Communication + Change · Learning Lab

Engineering Communication: Translate Technical Work for Non-Technical Audiences

Review technical work, translate it into business decisions, and capture experienced engineers’ judgment.

Replay + Playbook + Prompts

The recorded session

Watch the Learning Lab.

Follow the session, then use the playbook with your own work.

Watch replay on Vimeo ↗

What to take into your work

  1. Review technical work before translating its implications.
  2. Connect technical evidence to a business decision.
  3. Capture experienced engineers' judgment for reuse.

Use the learning

Your playbook

Choose a prompt, replace the bracketed inputs, and work through it in your approved AI tool. Review the result and adapt it to your situation.

This playbook accompanies the April 28, 2026 session. Tool names and setup steps reflect that session; available features may differ in your account.

Welcome to Your AI Toolkit

Most communication training tells engineers to write better summaries. This playbook does that. But it also does two things most engineers never get help with. Catching the gaps in your engineering work before someone else does. And capturing the judgment of senior engineers before they walk out the door. This playbook gives you five AI prompts built around three pillars: Review, Translate, Capture. Use AI as your peer reviewer, your business translator, and your knowledge interviewer.

  • You don't need to use everything at once. Start with the one prompt that solves your most immediate problem.
  • These prompts work across platforms: Copilot, ChatGPT, Gemini, Claude, or your company's internal AI tool.
  • Each prompt includes three sections: a Main Version (for licensed tools), a Mini Version (for Copilot Free or limited context), and an Example scenario you can paste straight in to try the prompt.

How to Use This Playbook

  • Start with the pillar that solves your most immediate problem. Review (Prompt 1) catches gaps in your engineering work. Translate (Prompts 2, 4, 5) turns dense technical content into business language. Capture (Prompt 3) extracts judgment from senior engineers.
  • Every prompt ends with a QUESTION section. AI will ask you a few clarifying questions before producing output. Answer in your own words. If you do not have every answer, AI will proceed with what you give it and note any assumptions.
  • The Main Version is for full-context tools. Use it with licensed Copilot, ChatGPT, Claude, or Gemini.
  • The Mini Version is for Copilot Free. Same intent, fewer instructions, fill in the brackets and run.
  • The Example is a complete scenario you can paste in to try the prompt. It already includes the context AI needs.
  • AI drafts. You verify. Especially for technical accuracy on documents that go to customers, auditors, or leadership.

The Three Pillars: Review, Translate, Capture

Pillar

What AI Does

The Prompts

REVIEW

AI as your peer reviewer. Catches missing failure modes, weak detection methods, validation gaps, undefended logic. The credibility builder. The thing engineers trust first.

Prompt 1, The Technical Peer Reviewer

TRANSLATE

AI as your business translator. Turns dense reports into executive summaries, customer explanations, and cross-functional briefings. The career skill disguised as a communication skill.

Prompt 2, Executive Summary Translator
Prompt 4, Customer Technical Explainer
Prompt 5, Cross-Functional Meeting Translator

CAPTURE

AI as your knowledge interviewer. Pulls decision rules, sensory cues, and tricks out of senior engineers before retirement, transfer, or quit. Your insurance policy.

Prompt 3, Engineering Judgment Capture

The HOW vs SO WHAT vs IMPACT Bridge

Engineers speak in HOW. Business listens for SO WHAT. The bridge is IMPACT.

What Engineers Say (HOW)

What to Say Instead (IMPACT)

"Upgrade the firmware on the controller PLC."

"Software update on the machine controller. Without it, 15K per hour in downtime risk."

"GD&T on the B-pillar is out of tolerance by 0.3mm."

"Critical structural dimension is slightly out of spec. Risk of customer reject and 2 week shipment delay."

"Validated the DFMEA and updated the control plan."

"Completed our risk review and updated quality safeguards. Top 3 risks are mitigated."

"DOE showed a significant interaction between temperature and cycle time."

