element Copilot prompt for a knowledge check task. Draft an agent instruction brief for a specific HR workflow. |
| Analyse | EvaluateCompareAssessCritiqueDistinguishBreak down | Evaluate an AI-generated learning objective against observable-verb criteria. Compare two Copilot prompts and identify which will produce more specific output. |
| Evaluate | JudgeValidateVerifyPrioritiseSelectJustify | Judge whether a Copilot output meets the original brief. Validate an agent action before confirming it in a production system. |
| Create | DesignProduceBuildDevelopSynthesiseCompose | Design a Cowork practice scenario for a specific role and AI tool. Produce a complete storyboard outline using Copilot assistance. |
By AI task phase — for three-layer objective design
| Phase | Verbs | Example objective fragment |
|---|---|---|
| Before (Task framing) |
DefineSpecifyFrameBriefScopeIdentify | "Specify the outcome, context, and constraints for a Copilot prompt before initiating the task." |
| During (Directing AI) |
InstructRedirectRefineConstrainAdjustIterate | "Redirect a Copilot output that has drifted from the original brief using a targeted corrective instruction." |
| After (Reviewing & owning) |
VerifyValidateReviewReviseApproveSubmit | "Verify a Copilot-generated document against the original brief before submitting it as your own work product." |
Evaluation Templates
Ready-to-use templates for the four Kirkpatrick levels, adapted for AI-focused learning programmes.
Level 1 — Reaction survey template (4 questions)
Q1 (Confidence — 1–5 scale):
"How confident are you now in using Copilot to [specific task from module]?"
1 = Not confident at all · 5 = Fully confident
Q2 (Transfer intention — open text):
"What will you do differently in the next five working days as a result of completing this module?"
Q3 (Friction — open text):
"Was there anything in this module that felt confusing, irrelevant, or wrong? If so, describe it briefly."
Q4 (Barriers — open text):
"Is there anything that would make it difficult to apply what you learned? If so, what would help?"
Level 2 — Performance assessment brief template
Task brief for learner:
"Using what you have learned in this module, write the Copilot prompt you would use to complete this task: [describe a specific, realistic task from the learner's role].
Submit: (1) the prompt you would use, and (2) a 2–3 sentence explanation of why you included each element of the prompt."
Rubric (4 criteria, each scored Strong / Adequate / Needs work):
1. Outcome specification: Does the prompt state what the output should be, precisely?
2. Context loading: Does the prompt include relevant audience, role, and situational context?
3. Constraint setting: Does the prompt specify what the AI should not do or include?
4. Format instruction: Does the prompt specify the output format, length, and voice?
Level 3 — Behaviour change baseline template
Learner self-assessment (1–5 scale for each):
- Confidence in writing a Copilot prompt for [task 1]
- Confidence in evaluating whether Copilot output meets a brief
- Confidence in giving corrective instructions to improve Copilot output
- Frequency of Copilot use in daily work (1 = never, 5 = multiple times daily)
Manager assessment (1–5 scale):
- Team member's ability to produce quality work products using Copilot
- Team member's critical evaluation of AI-generated content
- Observable change in how team member approaches AI-assisted tasks
Usage data baseline (from Microsoft 365 admin centre):
- Copilot feature activation date
- Weekly active Copilot usage (Teams, Outlook, Word)
- Prompt submission frequency (if available)
Craig Stanley Studio · ID Resources for Copilot & Agents · AI-ID Series · Direct link access only · © 2026