A combination of three threads in my career gives me a unique lens on this question.
Productivity Research
I build Excel and Excel Agent at Microsoft. Office products have been the backbone of productivity across industries, functions, and profiles. Understanding how people are productive at work, and how AI can enhance it, has been a core research objective for the last two years.
Career Building
As prep head at IIM Calcutta, I mentored 1,000+ students across three batches to find the right career fit and break into their target industries. I continue to mentor students and professionals through career coaching and corporate learning programs.
AI Coaching
Coached 100+ product managers in India through career transitions. Now coaching 100+ trainers building AI PM programs across the US. This gives a practitioner's lens on how AI is reshaping career paths.
Productivity research + career building + AI coaching = a unique point of view on how AI changes careers.
Claude Cowork launch wiped $1T from software stocks in one week. Two AI crashes in 13 months. TCS, Infosys hit.
Uncertainty
Outcomes become impossible to forecast.
Hinton says 5 years to AGI. LeCun says decades. OpenAI bets $74B it's sooner. The creators can't agree.
Complexity
Integrated systems create cascading effects.
246K tech layoffs in 2025. S&P 500 up 18%. Same companies. Same year. India IT: 25K fired + 35K AI hires at the same firms.
Ambiguity
Cause and effect become unclear.
Individual workers report 40% productivity gains (McKinsey). Economy-wide data shows near zero (Goldman, NBER). Same tool, two scales, opposite answers. Is AI creating value or just reshuffling it?
"The pace of change has never been this fast, yet it will never be this slow again."
Justin Trudeau, Davos 2018 · S&P 500 avg company lifespan: 67 years → 15 years(Innosight / McKinsey)
"Self-claiming some AGI milestone is just nonsensical benchmark hacking to me. The real benchmark is the world growing at 10%."
For 250 years, technology (A) amplified labor (L). AI is the first technology that substitutes for it.
EVERY WAVE CAUSED PANIC
1589 · Royal Court
"Consider thou what the invention could do to my poor subjects. It would assuredly bring to them ruin by depriving them of employment."
Queen Elizabeth I, refusing the stocking frame patent
1928 · The New York Times
"MARCH OF THE MACHINE MAKES IDLE HANDS"
"More and more the finger of suspicion points to the machine... machines had come into conflict with men."
1961 · TIME Magazine
"THE AUTOMATION JOBLESS: Not Fired, Just Not Hired"
"Each week, some 35,000 U.S. workers lose their jobs because of automation." US unemployment: 6.9%. By 1968: 3.4%.
1933 · The New York Times
"EINSTEIN TRACES SLUMP TO MACHINE"
Albert Einstein blamed the Great Depression on technological displacement. 25% of jobless Americans agreed.
1995 · Newsweek
"The Internet? Bah! Why Cyberspace Isn't, and Will Never Be, Nirvana"
"No online database will replace your daily newspaper." Author later: "Of my many mistakes, few have been as public."
2026 · Right Now
"AI Will Take All Our Jobs"
Sound familiar? The same fear, the same headlines, for over 400 years. Is this time different?
Each time, new jobs emerged that nobody predicted: software engineers, data analysts, UX designers.
The Lump of Labor Fallacy
The mistaken belief that there is a fixed amount of work in the economy. Economists have argued for over a century that technology always creates more jobs than it destroys. Is this time different?
Technology makes it easy for everyone to do a particular job. Once it is democratized, we all abstract a level higher to do a different kind of work.
Yesterday
Typing
Very few people were professional typists. Today, everyone types on their phone. You don't need someone else to type for you anymore.
Photography
Professionals operated expensive equipment. Today, everyone is a photographer. Pros shifted to art direction, storytelling, brand.
Today
Engineering
AI is democratizing software. Anyone can build an app or a product. The cost of building is approaching zero.
The New Expectation
Delivery is no longer the bottleneck. You need strategic vision, product thinking, understanding of competitive landscape, and a vision for where the product should go.
"The skill isn't building anymore. The skill is knowing what to build and why."
Replaced manual labor. Machines did the heavy lifting.
Each revolution let humans climb to higher-order work. AI targets the highest order.
U.S. Employment Share
Data spans 1800 – 2024 · Bureau of Labor Statistics
Agriculture
70%(1800)→1.6%(2024)
Manufacturing
33%(1950)→8%(2024)
Services
15%(1800)→84%(2024)
???
What's next?
ACT II
IS THIS TIME DIFFERENT?
If you don't need intellectual labor anymore...
What's left?
Every technology revolution has created new jobs to replace the ones it destroyed. Economists call the belief that there's only so much work to go around the "lump of labor fallacy." But this time, the debate is genuinely open.
Sam Altman
CEO, OpenAI
"AI will create new kinds of jobs we can't even imagine. The economy will grow. Everything gets cheaper."
From 49% to 81% in 18 months. Software is ground zero for AI capability.
