Your people are waiting for you to say yes.
Ethan Mollick's research shows that workers are already using AI — and hiding it. They are worried about organizational reaction, not about the technology itself. The permission gap — not the technology — is the largest barrier to AI adoption in your section.
Delivery Modes
The Week 5 deck is built to run in two modes from the same slides. Pick the mode before the calendar invite goes out — the prep, the room setup, and the speaker-notes track all change. The deck’s speaker notes are tagged [BRIEFING] and [JOINT] — follow the track that matches the mode you chose.
| Mode A · 30-minute briefing | Mode B · Joint session (60–90 min) | |
|---|---|---|
| Audience | Supervisors only — section leaders, OICs, SNCOs, civilians who approve work. | Supervisors and their builders in the same room (paired side-by-side). |
| Use it when… | You have one shot at a busy leadership team and need to land permission culture and the four evaluation questions in a SITREP-sized window. | Builders have already finished Weeks 1–4 and have at least one deployed tool to discuss; supervisors need to calibrate taste against real work from their own section. |
| Format | Narrated briefing. Two scenarios run as silent-think + one answer. Closes with verbal commitment. | Same slides, but each module ends with paired discussion. Adds Exercise C (Spot the Apprentice Risk) and a live tool review with builders. |
| Pre-work | Calendar invite, deck loaded, AI Review Drill memo ready in chat. | All of the above plus each builder pre-shares a URL for one deployed tool to demo (5 min each). |
| Outcome | Each supervisor commits aloud to one experiment with one Marine in the next 14 days. | Each supervisor · builder pair leaves with a written “what gets a Yes” rule the supervisor will apply this quarter, plus the same 14-day commitment. |
Default to the 30-minute briefing
If you have to guess, run Mode A. The 30-minute briefing is the program’s headline product and the highest-leverage use of a supervisor’s time. Reach for the joint session only when the builders are ready (Weeks 1–4 complete, deployed tool in hand) and the senior leader has explicitly carved out the longer window.
30-Minute Briefing Agenda Mode A
The default 30-minute supervisor briefing. Five short modules, two decision exercises, one verbal commitment at the close. Use this when you have one slot on a leadership calendar.
| Time | Module | Focus |
|---|---|---|
| 0:00 – 0:05 | Opening & Context | The permission gap, DoW AI Strategy, rapid adoption mandate |
| 0:05 – 0:10 | Permission Culture | What “yes” looks like, what kills adoption, what NOT to do |
| 0:10 – 0:20 | Evaluating AI-Assisted Output | Four questions, practice scenarios, evaluation exercises |
| 0:20 – 0:25 | The Apprentice Problem | Preserving junior development while gaining efficiency |
| 0:25 – 0:30 | Quick Reference & Wrap-Up | Supervisor quick reference card, closing action item |
Joint-Session Agenda (60–90 min) Mode B
The extended joint session runs the same slide deck, but every module ends with paired discussion between each supervisor and the builder sitting next to them. Plan on 60–90 minutes for the extended modules alone — or up to two hours if you include the optional live tool review and “Yes” criteria workshop documented in the Week 5 Facilitator Pack.
Room setup
Seat each supervisor next to the builder they supervise. Pairs discuss in place — no breakout rooms needed. The supervisor asks; the builder answers honestly, as if their own work were under review. The point of the joint session is calibration: by the end, builders should know exactly what “yes” sounds like from their supervisor, and supervisors should know what to ask their builders.
Extended timing
| Time | Module | What changes vs. the briefing |
|---|---|---|
| 0:00 – 0:10 | Opening & Context (Module 1) | After the data wall, ask the room: “Which of these five is most likely to land with your CO if you have to defend an AI decision next week?” Take 2–3 answers before advancing. |
| 0:10 – 0:20 | Permission Culture (Module 2) | Two-minute pair talk on the “what yes / what kills” slide. Each supervisor names a left-column phrase they will use this month; each builder names a right-column phrase they have heard at least once. Surface 2–3 commitments aloud. |
| 0:20 – 0:40 | Evaluating Output (Module 3, expanded) | Scenarios A and B run as 3-minute paired discussions instead of silent-think. Supervisor asks the four questions; builder responds about the scenario as if it were their own work. Surface 2–3 questions per scenario before each debrief slide. |
| 0:40 – 0:55 | The Apprentice Problem + Exercise C (Module 4, expanded) | Run the apprentice centerpiece, then anchor with Exercise C · Spot the Apprentice Risk (5–7 min full-room discussion). End with each supervisor naming one developmental task in their section that AI is currently doing instead of being a teaching tool. |
| 0:55 – 1:10 | Quick Reference & Commit (Module 5) | Have one supervisor read each of the three reference cards aloud. Builders confirm the cards reflect the work they actually do. Close with a paired commitment round — supervisor names who they’ll ask and by when; builder remembers who said their name. |
| +30–45 min | Optional · Live Tool Review | Each builder — five minutes. Show the deployed tool. Tell the room one thing AI did well, one thing it got wrong. Supervisor asks one hard question. Not an approval gate — a calibration of taste. |
| +30 min | Optional · “Yes” Criteria Workshop | Each supervisor · builder pair writes three rules that earn a Yes from this supervisor for AI use, and two that earn a No. Specific. No platitudes. Pairs read out; supervisor publishes the list to the section by end of week. |
Exercise C · Spot the Apprentice Risk
Exercise C is the joint-session-only addition. It only works with builders in the room because the lesson lives in supervisor and builder reacting to the same scenario at the same time.