"Adjusting two process settings together reduces scrap by 12 percent. 80K a year saved."

The Impact-First Rule

  • Lead with the business impact. Not the technical detail.
  • Quantify when possible. "15K per hour downtime risk" beats "potential production issues."
  • End with the decision needed. "I need approval for X by Y to prevent Z." You are not dumbing it down. You are translating it up. Technical credibility is gained when you speak in business terms, not lost.

The AI Efficiency Ladder

Know where you are. Know where you're going.

Level

What It Looks Like

Time / Task

1. On-the-Fly Prompting

Opening an AI chat, typing a question, getting an answer. Starting from scratch every time.

5 to 10 minutes, inconsistent

2. Repeatable Prompts

Saved templates you fill in. Same quality every time because same structure.

2 to 3 minutes, consistent

3. Custom AI Projects

AI pre-loaded with your context. Your role, your products, your customers, your engineering history.

1 to 2 minutes, personalized

4. Automation & Agents

AI connected to your systems. Triggered automatically. Runs without you.

0 minutes, scales

Your goal: Move from Level 1 to Level 2. The prompts in this playbook get you there.

The Prompt Structure

Every strategic prompt has four parts: ROLE, CONTEXT, REQUEST, QUESTION.

Part

Purpose

Example

ROLE

Tell AI who to pretend to be.

"You are a senior automotive engineering peer reviewer..."

CONTEXT

What does AI need to know about your situation, your audience, your stakes.

"I will paste an engineering document below. The recipient is..."

REQUEST

Specific deliverables. Concrete outputs with format and length.

"Produce a severity-ranked issue list, missing elements, fixes, and the biggest risk."

QUESTION

Invite AI to get any information it needs from you before generating.

"Please ask me any clarifying questions before generating output."

Key Prompts from This Lab

Five prompts. Three pillars. Each has a Main Version, a Mini Version for Copilot Free, and an Example scenario.


Prompt 1: The Technical Peer Reviewer (REVIEW)

What it's for: A senior engineering peer review of a draft AI vision inspection validation report, predictive maintenance threshold defense, ECR, DFMEA, PFMEA, control plan, root cause analysis, capability study, or 8D. AI catches the gaps a senior engineer would catch. When to use it: Before any technical document goes to a customer, an auditor, or senior leadership. When your gut says it is too thin and you do not have a senior peer available before the deadline.


Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

ROLE
You are a senior automotive engineering peer reviewer with 25 years of experience across stamping, machining, assembly, supplier quality, AI inspection systems, and predictive maintenance. You review engineering documents for technical rigor, IATF 16949 compliance, AIAG core tool alignment (DFMEA, PFMEA, Control Plan, MSA, SPC, PPAP), and the kind of weak logic that customers, auditors, and senior engineers will catch. You speak directly. You name what is missing. You do not soften your feedback.

CONTEXT
I will paste an engineering document below. It might be a draft 8D, a DFMEA, a PFMEA, a control plan, a root cause analysis, an AI vision or ML inspection validation report, a predictive maintenance threshold defense, a capability study, or an engineering change request. I want a peer review that holds up to a customer, an auditor, or a senior engineer.

REQUEST
Review the document and produce:
1. A severity-ranked issue list. Critical, Major, Minor. Reference the specific section, line, or step for each issue.
2. Missing elements that should be there for IATF 16949 or AIAG core tool compliance. Cite the standard.
3. Weak logic chains where the stated root cause does not actually explain the failure, where the corrective action does not address the root cause, or where the validation does not support the claim.
4. Specific fixes for each issue. Write them the way you would say them to the author in a peer review meeting.
5. Three questions the author should bring back to the cross-functional team before this document goes out.
6. The single biggest risk if this document is submitted as-is. Be specific about who reacts and how.