REAL-WORLD SIGNAL
41% of all code written in 2025 was AI-generated. 46% of GitHub Copilot output gets accepted. Google + Microsoft: ~25% AI-assisted code each. 95% of engineers now use AI tools weekly.
These aren't predictions. These are reports from the field.
Andrej Karpathy
@karpathy
"I've never felt this much behind as a programmer. The profession is being dramatically refactored... some powerful alien tool was handed around... the resulting magnitude 9 earthquake is rocking the profession."
"In the last 30 days, I landed 259 PRs, 497 commits, 40k lines added. Every single line was written by Claude Code. I shipped 22 PRs yesterday and 27 the day before, each one 100% written by Claude."
A product manager controls risk while enabling innovation. The biggest risk was always cost of production: 10 ideas, limited engineers, 3-5 month estimates, ruthless prioritization.
"At Google, we are moving from a writing-first culture to a building-first one. Writing was a proxy for clear thinking, optimized for scarce eng resources and long dev cycles. Now, when time to vibe-code prototype ≈ time to write PRD, PMs can SHOW not tell."
The cost of producing the asset is near zero. So the bottleneck has shifted: from building to distributing.
Blue Ocean
Uncontested market space where demand is created, not fought over. No competition yet.
Red Ocean
Existing market with defined boundaries, accepted rules, and companies fighting over shrinking demand.
AI turns every blue ocean red before you can finish swimming. When building costs nothing, barriers to entry collapse, and every market floods with competitors instantly.
"The new job for everyone has become marketing."
We started by saying software is ground zero. AI has completely disrupted ways of working and the means of production inside the software industry.
But make no mistake: any role where 100% of work is digital is already in the disruption zone. Financial analysts, content writers, designers, legal researchers. If your entire workflow exists as data, AI can already do your job.
For industries that involve the physical world, two barriers remain unsolved:
MR
Mike Rowe
Host, Dirty Jobs · mikeroweWORKS
"We've been telling kids for 15 years to learn to code."
"Well, AI is coming for the coders."
"It's not coming for the welders, the plumbers, the steamfitters, the pipefitters, the HVAC, or the electricians."
"In Aspen, I sat and listened to Larry Fink say we need 500,000 electricians in the next couple of years."
Organizations have decades of institutional knowledge, workflows, and decision-making patterns locked in people's heads. Collecting this context for AI is the #1 challenge.
Physical World
AI needs to sense and act in the real world. That's where robotics comes in.
Tesla Optimus
Factory floors
Unitree
Agile humanoids
Boston Dynamics
Advanced mobility
Solving these barriers is the biggest market opportunity of the next decade.
The same pattern from software is repeating everywhere: costs collapse, roles merge, fewer people produce more.
LEGAL
Harvey AI
$8B valuation. 50 of top 100 law firms. Lawyers save 4 hrs/week, worth $100K in billable time per lawyer annually.
The pressure flows downward: investors, boards, CEOs. The opportunity flows to whoever figures out how to diffuse AI across the organization.
: For You
FOR ORGANIZATIONS
Bottom-Up Learning Wins
Top-down AI training fails because every function has unique organizational context that no external consultant can replicate.
The most effective approach: give employees dedicated time, AI compute access, and aligned incentives to build agentic workflows that improve their own work.
78%
of AI users bring own tools without approval
95%
of gen AI pilots fail to scale beyond experiments
Manage Human + AI Teams
Managers must be empowered to build systems where they orchestrate both AI agents and human employees.
Org charts will evolve from hierarchies to agentic networks, where work is exchanged through tasks and outcomes, not reporting lines.
HBR: "Agent Manager"
A permanent new role. Within 12-18 months, a standard title in AI-first enterprises.
Embed First, Optimize Later
We have decades of process design that tells us why roles exist: who builds, who reviews, who approves. These touchpoints were created with intent.
Step 1: Embed AI into each existing touchpoint. Leverage proven process wisdom.
Step 2: Over time, identify opportunities to make workflows truly AI-native.
"AI is 20% algorithms, 80% organizational rewiring."
Lead First
Leaders must model AI use themselves. Vacate time, experiment openly, set the cultural tone from the top.
Allow Open Exploration
Early tasks may seem trivial. Comfort with the tool precedes solving high-value problems.
Grow Internal AI Explorers
Internal practitioners with org context outperform external consultants. They spot deployment opportunities and diffuse knowledge organically.
High-Friction First
Start where AI removes daily pain points. Then close the loop on end-to-end workflows for real productivity gains.
Build Knowledge Networks
Create mechanisms for teams to share AI context, patterns, and wins across the organization.
Build Evaluation Muscle
Enable employees to create agent evaluation mechanisms. Measure performance and quality of agentic workflows systematically.
Default to Agentic-First
Cultivate the instinct to ask "Can an agent do this?" before "Who should I assign this to?"
Redefine the Manager Role
Agent managers don't just delegate. They orchestrate workflows, set guardrails, and quality-check agent outputs.