Scenario · joint discussion (5–7 min)
A new SgtSec NCO is assigned to draft initial counseling statements for three junior Marines. They use ChatGPT to produce all three drafts in under five minutes. The drafts are technically correct. They sign and submit them. Their supervisor approves them without comment.
Discussion prompts:
- What was lost — for the NCO, for the Marines being counseled, for the section?
- What does a permission-culture supervisor do instead?
- Name one developmental task in your section that AI is currently doing instead of being a teaching tool.
Done looks like: the room can articulate the difference between AI-as-format-helper (acceptable for a counseling-statement template) and AI-as-judgment-replacement (not acceptable for the assessment paragraphs themselves).
Facilitator notes for joint mode
- Confirm the mode 24 h out. The senior leader chooses Mode A or Mode B. If they pick Mode B, coordinate that each builder pre-shares a URL for one deployed tool (used in the optional live review — and as a fallback artefact during Module 3 if the room runs cold).
- Press N to enable speaker notes in the deck. Notes are tagged [BRIEFING] and [JOINT]. In Mode B, follow the [JOINT] track at every tagged slide — the briefing track stays on the same slide content but skips the discussion prompts.
- You moderate — you do not narrate. In paired discussions and live tool reviews, supervisors and builders own the airtime. The facilitator’s job is to start the timer, surface 2–3 answers, and advance.
- Pacing guard rail. Module 3 (Evaluating Output) tends to overrun in joint mode because the scenarios produce real disagreements. Hold the line at 20 minutes for that block; the apprentice problem is non-negotiable.
- If builders arrive without tools to demo (optional review), pivot to the “Yes” criteria workshop — it works with or without a live tool.
- Closing accountability. Joint mode adds a paired commitment round at the end: supervisors say out loud who they’ll ask and by when; builders remember who said their name. This is the single highest-retention moment in the session — do not skip it.
For the printable companion
The Week 5 Facilitator Pack is the print-and-bring-to-the-room version of this agenda — same timing strip ([B] briefing rows and [J] joint extension rows), plus exercise prompts to read verbatim, recovery scripts for when things go sideways, and post-session homework. Keep it next to the deck during delivery.
Module 1: Why This Matters Now 5 min
The data is unambiguous. AI adoption with proper leadership support produces measurable results. AI adoption without it fails consistently.
- Microsoft research: Most workers abandon AI within weeks without organizational support.
- UK government study: 25 minutes per day in savings with proper support structures in place.
- Department of War (DoW) January 2026 AI Strategy: Year of Military AI Dominance — this is the directive, not a suggestion.
- Army: Created a dedicated AI/ML officer career field (49B).
- Marine Corps: Running generative AI workshops at Quantico.
AI is a SITREP item. Not optional anymore.
Module 2: Permission Culture 5 min
What “Yes” Looks Like
- “Try it. Show me what you build.”
- “I don't need to understand how it works. I need to understand what it does.”
- “If it saves time and meets quality standards, we should do it.”
- Protected time for learning.
- Public recognition for useful builds.
What Kills Adoption
- “I need to approve every AI interaction.”
- “Don't use AI for anything official.”
- “We'll wait until there's a formal policy.”
- Punishing experimentation failures.
- Treating AI use as suspicious or lazy.
What Supervisors Should NOT Do
- Don't ban AI use outright. Prohibition drives it underground where you can't guide it. You lose visibility and control.
- Don't require approval for every AI interaction. Creates a bottleneck that kills velocity. Review output, not every query.
- Don't assume AI output is automatically wrong. Skepticism kills adoption. Verify like any work product — evidence-based, not assumption-based.
