When over 100 executives log in to a webinar called “Resistance, Roll-outs, and AI” (​replay here​), you know the urgency to move from talking about AI to making it a success is real.

I invited Phil Kirschner, a former global McKinsey leader and the author of ​The Workline​, to examine the state of AI adoption and how to get stalled momentum moving again.

Over 60 lively minutes and dozens of audience questions, Phil shared the playbook for successfully introducing and implementing generative AI in companies regardless of size, industry, or geography.

Here’s a distilled set of takeaways I’m now sharing with clients, as well as in this newsletter.

1. The Launch Isn’t the Finish Line

Usually, Phil is the one with the “line” puns, but I couldn’t resist this one.

What the “finish line” looks like is a critical consideration that many overlook.

Phil opened Tuesday’s event with the statistic that should keep any transformation leader up at night: seventy percent of large change programs fail to achieve their intended goals.

But there’s a story behind this number, he explained.

Most organizations count “turned on the software” as success.

The real question is whether behavior shifted; that’s the line that needs to be crossed.

For example, if marketing is still spending most of their time writing newsletters and social posts with ChatGPT or Gemini sitting idle, those license fees are wasted, and the team is working in the past.

Phil’s tip: Anchor the roll-out to a business outcome before you start implementation.

2. Resistance Is Specific; Treat It That Way

Phil’s toolkit includes Lamarsh’s ten forms of resistance, but the headline is simple: people balk for different reasons.

Some genuinely don’t believe today’s workflow is broken. Others dislike the future you’re painting for them.

And while “cautious skeptics” can be won over, active resisters will be a pain. And that’s okay.

What rarely works is pretending everyone is excited about what you’re introducing – even if YOU believe the exciting opportunities AI brings.

Your job here is to diagnose the state of play and then focus on the middle portion of people who can be moved.

3. Lead With Emotion, Not Enablement

My personal highlight from Phil’s presentation was a case study from BetterUp.

The wellbeing platform recently kicked off its internal AI rollout by allowing employees to “sit with the fear” in facilitated sessions before considering training.

There are clear reasons why only ​8% of knowledge workers use AI daily​, and one of them is that no one wants to be training their robot successor.

Let’s address that before we start talking about prompts, techniques, and tools. As Alexandra Samuel ​told me last year​:

“The most perhaps foundational piece of this is that you need people to feel like their jobs are not at risk by [adopting AI]. You're not going to automate yourself out of work. And I think there's a lot of anxiety that people have around that.”

Try beginning your next all-hands with a frank acknowledgment: “Some of you worry AI will deskill your role. Let’s surface those concerns together.”

Addressing emotion up front buys you credibility for the technical deep dive that follows.

4. Map the Forces Before You Pick the Tool

One way to understand the blockers to adoption is through a model from Jobs-to-Be-Done.

On one page, list what pushes people away from the status quo, what pulls them toward the future, what anxieties hold them back, and what habits anchor them in place.

During our session, Phil applied the model to a pilot HR chatbot.

Within minutes, we understood the real hidden blocker: HR business partners feared losing their “trusted adviser” identity.

It’s exactly why another training video isn’t the solution here. Redefining the HRBP’s role around complex employee cases, work that the bot can’t mimic, would solve the problem.

Without deep listening that follows such frameworks, these kinds of insights are never uncovered.

5. Remember the Time-Lag on the Change Curve

Perhaps the most significant insight that should resonate with everyone who completed our ​Executive AI Boot Camp​: your newfound excitement doesn’t translate to those who are hearing about AI for the first time.

By the time a CEO is ready to celebrate AI initiatives, they’ve usually been processing the decision for months.

Middle managers hear the news on day one; frontline employees a few days later.

They are still at the top of the anxiety dip while leadership is already climbing the slope of enlightenment. Stating that gap out loud—“We’ve been exploring this for six weeks; today is your first exposure”—creates space for real questions and honest pace-setting.

This also answers an audience question from ​Henrik Jarleskog​, who noted that “Adoption often collapses because of the ‘not-invented-here’ syndrome.”

Phil’s solution was simple, and it solves the problem above: get people in as early as possible.

For any roll-out you’re planning, who could help provide insights today that would result in a better solution and inherent buy-in?

The Bottom Line: A Roll-Out Success

AI rollouts fail less because of the strength of the platform or model, and more because the change narrative is weak.

Treat emotion as data, involve employees in shaping the destination, and define success in human terms long before you send any money to OpenAI or Microsoft.

See you next week,

PS: We can help! We’re starting to roll out the first AI Snapshots, a new product to understand where a company’s biggest AI challenges and opportunities lie. 👉 Request more information here.

‍

Category Essentials: Deep Research

This week, I asked the community which Deep Research tool they're using regularly.

A level up from AI chat with web browsing, these Deep Research tools are real game-changers, especially for more complex and niche topics.  

The best part? You can try them (several times!) even if you're not on a paid plan.

Here's the order of preference from our community, with a few tips:

#1 ChatGPT Deep Research

ChatGPT’s Deep Research tool runs on its o3 reasoning model. What stood out is that it starts by clarifying your research question before diving in. That extra step helps ensure the final report is sharp—and that you don’t waste your credits chasing the wrong angle.

Even better, ​Antony Slumbers​ suggests using the 4o model to polish your prompt before sending it to o3 for Deep Research. (Same tip applies to Gemini, coming next.)

#2 Gemini Deep Research

Gemini takes a slightly different approach: it proposes a multi-step research plan first, which you can review and edit before it runs.

​Wendy Nguyen​ shared that even though she spends most of her time on ChatGPT, she still favors Gemini's Deep Research and now uses it just as often, because the output quality holds up.

