#35 | 10 Expert AI POVs + Blind Spots

TL;DR: I read 100 expert opinions on AI and found that half of their advice assumes you control your workplace. What if you don’t? Here’s what to do.

👋 Hi there,

A few days ago, I decided to read 100 expert opinions from AI’s brightest minds.

Before long, my desktop looked (digitally) cluttered, and my brain felt drained. 100 is a lot. What was number 42 about?

Anyway, buried in all that expert noise, I found 10 insights that might help you fine-tune your AI strategy further.

These experts don’t claim to have all the answers. But they’re certainly wrestling with better questions than most of us.

However, half of what they’re saying probably requires a reality check to make it work for you, me, and us.

What’s ahead:

→ Competitive Edge (4 strategies) How to spot advantages others miss

→ Smart Implementation (3 POVs) Making AI stick in the real world

→ Human-AI Balance (3 principles) Staying valuable as AI improves

→ The Brutal Question What happens when expert advice doesn’t work in your reality?

Want to explore all 100 expert insights? ​Click here​.

Competitive Edge

How to spot advantages others miss

1. “Jobs are bundles of tasks” – ​Sayash Kapoor​ (Researcher/Academic)

Kapoor discussed this during a heated debate about whether AI would replace radiologists. One panelist claimed, “AI will make doctors obsolete within five years.”

Kapoor pushed back: “AI won’t replace radiologists. It’ll handle the routine scan reviews so radiologists can focus on complex cases and patient consultation.”

What works: You can audit exactly where AI threatens your value versus where you’re still irreplaceable.

Reality check: When I applied this to my work, I discovered AI could handle 70% of my research tasks, but none of my strategy conversations with others (or myself). That insight strongly influenced my workflow.

Adjust: Track your time for three days in 30-minute blocks. Label each block as “Could AI do this?” or “Needs human judgment.” If more than 60% fall into the first category, consider shifting your role toward the second—fast.

2. “Before prompting, ask: what’s the evidence this is the best way?” – ​Dr. Philippa Hardman​ (AI Learning Specialist)

Hardman was reviewing her students’ dissertations when she noticed a pattern. Their AI-generated literature reviews looked impressive—perfect formatting, academic tone, dozens of citations.

However, some of the academic references provided didn’t exist. The students had automated their research process without understanding what good research looked like.

What works: Evidence-based prompting separates strategic users from people who get and accept confident but wrong answers.

Reality check: It’s easy to assume we know what “good” looks like in our field until we try to explain it to AI. Often, our methods work through intuition and experience rather than clear principles.

Adjust: Pick your most important recurring task. Write down the three criteria you use to judge if it’s done well. Can’t do this clearly? Consider spending a week learning the fundamentals before touching AI.

3. “If your competitors aren’t adopting AI, you might hit 30% margins” – ​Marc Bhargava​ (Investor/VC Managing Director)

Bhargava was consulting for a mid-sized accounting firm struggling against larger competitors. The firm spent hours on routine bookkeeping while competitors offered cheaper services.

When it used AI to automate 80% of basic transactions, it suddenly could offer premium strategy consulting—something its larger competitors couldn’t match at its price point.

What works: Early AI adoption creates operational leverage that translates directly to profit margins.

Reality check: This window closes fast. The accounting firm’s advantage lasted 18 months before competitors caught up. The key isn’t just being first—it’s what you build while you’re ahead.

Adjust: Identify your most time-consuming, low-skill task that clients still pay well for. Use AI to cut that time by 70%. Reinvest those hours into higher-value work that’s harder for competitors to replicate.

4. “Don’t automate your role. Rebuild it from scratch” – ​Juliet Bailin​ (Investor/VC Partner)

Bailin was working with a startup founder who spent six months building an AI system to automate his customer onboarding emails. The result? Customers got responses 50% faster, but still dropped off at the same rate.

When Bailin asked, “What if onboarding wasn’t about emails at all?” He redesigned it as a video call series, and retention jumped 300%.

What works: First-principles thinking creates advantages that can’t be copied by improving existing processes.

Reality check: This requires killing sacred cows. The founder had to admit that email onboarding—something he’d perfected over the years—was the wrong approach entirely.

Adjust: Choose your core work process. List every step. Now, imagine an alien observer asking, “Why do humans do it this way?” Can’t give a compelling answer beyond “that’s how we’ve always done it”? Design a completely different approach.

Smart Implementation

How to make AI stick in the real world

5. “Buying Copilot ≠ behavior change” – ​Conor Grennan​ (Chief AI Architect of NYU Stern)

Grennan watched NYU faculty get excited about AI tools, then barely use them. The turning point came when they stopped teaching “how to use ChatGPT” and started with “how to change your workflow.”

Instead of feature demos, they redesigned specific faculty routines, such as moving from “write lecture notes” to “create interactive discussions with AI prep.”

What works: AI adoption is change management, not software training.

Reality check: People don’t resist AI tools—they resist changing their habits. The technology works fine, but humans need rewiring.

Adjust: Pick one AI tool you’ve been meaning to use more. Instead of learning more features, redesign one specific daily routine to include it. Change the workflow first, then master the tool.

6. “Promote your internal AI champions—fast” – ​Edmundo Ortega​ (Investor/VC Partner)

Ortega studied a consulting firm where AI adoption stayed stuck at 20% for months. Then they gave their best AI user a promotion, a budget, and permission to train others.

Within three months, usage hit 75%. How? Seeing trusted colleagues share their success stories was the real spark behind this rapid AI adoption.

What works: Culture shifts through peer influence, not executive mandates.

