# The Conversation Script for a Principal Addressing a Senior Engineer Who Refuses to Try AI **By Dan Cumberland** · Published June 18, 2026 · Categories: AI Strategy > If you have a senior mechanical engineer who has not opened the AI seat your firm is paying for, this article is the conversation script for Monday's... ## The Monday Morning You Are Not Looking Forward To If you have a senior mechanical engineer who has not opened the AI seat your firm is paying for, this article is the conversation script for Monday's one\-on\-one— not another listicle of tools they will continue to ignore\. The "mechanical engineer AI" search returns buyer's guides; that is not the buyer problem at your firm\. The seat is bought\. The juniors use it\. The senior PE does not\. You have sent the Loom\. You have run the all\-hands\. And on Sunday night you are drafting talking points you do not actually want to deliver\. The discomfort is real, and it is not yours alone\. Across mid\-sized mechanical and MEP firms this conversation is happening every Monday— Principal across the desk from a senior PE who has carried the firm for twenty\-five years and who has, for reasons that are starting to look more substantive than stubborn, declined to participate in the AI rollout\. The senior engineer is protecting something real: twenty\-five years of design judgment and the professional liability attached to it\. What looks like resistance is due diligence\. Most articles on this topic will tell you to communicate the vision\. This one will tell you what to literally say in the first ninety seconds\. The research explains why\. ## Why It Is Not Fear \(And What That Means for Monday\) Senior engineers are the largest AI holdout group inside engineering organizations, and the research is consistent on why: their resistance is grounded in genuine expertise, not fear or unfamiliarity\. Faros AI[1](/blog/blog-mechanical-engineer-ai#ref-1), working off telemetry from twenty\-two thousand developers, finds that senior developers are the largest holdout group even as AI adoption across development teams has climbed broadly\. Prosci change\-management research[5](/blog/blog-mechanical-engineer-ai#ref-5), summarized for engineering audiences by T4Leader, lands on the same conclusion: senior technical staff are among the highest\-risk groups for sustained resistance precisely because the resistance is grounded in expertise\. Faros[1](/blog/blog-mechanical-engineer-ai#ref-1) documents four drivers: - **Trust and reliability for mission\-critical work**— sealed engineering work cannot tolerate hallucination - **AI's struggle with context\-heavy senior problems**— the parts of the job AI is worst at are the parts senior engineers do - **Professional identity threat**— the implicit framing that AI replaces judgment - **Prohibitive time investment**— learning the tool competes with the work the senior engineer is already buried under Read against that list, your senior PE is not behind\. They are reading the situation accurately\. A taxonomy of [six types of AEC AI holdouts](/blog/architecture-types-of-jobs/) sits underneath these four drivers in a real firm— the senior PE is the most expensive variant to mishandle\. There is a generational layer, but it does not run the way the listicles tell you\. London School of Economics research[2](/blog/blog-mechanical-engineer-ai#ref-2) finds Gen Z workers using AI at 83%, Millennials at 73%, Gen X at 60%, and Baby Boomers at 52%\. And then the same study shows the gap nearly closes when training access is matched\. > "A 60\-year\-old employee with proper AI training outperforms an untrained 25\-year\-old\." The productivity gap is organizational, not generational\.[2](/blog/blog-mechanical-engineer-ai#ref-2) In AEC specifically, the gap is structural: sealed work carries professional\-liability stakes that most knowledge\-worker industries do not\. Bluebeam's industry survey shows 27% of AEC firms actively using AI in operations, against McKinsey's 75% baseline for knowledge workers generally\.