1. A general-purpose technology
The right analogy isn't "new software." It's electricity. The phrase "AI is the new electricity" is from Andrew Ng (2017), and it's on target: electricity didn't improve one industry, it transformed all of them, because it plugged into everything.
Generative AI has that same "horizontal" character: the same tool writes an email, analyzes a sales spreadsheet, writes code, translates a contract. It's not one profession changing — it's all professions gaining a new layer.
- Steam enginelate 18th c.
- Electricitylate 19th c.
- Computing & internetlate 20th c.
- Artificial intelligencetoday
2. Signals that this isn't hype
Examples current as of May 2026 — the names and figures below will date; it's the logic that matters, not the specifics.
The "revolution" narrative is easy. What doesn't lie is where people and money go:
- Jeff Bezos came out of semi-retirement to co-found an AI company (Project Prometheus, several billion raised). When Amazon's founder gets back in the game, it's not for a trend.
- Andrej Karpathy, one of the biggest names in AI, left his other projects to join Anthropic. And he's not alone: several CTOs of billion-dollar companies accepted going back to being individual researchers to work on these models.
- The numbers: tech giants are committing hundreds of billions per year into AI infrastructure. You don't bet those sums on a passing trend.
The reading rule: when those with the most to lose are reallocating their time and capital to something, that something is serious.
3. What it concretely changes for you
For a business owner or retailer, the change isn't abstract:
- Work shifts from execution to supervision. You describe what you want and verify, instead of doing everything by hand. Your time goes to judgment, not raw production.
- A non-technical person can do technical things. I've shipped inventory management modules and advanced SEO without being a developer. The "you need to know how to code" barrier partly falls.
- Speed changes scale. Projects that used to take weeks get done in hours. That shifts the constraint: the bottleneck is no longer execution, it's knowing what to ask.
You do everything
- Find the info
- Write / produce
- Format it
- Fix, redo
You direct & verify
- Describe the need
- Let AI execute
- Check, adjust
- Validate & decide
4. What it does NOT change (and the pitfalls)
To stay honest:
- It doesn't replace judgment. AI proposes, you decide. It can be confidently wrong (see the guide on LLMs).
- Value shifts, it doesn't disappear. It moves toward those who know how to frame the right problem and verify the result — not toward those who delegate with blind trust.
- The advantage goes to those who start early. Not in theory: by practicing, on real cases.