- Don't skip verification just because the Marine is experienced. Seniority doesn't eliminate hallucination risk. Everyone verifies.
Guard Rails (Not Roadblocks)
- All tools go through the EDD SOP before deployment.
- AI-generated content for official use must be reviewed.
- Sensitive or classified information never enters unauthorized AI systems.
- Failed experiments are shared as learning, not punished.
The default answer should be “yes, with appropriate review.”
Module 3: Evaluating AI-Assisted Output 10 min
When a Marine brings you AI-assisted work, you need exactly four questions:
- Does it work? Can you demonstrate it?
- Is the output accurate? Verifiable facts, references, numbers?
- Does it follow the SOP? Proper review process completed?
- Does it save time? Before-and-after comparison?
You do not need to evaluate how AI produced it. Focus on output, not mechanism. Apply the same standards as any work product — if a Marine handed you this without mentioning AI, would you sign it?
Practice Scenarios
Supervisors must practice evaluation before they need it operationally. Work through these two scenarios. What questions would you ask?
Scenario A: AI-Generated Leave Policy Summary
A Marine presents you with a one-page summary of leave policy for Emergency Leave Authorization. The summary is clear, well-formatted, and cites MCO P1050.3K. The Marine tells you it was generated using ChatGPT and reviewed for accuracy. They want to distribute it to the section.
Suggested Evaluation Questions
- Can you verify the MCO citation? Pull up the actual order. Does the summary match?
- What did you use as the AI input? Did you paste the full MCO text, or ask general questions? Context quality matters.
- Who reviewed this for accuracy? AI-generated policy content requires subject-matter verification, not just proofreading.
- What happens if the policy changes? Is there a process to update this, or will outdated summaries circulate indefinitely?
Scenario B: Automated Reporting Tool
A Marine built a Python script that pulls data from a shared spreadsheet and auto-generates the weekly operations summary. It runs in 30 seconds. The manual process took 90 minutes. The Marine wants approval to use it for official reporting.
Suggested Evaluation Questions
- Can you demonstrate it with live data? Show me input, process, output. Walk through one full cycle.
- What happens when the input data is malformed? Does it fail gracefully, or generate incorrect reports?
- Who else has reviewed this code? Per EDD SOP, automation tools require peer review and security review before operational use.
- How do we verify accuracy each week? Sampling checks, output validation — what's the ongoing QA process?
Module 4: The Apprentice Problem 5 min
The Risk: Junior Marines Who Never Learn Fundamentals
If AI handles all the routine tasks that junior Marines traditionally learned on, they never develop professional judgment. They become dependent on tools they don't understand. Entry-level employment in AI-exposed fields has dropped 13% since 2022. The military cannot afford a generation of Marines who can operate AI but cannot operate without it.
Supervisors Must Ensure AI Augments, Not Replaces, Development
AI is a force multiplier for trained professionals. It is a crutch for untrained juniors. Your role is to ensure the former, prevent the latter.
Three Supervision Checks
- Can the Marine explain the output without referencing AI? If they can't walk you through the logic, they didn't learn — they copied.
- Require periodic tasks to be completed without AI. Preserve baseline competency. If the tool disappears, can they still operate?
- Use AI output as a teaching tool. “Is this correct? How would you verify it? What would happen if this number was wrong?”
Developmental Assignments Are Non-Negotiable
Certain tasks exist to build judgment, not just complete work. Writing after-action reviews, analyzing lessons learned, drafting initial counseling statements — these are formative assignments. AI can assist, but it cannot replace the learning value of the work itself.
Goal: AI-augmented Marines, not AI-dependent ones.
Supervisor Quick Reference 5 min
This is a reference card supervisors can screenshot, print, or save. Use it as a decision aid when evaluating AI-assisted work.
Supervisor Quick Reference: AI-Assisted Work
3 Questions to Ask Before Approving AI-Assisted Work
- Can you demonstrate it working? Live demo with real inputs.
- How did you verify accuracy? Specific checks, not general trust.
- Who reviewed this? Peer review, SME review, security review per SOP.
3 Signs of Healthy AI Use
- The Marine can explain the output without referencing AI. They understand the work, not just the tool.
- Failed experiments are shared openly. Learning culture, not fear culture.
- AI is saving time on routine tasks, not replacing judgment tasks. Force multiplier, not replacement.
3 Warning Signs of Unhealthy AI Use
- The Marine cannot answer basic questions about their own output. Dependency, not augmentation.
- AI use is hidden or apologized for. Sign of a fear culture you created.
- Junior Marines are skipping foundational tasks entirely. No baseline competency being built.