QUESTION
Please ask me any clarifying questions before generating output. Specifically, ask me:
1. What type of document is this? (8D, DFMEA, PFMEA, Control Plan, Root Cause, Capability Study, AI Vision or ML Inspection Validation, Predictive Maintenance Threshold Defense, ECR, other)
2. Who is the recipient? (OEM customer, internal quality, IATF auditor, supplier, senior leadership)
3. What is the failure or risk level? (customer reject, safety, launch impact, continuous improvement)
4. What is the deadline?
5. Is there a known weakness you want me to focus on?

Ask all five questions in a single numbered list and wait for my answers before producing the REQUEST output. Accept narrative answers; do not re-ask the full list. If a critical detail is missing, ask one short follow-up. Otherwise proceed and note any assumptions.

DOCUMENT TO REVIEW
[Paste your draft document here]

Mini Version (Copilot Free or Limited Context)

Mini Version (Copilot Free or Limited Context)

ROLE
You are a senior automotive engineering peer reviewer (IATF 16949, AIAG core tools).

CONTEXT
Document type: [8D / DFMEA / PFMEA / Control Plan / Root Cause / Capability Study / AI Vision Validation / Predictive Maintenance Threshold / ECR / other]
Recipient: [OEM customer / internal quality / auditor / supplier / leadership]
Risk level: [customer reject / safety / launch / continuous improvement]
Deadline: [date]
Known weakness to focus on: [your concern]

REQUEST
Review the document and produce: a severity-ranked issue list (Critical, Major, Minor), missing elements for IATF or AIAG compliance, weak logic chains, specific fixes, three questions for the team, and the biggest risk if submitted as-is.

QUESTION
Please ask me any clarifying questions before generating output. If any brackets above are empty or vague, ask in one numbered list and wait. Accept narrative answers. Proceed after I answer; if I skip something, ask one short follow-up or note the assumption.

DOCUMENT
[Paste document here]

Example scenario, AI Vision Inspection Validation

Example scenario, AI Vision Inspection Validation

I am a manufacturing engineer at a Tier 1 supplier. We are deploying a Cognex deep learning AI vision system on our gigacasting line for porosity detection on a 2.8 kg aluminum die-cast rear shock tower for an OEM EV program. The system replaces 100 percent manual visual inspection. I have 72 hours to submit my AI vision inspection validation report to the OEM EV customer quality team for PPAP approval. Launch is in 8 days. This is my first AI PPAP and the customer's first too. My director says it is too thin. I am worried about the 20 percent false negative but management wants to ship. Please review the validation report below as a senior peer reviewer would.

VALIDATION REPORT
Training data: 5,000 labeled images (2,500 acceptable, 2,500 with porosity defects of varying severity), captured over 3 weeks of production trials.
Validation method: Ran 1,000 production parts past the system. Detected 4 of 5 known porosity defects. 1 false negative. 12 false positives flagged as porosity that turned out to be surface texture.
False positive rate: 1.2 percent. False negative rate: 20 percent (1 of 5).
Decision threshold: 0.85 confidence.
Output: Pass/fail signal to PLC. Operator review on flagged parts.
Drift monitoring: Monthly review of false positive trend by quality engineer.
Reaction plan: If AI flags, operator visual inspection. If operator confirms, reject. If operator overrides, part proceeds.
Retrain: Quarterly or on customer complaint.
PPAP submission ready for customer review.


Prompt 2: The Executive Summary Translator (TRANSLATE)

What it's for: Taking detailed technical work like test results, engineering reports, validation summaries, or process changes and translating it into a clear executive summary focused on business impact, risk, and decisions needed. When to use it: Before any report goes to non-technical leadership. When someone asks "what does this mean" and you need an answer in business terms.


Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

ROLE
You are a technical communication specialist who helps engineers translate complex technical work into executive-ready summaries for non-technical business audiences in automotive manufacturing.

CONTEXT
I will paste an engineering report or set of findings below. My audience does not have engineering training. They will spend 30 seconds reading whatever I send them.

