This article presents my thoughts on AI and the transformation it will bring, as well as its specific impact on business.
During the Industrial Revolution, we underwent a fundamental change that proved machines could out-perform humans in highly mechanised tasks.
Although productivity soared, humans had to adapt, and there were casualties along the way; hand-spinning and many factory jobs vanished almost overnight.
Yet the erosion of labour value arrived with an explosion in demand for goods and services, letting workers redeploy their skills elsewhere. The coming AI wave will not follow the same script.
How Fast Has AI Developed?
Only two years ago I could outperform most large language models (LLMs) at logic, maths and anything geometric. I assumed we’d need comparable leaps in robotics before real labour disruption arrived.
Fast-forward to 2025: vision-language models can parse diagrams, Boston Dynamics robots do parkour, and South Korea has replaced roughly 10 % of its industrial workforce with robots. I expect that share to accelerate—and governments will be caught off-guard by the unemployment spike.
How AI Will Make SaaS Obsolete
Front-end builders such as Durable, Wix AI and Hostinger’s AI suite have already trivialised landing-page design. The real bottleneck is orchestration—testing, deployment, scaling.
Tools like Replit Ghostwriter hint at a near-future marketplace where an agent can act as a full-stack engineer on demand. Why pay a monthly seat fee when an LLM can build the feature on the fly?
Who Actually Wins?
“AI is a once-in-a-lifetime opportunity” — but mainly for those who own the picks and shovels:
| Tier | Why They Win | Examples |
|---|---|---|
| Infrastructure | Sell GPUs, cloud, compilers | NVIDIA, AMD, AWS |
| Foundation-model labs | Control data + training pipeline | OpenAI, Anthropic, Google DeepMind |
| Tooling layers | Orchestration, agents, security | LangChain, Replit, Pinecone |
How AI-First Businesses Could Shrink the Economy
1 · Winner-Take-Most Economics
NVIDIA booked $22.6 billion in Q1 FY25 data-centre sales—up 427 % YoY—and now supplies roughly 80 % of advanced AI GPUs (NVIDIA IR).
A handful of labs (OpenAI, Google, Anthropic, xAI) control >90 % of parameters shipping in production models, and most agent frameworks centre on LangChain or Replit’s stack. When replication costs run into billions, profits pool at the top.
2 · Large-Scale Labour Displacement
- 27 % of jobs in advanced economies face “high risk” from current AI according to a 2024 OECD employer survey.
- The World Economic Forum warns that 40 % of routine coding tasks could vanish by 2040 (WEF).
Fewer pay-packets means lower consumption just as AI pumps out more supply.
3 · Money Piles Up but Stops Moving
U.S. M2 velocity hit 1.39 in Q1 2025, half its 1997 peak (FRED). Cash stashed in corporate treasuries doesn’t buy groceries; velocity drops and deflationary stagnation looms.
4 · Why “Any B2B Is Safe” Is a Myth
- Tiny SaaS features are commoditised by prompt-built agents.
- Mid-tier vendors suffer shrinking margins as clients renegotiate.
- Infrastructure giants keep pricing power because everyone depends on them.
What to Watch Next
- Policy lag: governments react after layoffs spike, not before.
- Agent marketplaces: early signals on Replit and Hugging Face.
- Compute bottlenecks: GPUs today; energy tomorrow.
AI can lift productivity while draining spending power from the majority.
When money pools at the top and stops circulating, record corporate earnings can coexist with an economy that feels — and is — stagnant for most citizens.
Ultimately, we will have to refine our understanding of what it means to be human and what is needed for an economy to function.
Sources
- History.com – Industrial Revolution overview
- International Federation of Robotics – World Robotics 2023 Report
- NVIDIA Investor Relations – Q1 FY25 earnings
- OECD – Workers Most Affected by AI (2024)
- World Economic Forum – AI and Coding Jobs
- Federal Reserve FRED – Velocity of Money (M2V)