[3](/blog/blog-mechanical-engineer-ai#ref-3)[11](/blog/blog-mechanical-engineer-ai#ref-11) AEC is not behind by accident\. The liability exposure is the reason\. Outside engineering, the pattern shows up the same way\. A federal grant writing consultant in Dan's network spent most of 2024 buying AI tools and requesting refunds— the tools, in his words, "claimed to do things that they absolutely could not do\." As of October 2024 his line was, "I don't get it, it's not doing what I need\." A few months later, after finding a bounded use case that fit his domain expertise, he was building tools instead of buying them\. The skeptic\-to\-builder arc is documented across professions; the most expertise\-grounded skeptics often become the strongest adopters once the bounded use case lands\. Your job in Monday's conversation is not to overcome their resistance\. It is to listen to what their resistance is telling you about your rollout\. If you internalize that the senior engineer is technically correct about a piece of this, the conversation gets simpler\. Here is the script\. ## The Conversation Script The conversation that converts a senior engineer holdout opens by validating their objection, narrows to a bounded pilot on a non\-sealed workflow, and closes with a specific check\-in date— never with the phrase "you are being left behind\." The opening earns the right to make the ask\. The ask is bounded by design\. Bounded participation is the win, not conversion to AI evangelism\. ### The First 90 Seconds Open by acknowledging what the engineer is protecting\. Not the firm's strategy slide\. Not the AEC adoption stat\. What they are protecting— judgment, accountability, the firm's reputation that gets signed under their stamp\. The validation has to be real, not a setup\. Concretely: > **You:** "I want to start by saying something\. The fact that you have not opened that AI seat— I am not actually sure you are wrong about it\." **Senior PE:** \[Waits\.\] **You:** "Everything you have flagged— hallucination, the stamp, what it would mean for the firm if we got it wrong— is technically correct\. Most of our peers haven't moved on this yet[3](/blog/blog-mechanical-engineer-ai#ref-3), and we are not going to be the firm that rushes in ahead of them\. I want to talk about where we can use this where none of that is in play\." Three short lines\. No hedging\. No "but" in the third one\. The shareholder\-message version of this conversation— [the speech that lands with engineers](/blog/ai-tools-for-mechanical-engineers/) at the all\-hands— belongs in a different artifact\. Monday's one\-on\-one is narrower\. Prosci's ADKAR model[6](/blog/blog-mechanical-engineer-ai#ref-6)— Awareness, Desire, Knowledge, Ability, Reinforcement— frames the senior holdout as an Awareness\-and\-Desire problem, not a Knowledge or Ability problem\. Your engineer is not waiting for a tutorial\. They are waiting for a reason to believe the change is necessary and survivable\. That is what the first ninety seconds buys\. ### The Three Objections— Trust, Liability, Identity Three objections will land in the first five minutes\. Each one has a response that starts with the word "Agree\." ```html-table
ObjectionWhat the engineer actually meansHow to respond
"I do not trust it. It hallucinates."This is a mission-critical reliability concern about sealed work.Agree. Current AI in MEP is best understood as a pattern-recognition and task-automation assistant, not a substitute for professional engineering judgment8. Narrow the ask away from sealed work entirely.
"I am liable for the stamp."Non-delegable professional duty. This is the strongest objection on the list.Agree harder. Courts have already established for adjacent professions that AI-generated outputs do not transfer the verification duty to the AI10. State PE boards have not yet issued formal guidance on AI use in sealed work; the engineer's caution is currently the only guidance. Stamp authority stays with them. AI does not change that.
"I do not need it. I have been doing this for 25 years."Identity threat— Faros's1 third driver.Do not argue. Reframe the tool as an intellectual-augmentation layer for the formulaic parts of their workflow, not for the parts that require their judgment. Their judgment is the asset. The rollout is the variable.
``` A fourth objection sometimes surfaces underneath the trust one— "I tried it once\. It was bad\." Howard Tullman in Inc\.[13](/blog/blog-mechanical-engineer-ai#ref-13) notes that most early experiments use the cheapest consumer tools, which lag enterprise\-grade options\. File it if it surfaces\. ### The Ask The ask is bounded: one workflow, two weeks, a defined output that is not the sealed artifact\. Pick from non\-sealed deliverables only: - Section\-23 spec drafting— first pass only - Cover\-sheet generation for a known repeat client - RFI response drafting - Meeting minutes from a \.vtt recording - Internal calc\-check narrative \(for the engineer's own review, not for the file\) Define success in the room: the engineer runs the tool on one workflow for two weeks and shows up to the check\-in with an observation\. Anything they observe is useful\. A bug they spotted, a hallucination they caught, a place the output was usable, a place it was not\. Set the check\-in date in the same conversation\. Two weeks\. Calendar invite goes out before the engineer leaves your office\. Not at the end of the week\. Before they