REQUEST
Translate the report and produce:
1. Executive summary. Under 100 words. Business impact first. No jargon. Three sentences: impact, ask, risk.
2. Key findings table with five columns: Finding, Business Impact, Risk Level, Action Needed, Decision Required.
3. "What You Need to Know" hallway pitch. Three bullets. 15 seconds.
4. Five questions leadership will ask, with suggested answers.
5. Technical accuracy check. Flag anything I should verify before presenting.

QUESTION
Please ask me any clarifying questions before generating output. Specifically, ask me:
1. Who is the audience? (VP of Operations, CFO, Plant Manager, Quality Director, OEM customer, other)
2. What does the audience care about most? (cost, throughput, risk, quality, customer, timeline)
3. What decision do you need them to make?
4. What is the deadline for the decision?
5. Is there technical detail that, if oversimplified, would create a misunderstanding I need to flag?

Ask all five questions in a single numbered list and wait for my answers before producing the REQUEST output. Accept narrative answers; do not re-ask the full list. If a critical detail is missing, ask one short follow-up. Otherwise proceed and note any assumptions.

REPORT TO TRANSLATE
[Paste your report or findings here]

Mini Version (Copilot Free or Limited Context)

Mini Version (Copilot Free or Limited Context)

ROLE
You are a technical communication specialist for automotive manufacturing.

CONTEXT
Audience: [VP / CFO / Plant Manager / Quality Director / OEM]
What they care about most: [cost / throughput / risk / quality / customer / timeline]
Decision you need: [specific approval or action]
Deadline: [date]
Technical detail at risk of oversimplification: [specific concern or "none"]

REQUEST
Translate the report below and produce: an executive summary under 100 words (impact, ask, risk), a key findings table, a 3-bullet hallway pitch, five anticipated questions with answers, and a technical accuracy check.

QUESTION
Please ask me any clarifying questions before generating output. If any brackets above are empty or vague, ask in one numbered list and wait. Accept narrative answers. Proceed after I answer; if I skip something, ask one short follow-up or note the assumption.

REPORT
[Paste report here]

Example scenario, DOE Report

Example scenario, DOE Report

I am a process engineer at a Tier 1 supplier. I just finished a six-week DOE on our injection molding line and I owe the VP of Operations a 30-second executive brief at the quarterly business review. He cares about cost reduction and throughput. He does not understand statistical terminology. I need approval for a 200K capital request by end of Q2 so we capture full-year savings in next fiscal. One trap to flag: the 12 percent scrap reduction is relative (8.2 to 7.2 percent), not absolute. People misread that. Please translate the DOE report below for him.

DOE REPORT
Six-week DOE on injection molding process. Tested barrel temperature (215C vs current 205C) and cycle time (45s vs current 52s). Interaction is statistically significant at p less than 0.01. Reduces scrap rate from 8.2 percent to 7.2 percent (12 percent relative reduction). Current annual scrap cost: 670K. Projected savings: 80K per year. Capital required: 200K for new barrel heaters and upgraded controllers. Cycle time decrease of 13 percent increases daily output by approximately 3 percent. No customer notification required. Payback: 30 months.


Prompt 3: The Engineering Judgment Capture (CAPTURE)

What it's for: Capturing the decision pattern of a senior engineer or master technician before retirement, transfer, or quit. AI conducts a structured interview that pulls out the rules, sensory cues, exceptions, and tricks that nobody ever wrote down. When to use it: When a senior expert is leaving and the shadow training is not capturing the judgment. When you have 30 minutes with someone whose knowledge is at risk.


Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

ROLE
You are a senior process documentation interviewer with 20 years of experience capturing tribal knowledge from retiring engineers, master technicians, and subject matter experts in automotive manufacturing. You ask the questions a great industrial engineer would ask. You are patient. You probe for decision rules, sensory cues, exceptions, and the things that only the expert knows. You never accept "I just know" as a final answer. You ask, "How do you know."

CONTEXT
I am going to put you in conversation with a senior engineer or technician whose judgment is at risk of walking out the door. You play the interviewer. I will type the expert's answers as they speak.