Rethink Team Boundaries
Agents work across functions. Org structures must evolve from silos to fluid, task-based networks.
Unblock Individuals First
Ensure every employee can augment their own work before optimizing cross-team workflows.
Then Solve Cross-Team Friction
Build new playbooks for interconnected human + AI teams to move faster together.
Design AI-Native Workflows
Don't just layer AI onto old processes. Rethink workflows where humans and agents each play to their strengths.
Iterate in Production
Perfection delays adoption. Ship embedded AI early, gather feedback, refine. Speed of learning beats quality of planning.
88% of organizations use AI. Only a third have scaled it. The gap is not technology. It's management.
Like organizational behavior gave us frameworks for managing humans, you need deep understanding of how AI agents work, what they're good at, and where they fail.
Diagnose failures. Design guardrails. Know when to trust, when to verify.
Systems Thinking
You're no longer doing the work. You're managing interconnected systems that span functions and disciplines. Seeing how components interact matters more than mastering any one.
Vision Over Execution
Execution is commoditized. Any agent can execute. The value lies in how superior and well-thought-out your vision is compared to someone else's.
KNOW THE LIMITS
Karpathy's 6 Cognitive Deficits of LLMs
"LLMs are fallible people-spirits. Their output must be verified, not trusted outright."
1
Hallucinations — They make things up. "Hallucination is all LLMs do. They are dream machines."
2
Jagged Intelligence — Solve complex math, then claim 9.11 > 9.9. Capabilities are deeply uneven.
3
Poor Self-Knowledge — Cannot assess what they know vs. don't know. Confident on topics where they should express uncertainty.
4
Anterograde Amnesia — "A coworker who forgets everything after each conversation." No long-term memory consolidation.
5
Gullibility — Susceptible to prompt injection. Can be tricked into leaking data or giving harmful responses.
6
Spinning — Vague prompts cause repetitive correction loops. Gets stuck rather than making progress.
"AI will not deliver value because firms spend money on tools. It will deliver value when leaders develop the competencies to transform organizations."
Harvard Business Review
Not the model. Three things matter:
Compute
Raw processing power at scale
Context
Proprietary data and domain knowledge
Distribution
Reaching users where they already are
ACT V
YOUR MOVE
Instruct AI with precision. Clarity is the new management skill.
Measure AI output quality. If you can't evaluate, you can't improve.
Andrej Karpathy @karpathy
Ex-Tesla AI, Ex-OpenAI
"The hottest new programming language is English."
poly·math/ˈpɒlɪmaθ/— person of wide knowledge across multiple fields.
Leonardo da Vinci: painter, engineer, anatomist, architect, inventor. The original one-person team.
→ AI is making this achievable for everyone.
WHAT WE'RE SEEING
Roles Are Fusing
Specialists built the old world. AI collapses boundaries between disciplines. A single person now has the toolkit of an entire team: design, code, market, analyze.
Abstraction Upward
Entry-level execution is being filled by agent workforces. Humans move upward to strategy, judgment, and orchestration. The floor rises.
The Gap Narrows
As agents handle more of the value chain, the distance between capital owners and the working individual (labor) shrinks dramatically. Everyone becomes closer and closer to becoming a capital owner itself.
THE OPPORTUNITY
The One-Person Billion-Dollar Company
A single founder with AI agents can ideate, build, ship, market, and scale. The generalist who understands the whole system wins.
Sam Altman @sama
CEO, OpenAI
"...there's this betting pool for the first year that there is a one-person billion-dollar company."
"I skate to where the puck is going to be, not where it has been."
Wayne Gretzky
We can't optimize for where we are today. We have to prepare for where AI will take us.
AND THAT BRINGS US TO THE ABUNDANCE ERA
When AI reaches L4-L5, it doesn't just execute. It innovates. Problems unsolved for decades become tractable. Technology compounds so fast that our ability to address humanity's biggest challenges explodes.
Sam Altman @sama
CEO, OpenAI
"In the 2030s, intelligence and energy are going to become wildly abundant. The ability for one person to get much more done will be a striking change."
"We won't experience 100 years of progress in the 21st century. It will be more like 20,000 years of progress."
Ray Kurzweil, Law of Accelerating Returns
Each of you will have the opportunity to find a unique value proposition, build your own AI organization, and solve problems that matter. The problem space is no longer closed. It's wide open.
AI Budget
Allocate personal time and money to AI tools. Treat it as an investment, not an expense.
Build AI Fluency
Use AI daily. Push its limits. Break it. Learn what it can and cannot do.
Think in Workflows
Decompose your job into steps. Identify which steps an agent can own entirely.
Become a Generalist
The age of deep specialization is ending. Learn to orchestrate across domains.
"So let me ask again... do you feel anxious?"
Every previous revolution replaced muscle. This one replaces mind. We have never been here before. But if you have taste, instinct, and the willingness to explore, this is the greatest era to be alive.