0 / 6 done
Lesson 1 of 6

Useful AI Has Arrived

For two years AI was a clever party trick. Jensen Huang's headline is that the trick just became a worker — and that changes the business math entirely.

⏱️ ~6 min read🎬 GTC Taipei 2026

Jensen opens the keynote with a claim that sounds simple but is actually the whole story: AI has crossed from "interesting" to "useful." It's not just answering questions anymore — it's doing work.

"Today we can say that agentic AI has arrived, that useful AI has arrived."

— Jensen Huang, NVIDIA GTC Taipei 2026

From generative to agentic

Quick vocabulary, because the whole keynote rides on it:

  • Generative AI — makes content: a paragraph, an image, some code. You ask, it produces. (This is the chatbot era.)
  • Agentic AI — does work: it can observe a situation, reason about it, make a plan, and take actions using tools (a browser, a spreadsheet, a database). You give it a goal; it gets it done.

That jump — from "writes me a draft" to "handles the task" — is why Jensen keeps saying useful. An AI that can use tools and finish jobs is an AI a business will actually pay for.

Why business suddenly cares: tokens = money

Here's the line that made every CEO sit up. A "token" is just the unit of AI output (roughly, a chunk of a word). Jensen reframes it as a unit of production:

"Tokens are now profitable units of revenue… AI is now a profit generator. AI is now a GDP generator."

— Jensen Huang

Translation: once AI output became both cheap to produce and genuinely useful, every token an AI generates can make money. That flipped AI from a research cost into a revenue engine — which is why demand for computing power "skyrocketed."

The productivity multiplier

Jensen's favorite proof is software developers. The world's ~30 million developers earn roughly $3 trillion in salaries — and with AI coding tools, their output has multiplied:

"It's effectively $9 trillion of productivity from $3 trillion of salaries."

— Jensen Huang

Same people, same pay, roughly three times the output. That ratio — more value per dollar of labor — is the entire economic case for AI in one number.

MIS in one line: This is the classic "business value of IT" question your field was built on. A new technology only matters if it changes the value equation — and a 3× output story is exactly that kind of change.

So… does AI kill jobs?

Jensen's answer is blunt:

"People talk about AI reducing jobs — complete nonsense. It's causing more software engineers to be hired."

— Jensen Huang

His logic: if one engineer can now generate far more value, companies want more of them, not fewer. When something becomes more productive and there's hunger for more of it, you hire up.

Fair to note: Jensen sells the picks and shovels, so he's optimistic by nature, and not every job behaves like software (where demand seems almost bottomless). Reasonable economists debate which roles get augmented, which get automated, and how fast. The honest takeaway isn't "no jobs change" — it's "the people who learn to work with agents are the ones who win."

"Does IT actually create value?" has haunted MIS for decades (look up the productivity paradox). Jensen is making a strong claim that, this time, the value shows up clearly and fast. Your job as an MIS pro is to be the one who can actually measure it — output per worker, cost per task, revenue per token — instead of taking hype or fear at face value. You're also the person who frames AI as augmentation (making people more productive) versus blunt automation (replacing them), which leads to very different strategies.

Quick check

When Jensen says "useful AI has arrived," what's the core business shift he means?

The shift is from generating content to doing work. Once AI completes tasks using tools, each token of output can create real economic value — that's why he calls tokens "profitable units of revenue."
🧠 Think like an MIS analyst

"AI gives 3× productivity — so cut headcount 50%"

Your CEO reads the "$9 trillion from $3 trillion" stat and announces a plan to cut the team in half. She asks you to make it happen. What do you advise?

You slow the leap from "3× output" to "half the people." Jensen's own point is that higher productivity led companies to hire more, because the extra output was worth selling. The real question is demand: can we profitably use 3× more output? If yes, you grow, not shrink. You'd also ask which specific tasks AI actually accelerates (some, not all), pilot it on those, measure the real lift, and redeploy people to higher-value work. Cutting blindly can leave you unable to capture the upside — and unable to handle the cases where AI still needs a human. Outcome over hype.

✍️ Your turn

Think of a job or task you know well. What's one part an AI agent could genuinely do (not just draft) — and one part that still clearly needs a human?

Do you find Jensen's "AI creates more jobs" argument convincing? Where might it hold, and where might it not?

Key takeaway: "Useful AI" means AI that does work, not just generates content. That turns AI output (tokens) into a unit of production — a real profit and productivity engine. The business question isn't "will AI replace people?" but "how do we capture the new value, and who learns to work with agents?"
↑ UpModule home Next →2 · Meet the AI Agent