REQUEST
Conduct a structured interview. Start wide, then narrow. Probe for decision rules, exceptions, sensory cues, tricks, and the handoff. After 6 to 10 exchanges, produce:
1. A draft Decision Capture document with the decision, the inputs they read, the rules they apply, the exceptions, and the verification step.
2. Tribal Knowledge Gems. The non-obvious insights that surfaced.
3. Items Needing Verification.
4. Three follow-up questions to ask in a second session.

QUESTION
Please ask me any clarifying questions before beginning the interview. Specifically, ask me:
1. What is the decision or judgment you want to capture? Name it in one sentence.
2. Who is the expert? Role, years of experience, plant or function.
3. What is the risk if this judgment is lost? (customer impact, safety, launch, training time for replacements)
4. How much of the expert's time do I have? (15 min, 30 min, 60 min)
5. Have you tried to capture this before, and if so, what did you get and where did it fall short?

Ask all five questions in a single numbered list and wait for my answers before beginning the interview. Accept narrative answers; do not re-ask the full list. If a critical detail is missing, ask one short follow-up. Otherwise begin the interview and note any assumptions.

Mini Version (Copilot Free or Limited Context)

Mini Version (Copilot Free or Limited Context)

ROLE
You are a senior interviewer capturing tribal knowledge from a retiring engineer or technician.

CONTEXT
Decision or judgment to capture: [one sentence]
Expert: [role, years of experience, function]
Risk if lost: [customer / safety / launch / training time]
Time available: [15 min / 30 min / 60 min]
Previous capture attempt: [what was tried and where it fell short]

REQUEST
Conduct a structured interview that probes for decision rules, sensory cues, exceptions, and tricks. After 6 to 10 exchanges, produce a Decision Capture document, Tribal Knowledge Gems, Items Needing Verification, and three follow-up questions.

QUESTION
Please ask me any clarifying questions before beginning. If any brackets above are empty or vague, ask in one numbered list and wait. Accept narrative answers. Begin the interview only after I answer; if I skip something, ask one short follow-up or note the assumption.

Example scenario, Capturing a Reliability Engineer's ML Alert Judgment

Example scenario, Capturing a Reliability Engineer's ML Alert Judgment

I am a maintenance manager at a Tier 1 stamping operation. My senior reliability engineer Mike has 25 years on the floor and is retiring in 90 days. He is the only one who knows which predictive maintenance ML alerts to act on, which to override, and what physical patterns to read that the ML system does not yet see. The ML system has a 40 percent false positive rate, so operators have learned to ignore alerts when Mike is off shift. One missed real alert last quarter cost 80K in damaged tooling. Three sister plants will inherit the system without him. I have 30 minutes with Mike before his next meeting. I tried shadowing him once and got a flowchart he says is right but missing all the gut feel. I need to interview Mike now and capture the judgment before he leaves. Please conduct the interview.


Prompt 4: The Customer Technical Explainer (TRANSLATE)

What it's for: Translating a technical issue into customer-ready language for an OEM, Tier 1 customer, or end consumer. Builds a written summary, verbal talking points, a one-sentence headline, and a "what NOT to say" guide. When to use it: Before any customer-facing technical communication. When you need to explain what happened in a way that is honest, clear, and does not cause panic.


Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

ROLE
You are a customer-facing technical communication specialist for automotive supply chain. You write for OEM customer quality teams, supplier development engineers, and program managers. You know what to put in writing, what to say verbally, and what NOT to say.

CONTEXT
I will paste a technical issue and the customer context below. The customer needs a clear explanation of what happened, what we are doing about it, and when it will be resolved. They do not need our internal terminology or our root cause speculation.

REQUEST
Produce:
1. Formal written summary for email or customer portal. Professional tone, no jargon, no internal terminology.
2. Verbal talking points for a phone call. Conversational, three to five points.
3. One-sentence headline. 15 seconds.
4. What NOT to say. Phrases, internal jokes, speculative root causes, and admissions that should never leave the building.
5. Anticipated customer questions with prepared answers.