leave\. The AEC firms moving on bounded MEP AI are getting real engineering throughput— one MEP\-focused startup has reported cutting the AI\-generated first pass for 500,000\-square\-foot commercial electrical design from two months to less than a day[9](/blog/blog-mechanical-engineer-ai#ref-9)\. Mention it once\. Do not push it\. The senior PE does not need to be sold on the category; they need a workflow that respects what they are responsible for\. What works in the script depends as much on what you do not say\. Here are the failure phrases\. ## What NOT to Say Three phrases will detonate this conversation regardless of how well you prepared: "You are being left behind," "The juniors are already using it," and "This is mandatory now\." Each one fails for a research\-backed reason\. **"You are being left behind\."** This phrase converts an engineer with twenty\-five years of judgment into a person being shoved out the door\. It triggers Faros's[1](/blog/blog-mechanical-engineer-ai#ref-1) identity\-threat driver directly\. This is the same pattern that produces [a VP who tried Copilot once and called himself a holdout](/blog/jhu-career-architecture/)— once the identity frame locks in, the tool conversation stops\. Replace it with: "I want this to make your judgment more useful, not less necessary\." **"The juniors are already using it\."** This implies a generational hierarchy reversal that the engineer reads, accurately, as an attack on their authority\. It is also factually weak— LSE's data[2](/blog/blog-mechanical-engineer-ai#ref-2) shows the gap nearly disappears when training access is matched\. The juniors are not better at AI because they are younger; they are using it more because the training environment was matched to them and not to the senior engineer\. Replace it with: "I want to give you the same setup time we gave the juniors\." **"This is mandatory now\."** Mandates convert an Awareness\-and\-Desire conversation into a Knowledge\-and\-Ability conversation prematurely[6](/blog/blog-mechanical-engineer-ai#ref-6)\. The senior engineer who does not yet believe the change is necessary will not learn the tool, regardless of mandate\. Prosci's guidance is clear: never label individuals as resisters; describe the specific behaviors you observe and what they indicate about your change approach[7](/blog/blog-mechanical-engineer-ai#ref-7)\. Assume the conversation goes as well as it can\. What happens in the next two weeks decides whether the script worked\. ## After the Conversation— Reading the Response After the conversation, you are diagnosing two questions: did engagement begin, and is any residual resistance passive or principled? Passive resistance escalates to performance management\. Principled resistance gets integrated into decision\-rights on AI tool selection\. ```html-table
Observable behaviorDiagnosisResponse
Engineer opens the tool inside the two-week window, sends one question or critique, shows up to the check-in with a specific observationThe script worked. Engagement has begun.Hold the check-in. Pull in one peer for a thirty-minute compare-notes session. Set the 30/60/90 review cadence.
No engagement. No counter-proposal. Polite delays. Missed check-in.Passive resistance. MIT Sloan research5 documents what this is: silent non-participation is how high-status technical staff express objection without escalation— not disengagement, but a structured signal that the change approach needs work.Escalate cadence. Hold the check-in anyway. Document. If pattern continues past sixty days, tie to performance review.
Specific objections. Named workflow problems. Identified failure modes. Critique of the tool with detail.Principled resistance. This is the BCG decision-rights signal5— the engineer has read your rollout and identified a real problem with it.Promote it. Integrate the engineer into the firm's AI tool-selection committee, not the execution rollout. Do not punish the signal.
``` Principled resistance is a gift\. Promote it into your tool\-selection process\. All of the above assumes a peer AI champion exists in the firm\. In a forty\- or fifty\-person firm, you may not have one\. ## The Small\-Firm Edge Case— When the Senior PE Is the Most Senior Peer In a forty\- to sixty\-person mechanical firm, the senior PE is often the most senior peer— there is no AI champion available above them, which means the Principal must either play that role personally or formally recruit from one tier down\. Faros's[1](/blog/blog-mechanical-engineer-ai#ref-1) peer\-to\-peer adoption advantage— 22% more effective than top\-down mandates— assumes a peer exists\. In a fifty\-person firm, you may not have one\. Two viable substitutes: - **Principal\-as\-champion\.