Reality check: Too many “AI champions” (early adopters) get extra work but no additional authority. They become unpaid help desk agents for confused colleagues instead of change leaders.

Adjust: Find your organization’s best AI user. Ask leadership to give them a training budget, presentation opportunities, and measurable goals. Make their wins visible to everyone else. Or, if it is you, ask for the budget, but pitch your ideas too.

7. “Prompt like it’s a person—tone, context, goal” – ​Shiv Singh ​(Co-founder AI Trailblazers)

Singh was reviewing his marketing team’s AI output when he noticed everything sounded generic. Their prompts looked like search queries: “Write an email about a product launch.”

When he taught them to prompt conversationally—”I’m a customer success manager writing to clients who haven’t used our new feature. Be encouraging but not pushy. Help them see the value without overwhelming them”—engagement rates doubled.

What works: Conversational prompting unlocks AI’s ability to match tone and context to specific situations.

Reality check: We often use AI like we’re filling out a form instead of briefing a capable colleague. The difference in output quality is dramatic.

Adjust: Take your last generic AI prompt. Rewrite it including: your role, the recipient, the desired tone, and the specific outcome you want. Compare the results.

Human-AI Balance

How to stay valuable as AI improves

8. “Use AI. But never fully trust it” – ​Dr. Philippa Hardman​

Hardman discovered her students were submitting papers with fabricated citations that looked completely legitimate.

One student’s AI-generated bibliography included 15 studies that didn’t exist, complete with realistic journal names and DOI numbers. The formatting was perfect, but the content was fiction.

What works: AI’s confidence is not correlated with its accuracy. Validation is your competitive advantage.

Reality check: AI lies convincingly. The better it gets at sounding authoritative, the more dangerous blind trust becomes. Sure, it’ll draft a great email, but you might want to check for accidental Shakespeare quotes.

Adjust: Create a verification checklist for any AI output others will see. Include at least three specific checks (sources, logic, factual claims). Use it every time, especially when the results look impressive.

9. “Learning is about the experience, not the output” – ​Mark Daley​ (Academic/Chief AI Officer)

Daley addressed professors worried that AI would make education pointless.

His response: “People still cook even though restaurants exist. They garden even though supermarkets sell vegetables. They make music even though Spotify has millions of songs. Learning isn’t about efficiency—it’s about becoming.”

What works: Skill-building creates value beyond the immediate output, even when AI can produce similar results faster.

Reality check: This assumes people choose effort over ease when they have the option. That’s optimistic about human nature, but maybe not realistic for many.

Adjust: Identify one skill where AI could replace your output. Learn it manually for one month anyway. Document what the learning process teaches you that the AI output doesn’t. (Ha, I can already see you skipping this!)

10. “Redesign your org in 3 steps: Tasks → Skills → Roles” – ​Sania Khan​ (former Chief Economist of Eightfold AI)

Khan was hired when a software company’s AI initiative stalled. The company had jumped straight to hiring “AI specialists” without understanding which tasks AI could actually handle.

After mapping every team’s work, they discovered 40% of engineering tasks could be automated. Still, they needed three new roles they hadn’t planned for: AI prompt designers, output validators, and human-AI workflow coordinators.

What works: Systematic change requires understanding what’s happening before deciding what should happen.

Reality check: This framework works perfectly for leaders with authority. If you’re still testing the waters with AI adoption, apply it to your role first to build credibility.

Adjust: Map your or your team’s weekly tasks in a spreadsheet. Rate each task’s AI automation potential. Identify which skills become more valuable and which become obsolete. Present this analysis to leadership as a strategic opportunity?

The Brutal Question

What would you tell someone who implements all 10 insights and still struggles with AI adoption?

Well, you’re probably solving the wrong problem.

These insights assume you control your context, but what if you work for a company that treats AI like a security threat?

What if you’re in a regulated industry where approval takes six months? What if your role gives you zero authority to change workflows?

The experts operate in environments designed for innovation. Yours might not be.

Sometimes the most innovative AI strategy isn’t optimizing your current situation—it’s escaping it.

If your context blocks strategic AI use, focus on building portable skills instead.

Master evidence-based prompting, validation frameworks, and change management. These transfer when you move somewhere that values innovation.

Document your experiments. Build a portfolio of small wins. Create proof that opens doors to places where these insights can work.

The brutal truth? Expert advice is powerful. But it only works when you have the right environment to apply it.

If you don’t have that environment yet, work on changing your environment first.

Your Strategic Challenge

Pick the insight that makes you most uncomfortable.

That discomfort? It’s pointing toward growth.

Test it for one week. Document what happens—success, failure, or the messy middle.

If it doesn’t work, ask: “Is this insight wrong for my situation, or is it wrong for this insight?”

Then decide: adapt the insight to fit your constraints, or start building toward different constraints?

Both paths are valid. But drifting without choosing gets you nowhere.

And remember, expert insights are starting points. Your context determines what happens next.

Want to explore all 100 expert insights? ​Click here​.

Thanks a lot for reading this far. Happy Friday,

Mark
The AI Learning Guy
👋⚡😎

Sources and books

1. Best AI Books 2025. ​View on Amazon*​.

Note: No single website has all the answers. This list serves as a starting point for those who want to explore or satisfy their curiosity about AI.
Links: Links with * are affiliate links. See disclosure below.

​Affiliate disclosure: To cover the cost of my email software and the time I spend writing these emails, I sometimes link to products. Please assume these links are affiliate links. If you choose to buy through my links, a big THANK YOU – it will make it possible for me to keep doing this.