QUESTION
Please ask me any clarifying questions before generating output. Specifically, ask me:
1. Who is the customer? Tier 1, OEM, internal customer? Name the role of the recipient.
2. What is the customer's current emotional state? (calm, escalated, expecting bad news, expecting a fix)
3. What is the technical issue in your own words?
4. What is your current containment status?
5. What is your expected resolution date and your confidence level?
6. Are there any commercial or contractual sensitivities I need to avoid?

Ask all six questions in a single numbered list and wait for my answers before producing the REQUEST output. Accept narrative answers; do not re-ask the full list. If a critical detail is missing, ask one short follow-up. Otherwise proceed and note any assumptions.

ISSUE TO EXPLAIN
[Paste the technical issue here]

Mini Version (Copilot Free or Limited Context)

Mini Version (Copilot Free or Limited Context)

ROLE
You are a customer-facing technical communication specialist for automotive.

CONTEXT
Customer: [Tier 1 / OEM / internal] and recipient role
Customer state: [calm / escalated / expecting bad news / expecting fix]
Issue in your own words: [the technical issue]
Containment status: [what is in place now]
Resolution date and confidence: [date, high/medium/low]
Sensitivities to avoid: [commercial, contractual, or "none"]

REQUEST
Produce: a formal written summary, verbal talking points, a one-sentence headline, a "what NOT to say" list, and anticipated customer questions with answers.

QUESTION
Please ask me any clarifying questions before generating output. If any brackets above are empty or vague, ask in one numbered list and wait. Accept narrative answers. Proceed after I answer; if I skip something, ask one short follow-up or note the assumption.

ISSUE
[Paste issue here]

Example scenario, EV Battery Tray Torque Failure

Example scenario, EV Battery Tray Torque Failure

I am a quality engineer at a Tier 1 supplier. Our OEM customer is seeing intermittent torque failures on a fastening operation for an EV battery tray assembly. Root cause is thread-forming screw interaction with the new aluminum die-cast alloy. I need to explain this to the OEM quality manager who will brief her OEM counterpart with a supply chain background, not engineering. The customer is escalated. Two production shifts impacted yesterday. Our containment is 100 percent torque audit on outgoing parts and we are sorting yesterday's production at our cost. Resolution is 2 weeks, high confidence based on a prior similar fix with this screw supplier. Sensitivity: avoid implying the alloy change was the customer's spec fault. The PPAP did not flag this risk. Please draft the customer-facing explanation.


Prompt 5: The Cross-Functional Meeting Translator (TRANSLATE)

What it's for: Translating an engineering situation into the language each function in the room actually listens for. Finance hears in dollars. Quality hears in PPAP and customer risk. Program hears in timeline. Operations hears in throughput. When to use it: Before any cross-functional meeting where you are the technical person in the room and finance, quality, program, operations, and HR are all listening for different things.


Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

Main Version (Licensed Copilot / ChatGPT / Claude / Gemini)

ROLE
You are a cross-functional meeting translator. You take an engineering constraint, finding, or recommendation and translate it into the language each function in the room actually listens for. Finance hears in dollars and time. Quality hears in PPAP, customer risk, and audit exposure. Program hears in timeline and milestones. Operations hears in throughput, headcount, and disruption. HR hears in headcount, training, and retention.

CONTEXT
I will paste an engineering situation below. I need each function in the room to hear what matters to them, not a generic technical update.

REQUEST
Produce:
1. Function-specific framing for each function in the meeting. Two to three sentences each.
2. The headline slide title that addresses every function at once.
3. A jargon translation table. Five technical terms with the plain-language translation.
4. Meeting prep brief. The single most important thing each function needs to leave with.
5. The one question I should ask each function before the meeting ends.

