We Have a Tech Surplus and a Trust Deficit
"Real change management isn't training people on how to use the tool; it's convincing them why using it won't hurt them."
— Synthesized from WEF 2026 Panel (Visa, Philips, Accenture)What You Need to Know, Do, and Rethink
Whether you're redesigning a curriculum or transforming a business, the core challenges are the same: evolving your practices while bringing people along for the journey.
Redesigning Education for the Agentic Age
Your students will graduate into a world where entry-level jobs of "doing the task" are evaporating. Your job is to prepare "Orchestrators," not executors.
Rethink Assessment
- Stop Catching Cheaters, Start Assigning the Impossible If an assignment can be solved by ChatGPT in 5 seconds, the assignment is the problem—not the AI. Assign problems that require AI assistance to solve.
- Grade Process and Judgment, Not Just Output Ask students to submit their "conversation log" with AI. How did they prompt? How did they verify? How did they iterate? That's the skill.
- Teach "Evals," Not Just Generation Give students broken, hallucinating AI code and ask them to create the evaluation metric to catch the error. This mirrors the actual job market.
Evolve Technical Literacy
- "Vibe Coding" is Valid Pedagogy Stop failing students for not knowing syntax. Start grading them on their ability to debug AI output. The skill is describing the problem, not writing the loop.
- Teach Data Curation, Not Just Analysis Cleaning and selecting data ("Golden Datasets") is more important than the algorithm itself. AI is useless without data governance.
- Model Vulnerability Live in Class Bring a problem you don't know how to solve. Open Claude on the projector. Show students how you talk to the bot, spot hallucinations, and iterate.
Navigate Institutional Change
- Fight for Tool Flexibility One-size-fits-all mandates (e.g., "everyone uses Copilot") fail. Advocate for a "Traffic Light" system: Red data (FERPA/Health), Yellow data (Coursework), Green data (Public research).
- Trust in Faculty Judgment Problems have known desired outcomes but unknown solutions. We hire professors because they navigate this ambiguity. Let them choose the tools for their discipline.
- Cease the "Detection" Arms Race It's a losing battle that erodes trust. Redirect that energy toward reimagining assignments that AI makes more valuable, not obsolete.
Transforming Organizations in the Agentic Age
If you use AI only to cut costs, you race to the bottom. If you use AI to do things that were previously impossible, you create new value.
Rethink Strategy
- Start with Problems, Not Solutions Companies asking "How can we use GenAI?" are starting with the solution. Winners ask "How can we reduce customer waiting time?" and happen to use GenAI.
- Focus on What Won't Change Bezos's secret: customers will always want lower prices, faster delivery, better selection. Build your AI strategy around these immutable needs, not the latest model.
- Re-Imagine, Don't Just Automate "Pilot Purgatory" happens when you pave the cow path. Philips moved from "selling devices" to "selling health monitoring." That's the hard creative work.
Build the Foundation
- Data Silos Kill AI Agents If Sales lists "Company Inc." and Billing lists "Company, LLC," the AI fails. The "sexy" model is useless without the "boring" data cleanup.
- Curate "Golden Datasets" LinkedIn tried pointing AI at all their Google Docs—it failed miserably. Success came from manually curating 50 perfect product specs. Quality over quantity.
- Don't Outsource Your AI Strategy Agencies don't know your business. You or a "nerd" on your team must do the grunt work of reading AI outputs and correcting them for 30 days. Outsource training, outsource quality.
Lead the Change
- Leader-Led Learning is Non-Negotiable Visa's breakthrough came when they locked 300 executives in a room for 2 days and forced them to build agents. If leaders don't know how it works, troops won't use it.
- Spend Equal on Adoption and Technology If you spend $1M on GPUs, spend $1M training staff to trust and use the output. The mistake: 50 engineers to build a tool, 1 change manager to send an email about it.
- "Human in the Lead," Not "Human in the Loop" "Loop" implies AI is driving and human is a brake. "Lead" means human sets the destination, AI is the engine. This psychological framing is everything.
The AI Career Quadrant
Four competencies emerged from our analysis of 19+ expert conversations. This framework applies whether you're designing a curriculum or restructuring a team.
The Compass
Focus on immutable customer needs, not trendy technologies. Solutions are temporary; customer problems are permanent. Fall in love with problems, not tools.
The Orchestrator
Shift from "doing the work" to "judging the work." Your taste, ethics, and critical thinking define output quality. You're paid to decide, not to type.
The Foundation
AI is useless without clean data and human trust. If your data is siloed or your employees don't trust the system, the smartest agent fails.
The Toolkit
The specific tool matters less than the meta-skill of learning. Be the "Willow Tree": deep roots in logic, flexible branches in tooling.
Mindset Shifts for Leaders and Educators
In an AI-first world, the half-life of a skill is measured in months. These mindsets will outlast any software—and they're what you need to model for your students and teams.
From Control to Enablement
Organizations that treat users as "liabilities to be managed" lose their best talent. The winning organizations treat them as "talent to be unleashed."
From Efficiency to Expansion
If you only use AI to cut costs, you can only cut to zero. If you use AI to do things previously impossible, there's no ceiling on value creation.
From Teaching to Learning
You are no longer the source of knowledge; you are the guide to discernment. Model lifelong learning by openly using AI to solve problems you don't know how to solve.
Use AI to Make Work Harder, Not Easier
Most people use AI to make mediocre work faster—a race to the bottom. Instead, use AI to challenge you: "Where is my argument weak? Give me direct pushback."