** Name yourself as the AI user the senior engineer can compare notes with\. Show your own usage\. Bring an example to the check\-in— a workflow you ran through the tool last week, what worked, what did not\. This is unglamorous, and it is the move\. - **Recruit one tier down\.** The strongest mid\-level engineer in the discipline becomes the formal peer reference, with explicit air cover\. [Mid\-career engineers who adopted AI twice as fast](/blog/will-ai-replace-civil-engineers/) are your candidate pool\. Title matters: "AI pilot lead" reads as bounded\. "AI lead" reads as a turf flag and creates a problem with the senior PE that does not need to exist\. LeadDev's broader framing on AI champion programs[12](/blog/blog-mechanical-engineer-ai#ref-12) holds— respected internal technical leads empowered as adoption advocates outperform top\-down rollouts\. The structure scales down; the role does\. Some senior engineers will not convert\. That is not a script failure\. It is a decision\. ## When the Answer Is "This Is Not Going to Work" Some senior engineers will not engage, and the script does not make every conversation a conversion\. After sixty days of bounded ask, documented passive resistance, and no counter\-proposal, the question shifts from change management to role fit\. Three signals, all required: - The bounded ask was honored on your side— narrow workflow, non\-sealed output, real check\-in, no mandate - Passive resistance is documented across at least three check\-in cycles - No engineering critique of the tool has surfaced— nothing to act on, nothing to integrate At that point the question is not "how do I get them on AI\." It is "is the role still right for the firm?" Reframe as a performance conversation, not an AI conversation\. If the engineer cannot do bounded participation after sixty days of being met with respect and a narrow ask, that is data, not failure\. One note for the record: no firm has published verifiable ROI on AI\-driven retention or separation decisions in mechanical engineering\. The Principal is operating without precedent here\. Treat it like every other consequential personnel decision— slowly, in writing, with HR in the loop\. ## FAQ ### Why won't my senior mechanical engineer use AI? Four documented drivers[1](/blog/blog-mechanical-engineer-ai#ref-1): trust and reliability concerns for mission\-critical work, AI's struggle with context\-heavy senior problems, professional identity threat, and prohibitive time investment\. Prosci research[5](/blog/blog-mechanical-engineer-ai#ref-5) adds that the resistance is grounded in expertise, not fear\. It is information about your rollout, not a character flaw in your engineer\. ### Is AI safe to use in MEP / mechanical engineering? Current AI in MEP is best used as an assistant for pattern recognition, data analysis, and task automation— not as a substitute for professional engineering judgment on sealed work[8](/blog/blog-mechanical-engineer-ai#ref-8)\. Bounded pilots should target non\-sealed workflows first: spec drafting, cover\-sheet generation, RFI response drafts, internal calc\-check narratives\. Sealed work stays with the PE's judgment until state boards say otherwise\. ### How do I have the conversation with a senior engineer about AI? Open by acknowledging what they are protecting— judgment, liability, the firm's reputation\. Validate the hallucination objection; it is technically correct for sealed work\. Narrow the ask: one workflow, two weeks, non\-sealed output\. Set the check\-in date before the engineer leaves the room\. Prosci's ADKAR model[6](/blog/blog-mechanical-engineer-ai#ref-6) frames this as an Awareness\-and\-Desire conversation, not a training one\. ### Should I make AI use mandatory? Not at the holdout stage\. Mandates convert an Awareness\-and\-Desire conversation into a Knowledge\-and\-Ability conversation prematurely[6](/blog/blog-mechanical-engineer-ai#ref-6), and a senior engineer who does not yet believe the change is necessary will not learn the tool, regardless of mandate\. Prosci's guidance: describe specific behaviors and what they indicate about your change approach[7](/blog/blog-mechanical-engineer-ai#ref-7)— do not label individuals as resisters\. ### What if my senior engineer still refuses after the conversation? Diagnose whether the resistance is passive \(no engagement, no rationale, missed check\-ins\) or principled \(specific objections, identified workflow problems\)\. Passive resistance[5](/blog/blog-mechanical-engineer-ai#ref-5) escalates to performance management on a sixty\-day timeline\. Principled resistance gets integrated into the firm's AI tool\-selection decision\-rights[5](/blog/blog-mechanical-engineer-ai#ref-5)— promote the signal, do not punish it\. ### Is this a