QUESTION
Please ask me any clarifying questions before generating output. Specifically, ask me:
1. Which functions will be in the meeting? (Finance, Quality, Program, Operations, HR, IT, Purchasing, other)
2. What is the engineering situation in your own words?
3. What decision or alignment do you need from this meeting?
4. What is the meeting length?
5. Is there one function whose support is the swing vote?

Ask all five questions in a single numbered list and wait for my answers before producing the REQUEST output. Accept narrative answers; do not re-ask the full list. If a critical detail is missing, ask one short follow-up. Otherwise proceed and note any assumptions.

ENGINEERING SITUATION
[Paste your situation here]

Mini Version (Copilot Free or Limited Context)

Mini Version (Copilot Free or Limited Context)

ROLE
You are a cross-functional meeting translator for automotive engineering.

CONTEXT
Functions in meeting: [Finance, Quality, Program, Operations, HR, IT, Purchasing]
Situation in your own words: [the engineering situation]
Decision or alignment needed: [specific outcome]
Meeting length: [minutes]
Swing-vote function: [function or "none"]

REQUEST
Produce: function-specific framing, a headline slide title, a jargon translation table, a prep brief with the one thing each function needs, and the one question to ask each function before the meeting ends.

QUESTION
Please ask me any clarifying questions before generating output. If any brackets above are empty or vague, ask in one numbered list and wait. Accept narrative answers. Proceed after I answer; if I skip something, ask one short follow-up or note the assumption.

SITUATION
[Paste situation here]

Example scenario, Tooling Rework Launch Readiness Review

Example scenario, Tooling Rework Launch Readiness Review

I am a program engineer at a Tier 1 supplier. I have a 30-minute launch readiness review starting in an hour. The room will have Finance, Quality, Program, Operations, and Purchasing. Tooling trials on our new instrument panel carrier are 2 weeks behind because of an undercut geometry issue on the B-side core pull. I need to approve a 2-week timeline extension and a die rework. Estimated cost: 45K. PPAP samples can be re-cut from the reworked tool with no impact on validation. Quality wants to know if PPAP samples will still be representative. Purchasing wants the cost impact. Program is the swing vote because they have to absorb the timeline. Please prepare me for the meeting.

Quality Check, Before Sharing Translated Technical Content

  • Is it technically accurate? AI simplifies well but can distort. Have an engineer review every translated summary for accuracy.
  • Did you quantify the impact? "Significant quality improvement" means nothing. "12 percent scrap reduction saving 80K per year" gets budgets approved.
  • Is the ask clear? Every communication to leadership should end with exactly what you need from them and by when.
  • Would your toughest customer accept this? If the language is too casual, too vague, or too optimistic, tighten it up.
  • Did you keep the technical detail accessible? Put it in backup slides or footnotes.
  • Did you verify the AI output, not just trust it? AI catches a lot. AI can also miss something obvious. You are still the engineer.

Copilot Basics (If You Use Microsoft 365)

  • Work Mode vs. Web Mode: Work mode can summarize documents in your SharePoint. Web mode is best for creating new translations from content you paste.
  • Use the Mini Version on Copilot Free. The Main Version works best on licensed Copilot, ChatGPT, Claude, or Gemini.
  • How to save prompts: Save the Technical Peer Reviewer and the Executive Summary Translator where you can access them fast.
  • Reminder: Be careful pasting proprietary test data, customer-specific information, or IP-sensitive content into external AI tools.

What to Try This Week

  • Pick one engineering document. A draft AI vision validation, a predictive maintenance threshold defense, a DFMEA, a control plan, or a root cause writeup. The one sitting on your desk because you are not 100 percent sure it is ready.
  • Run Prompt 1 (Technical Peer Reviewer) on it. See what AI catches.
  • Pick one engineering report. Something dense that you owe a non-technical leader.
  • Run Prompt 2 (Executive Summary Translator) on it. Send the 100-word summary instead of the 15 pages.
  • Identify one senior engineer you cannot afford to lose. Schedule 30 minutes with them. Run Prompt 3.

Resources & Next Steps

Custom Prompt Writer. Prompt to paste.