#3 Perplexity Deep Research

If you’re already Perplexity-ing more than Googling, try its Deep Research mode. It draws from around 20x more sources than the standard search mode and 10x more than the Pro search mode, yielding deeper and sharper answers.

What’s unique here is how it suggests follow-up questions and ideas for you to explore the topic further. If you’re the type who enjoys rabbit holes (in a good way), this one’s worth a go.

Want me to cover a specific category and/or AI tool next? Let me know here.

Key Levers to Scale Enterprise GenAI, Against ‘Brain Damage,’ AI Makes Us More Human Leaders

I read dozens of AI newsletters weekly, so you don’t have to. Here are the top 3 insights worth your attention:

#1 Scaling GenAI: An Example Across 20,000 Employees

Here’s another great example and set of lessons on rolling out GenAI at scale. And it’s from Novo Nordisk, the $500B pharma giant, deploying Microsoft Copilot across 20,000 employees.

In this MIT Sloan article, the authors share what actually moved the needle: targeted and role-specific training, champion networks (led by senior staff who are not digital natives!), and honest feedback loops to overcome adoption dips and AI shaming.

👉 ​Read the full story and the takeaways here.​

#2 Against ‘Brain Damage’

Honored to highlight this topic yet again: AI doesn't melt your neurons, but it can dull your thinking if you're not careful. That's from Ethan Mollick.

In his latest post, Ethan walks through studies showing how AI can short-circuit learning, creativity, and collaboration when used mindlessly.

But he also shares how the right prompts, structure, and sequencing (think before you prompt) can actually make you sharper.

👉 ​Read it here.​

#3 GrowthLoop CEO: AI Make Me a More Human Leader

Chris O’Neill (ex-Google, Evernote, now CEO at GrowthLoop) argues that the real value of AI isn’t just speed—it’s space.

By offloading the mental clutter (status updates, research, drafts), AI freed him up to mentor, think, and lead with intention. Less reaction, more reflection.

His take: leadership isn’t immune to AI disruption, and shouldn’t be. If everyone will lead a team of AI agents soon, execs better start designing for that now.

👉 ​Worth the read here.

‍

Prompt of the Week

A good prompt makes all the difference, even when you're just using a core LLM.

This one flips the usual script and makes AI actually your personal $500/hr consultant. Instead of giving ChatGPT a task and hoping for the best, it starts by making the AI ask you smart, targeted questions first.

Try it for more useful, personalized outputs, whether it's an email, plan, or a strategy doc. Just copy everything and paste to your AI like ChatGPT, then send it your vaguest, most half-assed request and it will guide you.

Your $500/Hour Consultant

You are Lyra, a master-level AI prompt optimization specialist. Your mission: transform any user input into precision-crafted prompts that unlock AI's full potential across all platforms.

You are Lyra, a master-level AI prompt optimization specialist. Your mission: transform any user input into precision-crafted prompts that unlock AI's full potential across all platforms.

## THE 4-D METHODOLOGY

### 1. DECONSTRUCT
- Extract core intent, key entities, and context
- Identify output requirements and constraints
- Map what's provided vs. what's missing

### 2. DIAGNOSE
- Audit for clarity gaps and ambiguity
- Check specificity and completeness
- Assess structure and complexity needs

### 3. DEVELOP
- Select optimal techniques based on request type:
 - **Creative** → Multi-perspective + tone emphasis
 - **Technical** → Constraint-based + precision focus
 - **Educational** → Few-shot examples + clear structure
 - **Complex** → Chain-of-thought + systematic frameworks
- Assign appropriate AI role/expertise
- Enhance context and implement logical structure

### 4. DELIVER
- Construct optimized prompt
- Format based on complexity
- Provide implementation guidance

## OPTIMIZATION TECHNIQUES

**Foundation:** Role assignment, context layering, output specs, task decomposition

**Advanced:** Chain-of-thought, few-shot learning, multi-perspective analysis, constraint optimization

**Platform Notes:**
- **ChatGPT/GPT-4:** Structured sections, conversation starters
- **Claude:** Longer context, reasoning frameworks
- **Gemini:** Creative tasks, comparative analysis
- **Others:** Apply universal best practices

## OPERATING MODES

**DETAIL MODE:**
- Gather context with smart defaults
- Ask 2-3 targeted clarifying questions
- Provide comprehensive optimization

**BASIC MODE:**
- Quick fix primary issues
- Apply core techniques only
- Deliver ready-to-use prompt

## RESPONSE FORMATS

**Simple Requests:**
```
**Your Optimized Prompt:**
[Improved prompt]

**What Changed:** [Key improvements]
```

**Complex Requests:**
```
**Your Optimized Prompt:**
[Improved prompt]

**Key Improvements:**
• [Primary changes and benefits]

**Techniques Applied:** [Brief mention]

**Pro Tip:** [Usage guidance]
```
## WELCOME MESSAGE (REQUIRED)

When activated, display EXACTLY:

"Hello! I'm Lyra, your AI prompt optimizer. I transform vague requests into precise, effective prompts that deliver better results.

**What I need to know:**
- **Target AI:** ChatGPT, Claude, Gemini, or Other
- **Prompt Style:** DETAIL (I'll ask clarifying questions first) or BASIC (quick optimization)

**Examples:**
- "DETAIL using ChatGPT — Write me a marketing email"
- "BASIC using Claude — Help with my resume"

Just share your rough prompt and I'll handle the optimization!"

## PROCESSING FLOW

1. Auto-detect complexity:
  - Simple tasks → BASIC mode
  - Complex/professional → DETAIL mode
2. Inform user with override option
3. Execute chosen mode protocol
4. Deliver optimized prompt

**Memory Note:** Do not save any information from optimization sessions to memory.

‍

👉 Try it, tweak it, and save it for your future use. If this prompt is helpful (or if you made it better), I’d love to hear how here.