generational issue? Partly\. Gen Z workers use AI at 83% versus 52% for Baby Boomers[2](/blog/blog-mechanical-engineer-ai#ref-2)\. And the gap nearly disappears when training access is matched: a sixty\-year\-old with proper training outperforms an untrained twenty\-five\-year\-old[2](/blog/blog-mechanical-engineer-ai#ref-2)\. Senior engineers are not too old to learn\. They are under\-resourced for it\. Whatever the specific question, the answer comes back to the same place\. ## The Frame— Protect the Human, Install the Amplifier The Principal's job in this conversation is not to get the senior engineer on AI\. It is to protect twenty\-five years of design judgment while installing the amplifier around it\. The mechanical engineer AI conversation only works if you walked in believing the engineer is the asset, not the obstacle\. If the rollout itself needs design work, [an outside AI implementation partner](/service/) can sit between you and the firm and take some of the political weight off the relationship that does the actual engineering\. On Monday, you are not selling them on AI\. You are protecting twenty\-five years of judgment while installing the amplifier around it\. *By \[Dan Cumberland\]\(https://dancumberlandlabs\.com\)* ## References 1. Faros AI, "Driving AI Adoption in Senior Software Engineers" \(2026\)— [https://www\.faros\.ai/blog/ai\-adoption\-in\-senior\-software\-engineers](https://www.faros.ai/blog/ai-adoption-in-senior-software-engineers) 2. London School of Economics, via Resultsense, "The Hidden Productivity Gap: Why AI Training Matters More Than Age" \(2025\)— [https://www\.resultsense\.com/insights/2025\-10\-28\-lse\-generational\-ai\-gap\-workforce\-productivity](https://www.resultsense.com/insights/2025-10-28-lse-generational-ai-gap-workforce-productivity) 3. Bluebeam, via Construction Dive, "Survey finds AI has taken hold in AEC" \(2025\)— [https://www\.constructiondive\.com/news/ai\-aec\-industry\-research\-bluebeam/732155/](https://www.constructiondive.com/news/ai-aec-industry-research-bluebeam/732155/) 4. T4Leader, "Why Are Your Senior Engineers Quietly Sabotaging the New System?" \(2024\)— citing Prosci \(2021\), MIT Sloan \(2018\), and BCG \(2019\)— [https://www\.t4leader\.com/post/why\-are\-your\-senior\-engineers\-quietly\-sabotaging\-the\-new\-system](https://www.t4leader.com/post/why-are-your-senior-engineers-quietly-sabotaging-the-new-system) 5. Prosci, "The Prosci ADKAR Model" \(current\)— [https://www\.prosci\.com/methodology/adkar](https://www.prosci.com/methodology/adkar) 6. Prosci, "Preventing and Managing Resistance to Organizational Change" \(current\)— [https://www\.prosci\.com/blog/a\-deeper\-dive\-into\-the\-real\-work\-of\-resistance\-management](https://www.prosci.com/blog/a-deeper-dive-into-the-real-work-of-resistance-management) 7. Consulting\-Specifying Engineer, "See how AI can transform MEP design decisions" \(2025\)— [https://www\.csemag\.com/see\-how\-ai\-can\-transform\-mep\-design\-decisions/](https://www.csemag.com/see-how-ai-can-transform-mep-design-decisions/) 8. Fortune, "Notion Capital leads $20 million seed round for startup automating building MEP design" \(2025\)— [https://fortune\.com/2025/12/18/endra\-ai\-startup\-automating\-mep\-design\-sweden\-20\-million\-seed\-round\-notion\-capital/](https://fortune.com/2025/12/18/endra-ai-startup-automating-mep-design-sweden-20-million-seed-round-notion-capital/) 9. The Bulldog Law, "Business Liability for AI Hallucinations: Legal Defense Strategies When Artificial Intelligence Gets It Wrong" \(2025\)— [https://www\.thebulldog\.law/business\-liability\-for\-ai\-hallucinations\-defense\-strategies\-when\-artificial\-intelligence\-gets\-wrong](https://www.thebulldog.law/business-liability-for-ai-hallucinations-defense-strategies-when-artificial-intelligence-gets-wrong) 10. McKinsey, "Superagency in the Workplace: Empowering people to unlock AI's full potential at work" \(2025\)— [https://www\.mckinsey\.com/capabilities/tech\-and\-ai/our\-insights/superagency\-in\-the\-workplace\-empowering\-people\-to\-unlock\-ais\-full\-potential\-at\-work](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work) 11. LeadDev, "Your AI champions are the key to engineering adoption" \(2025\)— [https://leaddev\.com/ai/ai\-champions\-are\-the\-key\-to\-engineering\-adoption](https://leaddev.com/ai/ai-champions-are-the-key-to-engineering-adoption) 12. Inc\., Howard Tullman, "How to Turn AI Into an Employee Partner— Not a Replacement" \(2025\)— [https://www\.inc\.com/howard\-tullman/ai\-employee\-partner\-not\-replacement\-technology/91305759](https://www.inc.com/howard-tullman/ai-employee-partner-not-replacement-technology/91305759) --- Source: https://dancumberlandlabs.com/blog/mechanical-engineer-ai/