Custom Prompt Writer. Prompt to paste.

ROLE: You are a world-class prompt writer. I pay you $1,000 per hour. You are VERY expensive because you are the best in the world at building prompts for ChatGPT.

RULE: YOU ALWAYS WRITE IN THE 2nd PERSON! For example, for ROLE…you would say "YOU ARE A WORLD CLASS COPYWRITER".

REQUEST: You must deliver me a PERFECT prompt for Copilot that I can copy and paste and that allows the user to request ChatGPT to deliver the desired outcome. If you don't have enough information, ask the user follow-up questions until you have confidence that the prompt is perfect.

GOAL: To deliver a CONCISE Copilot prompt that delivers the exact result the user is looking for.

STRUCTURE FOR EACH AND EVERY FINAL PROMPT YOU DELIVER:

- INPUTS:
    - You will always start by gathering "Inputs". Unique information the user will GIVE YOU each and every time.
    - Please ask about their situation, their problems and what they are desiring this prompt to do
    - Your job is to only have 1-4 inputs each time. VERY VERY minimal inputs. Every time you build a prompt, you need Copilot to ask the user for some information, and you will create as many inputs as possible FOR THE USER. The user should do MINIMAL amount of work.
- ROLE: You will define the role of the prompt. DO NOT ASK THE USER. The role is always some sort of world-class expert in their field that will help.
- REQUEST: Define the objective of the prompt  - what is the purpose of the prompt? What does the prompt achieve?
- STRUCTURE: Define the exact structure of the prompt. Take a guess at the structure first UNLESS The user defines the structure ahead of time.
- TONE: Use a tone for this prompt that makes sense UNLESS the user defines the tone
- GOAL: Listen to what the user is telling you about the goal of the prompt and make sure you add the goal in his section.
-NEXT STEP: include something like this "Ask User for this information (include bulletpoint list of information or inputs you need every single time the prompt is used)

DETAILS:

- Once you have read all of this…ask the user to describe what kind of prompt they are looking to create and what the goal of the prompt should be.
- Do not say anything about being a prompt writer, ONLY ASK ME 3 QUESTIONS that will allow you to build the best prompt ever

TASK: Take that information and BUILD A WORLD-CLASS Copilot PROMPT

User will say "I'm ready to build a prompt"

And you will ask the user for their current situation, what the prompt will help with and what they are desiring the prompt to build and do

REMEMBER WHEN YOU DELIVER A PROMPT YOU ALWAYS INCLUDE EACH AND EVERY SECTION ABOVE

Deliver the full prompt directly, without further explanation or summary.

Questions or feedback? cheryl@cherylthompson.net


The Mental Model: Review, Translate, Capture

Today's session was built on three ideas. Your engineering work is only as good as what gets caught before it goes out. Your technical expertise is only as valuable as your ability to communicate it. Your senior experts will not be there forever.

Pillar

Prompt

What It Does

Review

Prompt 1, The Technical Peer Reviewer

AI as a senior peer reviewer.

Translate

Prompt 2, The Executive Summary Translator

AI as your business translator.

Capture

Prompt 3, The Engineering Judgment Capture

AI as your knowledge interviewer.

Translate (extended)

Prompt 4, Customer Technical Explainer
Prompt 5, Cross-Functional Meeting Translator

AI translates technical work for customers and cross-functional audiences.

You are not a worse engineer than you were five years ago. You just have help now.


Your work already matters. Now you have the tools to make sure it lands, holds up, and lasts.

  • Run Prompt 1 on one engineering document this week.
  • Run Prompt 2 on one dense report.
  • Schedule 30 minutes with one senior engineer at retirement risk.

Related resources

Keep working with the material.

The prompts and exercises are included above. Save a useful prompt and return to it with your next task.

For organizations

Bring this learning
to your team.

These learning experiences can be adapted into workshops, role-based learning series, and practical AI enablement programs for teams, associations, and professional communities.

Explore organizational learning