Default Posture: Yes, with appropriate review.
Tools Your Marines Will Use
GenAI.mil is the Marine Corps enterprise AI platform per MARADMIN 018/26. It is IL5-authorized for CUI and should be the default tool for all AI-assisted work on DoD networks. CamoGPT (Army-managed) provides supplementary capabilities including API access and IL6/SIPR support. Commercial tools (ChatGPT, Gemini via their websites) are approved for unclassified work only. ChatGPT, Gemini, and Grok are also available on GenAI.mil. PII must be anonymized on all AI platforms unless a PIA authorizes it. See the Approved Tools page for the full list.
Closing
The EDD program provides structure. Your people have motivation. The Department of War has directed priority. The only thing between your section and measurable productivity gains is your permission. Give it.
Instructor Note
This is the highest-leverage 30 minutes. Respect the time constraint. Lead with operational relevance. End with a specific ask: “Within the next week, ask one person in your section what they'd build if they had permission.”
Knowledge Check
According to research cited in this course, what is the largest barrier to AI adoption in most organizations?
Ethan Mollick's research shows workers are already using AI but hiding it. They worry about organizational reaction, not the technology itself. The permission gap is the largest barrier.
The UK government study on AI adoption found that workers with proper support structures saved approximately how much time per day?
The UK government study found 25 minutes per day in savings when proper support structures were in place. AI adoption with leadership support produces measurable results.
What does the Department of War's January 2026 AI Strategy represent for military units?
The DoW January 2026 AI Strategy declared the Year of Military AI Dominance. This is a directive, not a suggestion. AI is a SITREP item — no longer optional.
Knowledge Check
Which of the following is an example of what “yes” looks like in a permission culture?
“Try it. Show me what you build” is a permission culture statement. The other options are adoption-killing behaviors that create fear cultures and drive AI use underground.
Why should supervisors NOT ban AI use outright?
Prohibition drives AI use underground where you can't guide it. You lose visibility and control. The goal is guard rails (not roadblocks) with appropriate review processes.
What is the recommended default posture for supervisors when evaluating AI use?
The default answer should be “yes, with appropriate review.” This creates a permission culture while maintaining necessary guard rails through the EDD SOP process.
Knowledge Check
When a Marine brings you AI-assisted work, which of the following is one of the four critical evaluation questions?
The four critical questions are: Does it work? Is the output accurate? Does it follow the SOP? Does it save time? Focus on output quality, not on how the AI produced it.
When evaluating AI-assisted output, supervisors should focus on:
You do not need to evaluate how AI produced the output. Focus on output, not mechanism. Apply the same standards as any work product — if a Marine handed you this without mentioning AI, would you sign it?
In the practice scenario where a Marine built an automated reporting tool, what is the FIRST evaluation question a supervisor should ask?
“Can you demonstrate it with live data?” is the first evaluation step. Show input, process, output — walk through one full cycle. Verification starts with demonstration.
Knowledge Check
What is the “apprentice problem” in the context of AI adoption?
If AI handles all routine tasks that junior Marines traditionally learned on, they never develop professional judgment. They become dependent on tools they don't understand. The goal is AI-augmented Marines, not AI-dependent ones.
Which of the following is one of the three supervision checks for preserving junior development?
One of the three supervision checks is requiring periodic tasks to be completed without AI. This preserves baseline competency — if the tool disappears, can they still operate?
Why are developmental assignments like after-action reviews and initial counseling statements considered “non-negotiable” for junior Marines?
Certain tasks exist to build judgment, not just complete work. AI can assist, but it cannot replace the learning value of formative assignments. The goal is AI-augmented Marines, not AI-dependent ones.
Knowledge Check
Which of the following is a sign of HEALTHY AI use in a section?
Failed experiments being shared openly is a sign of a healthy learning culture. When Marines hide or apologize for AI use, that signals a fear culture created by the supervisor.
Which of the following is a WARNING sign of unhealthy AI use?
Junior Marines skipping foundational tasks entirely means no baseline competency is being built. This is dependency, not augmentation. Supervisors must ensure AI augments development, not replaces it.
According to the supervisor quick reference, what are the three questions to ask before approving AI-assisted work?
The quick reference card lists three questions: (1) Can you demonstrate it working? (2) How did you verify accuracy? (3) Who reviewed this? These focus on output quality and process compliance, not on the AI mechanism.
Course Completion Checklist
Check off each item as you complete it. Your progress is saved automatically in your browser.
Course Complete!
You have completed all items in the Supervisor Orientation course. View your progress dashboard or generate a completion certificate.