The "Barbell Effect" of Intelligence
AI is great at the low end (emails) and unexpectedly great at the high end (impossible optimization problems). It fails in the "mediocre middle." Push to the extremes.
39% of Skills May Be Obsolete in 5 Years
World Economic Forum data. The tools you teach today will change in 6 months. Focus on meta-skills: judgment, systems thinking, adaptability, and clear communication.
Ethics and Responsible AI Integration
If the AI provides the answer, the human provides the Trust. In an agentic world, one employee lacking integrity won't just make one bad decision—they'll program an agent to make 10,000 bad decisions per minute.
The New Ethical Imperatives
Technology acts as a multiplier for existing forces—including greedy or inequitable systems. We need humans in the lead setting the moral compass, not just in the loop watching the machine.
Teach AI Civics, Not Just AI Skills
Just as we teach students not to litter, teach them: "Just because you can scrape that data, should you?" If an AI denies a loan, a human must be accountable. We cannot outsource morality to a server.
Reliability is Now "Safety Critical"
In an agentic world, students aren't just doing work—they're supervising work at scale. Integrity, consistency, and ethical judgment become non-negotiable baseline requirements.
Design for Diverse Contexts
Delight can backfire. A delivery app sent "Missed call from Mom" as a Mother's Day promo—traumatic for users who had lost their mothers. Validate "helpful" AI features against diverse experiences.
Cultivate Deep Skepticism
The "Talking Dog" theory: treat AI with awe (it's a miracle!) but don't trust it for financial advice just because it talks. This skepticism is the new definition of critical thinking.
Leadership, Culture, and Change Management
Conway's Law is real: you can't change the output without changing the team structure. Technology adoption is organizational change—and that's the 90% most people ignore.
Structure Dictates Technology
Block had to reorganize from "General Manager" silos to a "Functional" structure to succeed with AI. If your org chart is broken, your AI will be broken.
Trust Precedes Adoption
If leadership uses AI to fire people, they destroy trust. If they use it to remove grunt work so people can do more interesting things, they build it. Choose wisely.
Start Small, Stay Scrappy
Block's massive AI tool "Goose" started with one engineer building something cool on the side—not a million-dollar mandate. Encourage your "mad scientists" to tinker.
Everyone is Now a Manager
Even entry-level employees will manage 5-10 AI agents. The skills of management—clear instructions, defining outcomes, quality control—are now entry-level requirements.
Audit the "Miracle Step"
Seth Godin: If your strategy has a step that says "...and then the AI makes it work," you're failing. Map the exact workflow: Ingestion → Orchestration → Training. No magic.
Hire for Taste, Not Just Skill
Appoint a "Benevolent Dictator" for quality—someone with impeccable product taste who decides if AI output is good enough. "Vibes" don't scale; rigorous human taste does.
Your Monday Morning Checklist
-
1
Identify One "Impossible Problem"
Find a problem in your curriculum or business that has stumped you. Use AI to work through it—and document your process as a teaching moment.
-
2
Audit Your Data Foundation
Before any AI initiative, ask: "Is our data clean, connected, and semantically tagged?" If not, start there—not with the shiny model.
-
3
Schedule "Hands-on-Keyboard" Time
If you're a leader, block 2 hours to build something with AI yourself. If leaders don't know how it works, adoption stalls.
-
4
Reframe One Assessment or Process
Take one assignment or workflow that AI makes trivial. Redesign it to require AI—and grade the judgment, not the output.
-
5
Have the Trust Conversation
Ask your team or students: "What are you afraid AI will be used for?" Address the fear directly. Trust is the prerequisite for adoption.
-
6
Focus on One Immutable Need
Ask: "What do our customers/students need that will still be true in 10 years?" Build your AI strategy around that, not the latest tool.
Expert Sources
This guide synthesizes insights from 19+ hours of conversations with industry leaders across tech, consulting, enterprise, and academia.
Leadership & Change Management
- Scaling AI: Now Comes the Hard Part WEF 2026 Panel — Visa, Philips, Accenture, Aramco
- From Managing People to Managing AI Julie Zhuo, Former VP of Design at Facebook
- The Agentic Organization Sandra Durth, Partner at McKinsey
- Replacing Sales Teams with AI Agents Jason Lemkin, Founder of SaaStr
AI Strategy & Implementation
- How Block is Becoming AI-Native Dhanji R. Prasanna, CTO of Block
- The Full Stack Builder at LinkedIn Tomer Cohen, CPO of LinkedIn
- AI Data Foundation (Don't Fail with Agentic AI) Christian Kleinerman, EVP of Snowflake
- AI as a Scientific Tool Demis Hassabis, CEO of Google DeepMind
Skills & The Expert Economy
- The Expert Economy & Scale AI Jason Droege, CEO of Scale AI
- AI Evals: The Hottest New Skill Hamel Husain & Shreya Shankar
- Experts as the New Gold Mine Brendan Foody, CEO of Mercor
- Building AI Skills IBM Technology
Strategy & Mindset
- Things That Won't Change Jeff Bezos at Italian Tech Week
- Finding the Problem is Hard Kevin Systrom, Co-founder of Instagram
- The "Talking Dog" Theory of AI Seth Godin
- AI Code Generation: Wins, Fails, and the Future Gabe Goodhart, IBM Chief Architect
The Technology is Here.
The Hard Part is Human.
The institutions that treat AI as a "software upgrade" will fail. The ones that treat it as a "cultural upgrade"—re-skilling, re-trusting, re-imagining—will win.
"The more artificial the intelligence becomes, the more premium the humanity becomes."