RESHUFFLE An interactive companion to the book
← All ideas
/10 ·Chapter 5 ·~6 min

The Sommelier

Why reskilling is a losing game in a system that's already changed

Idea /10 — The sommelier
A wine bottle · a phone · a paradox Pick first. Then scroll.
$80 in store · $400 in restaurant Why?
Info commoditized · phones know everything Sommelier should be extinct
Sommelier income · up The opposite of extinction
Three kinds of constraint Scarcity · risk · coordination
Where the value moved From info → confidence
The new bundle Curation · storytelling · taste
The reframe Don't reskill — rebundle
WINE $80 $400 ▍ AT THE STORE $80 JUST THE BOTTLE AT THE RESTAURANT ▍ $400 SAME BOTTLE · 5× THE PRICE ▍ OLD · SCARCITY · BROKEN INFO · ON EVERY PHONE ▍ NEW · RISK DECISION ANXIETY ▍ NEW · COORDINATION PAIRING + STORY ▍ NEW · AUTHORITY CURATION + TASTE ▍ THE SOMMELIER · 2024 REBUNDLED, NOT RESKILLED VALUE MOVED. THE SOMMELIER MOVED WITH IT.
▍ Quick prediction

Every phone has all the wine information in the world. So why does the sommelier still command a skill premium?

Pick one. Then scroll.

Idea /10 · The sommelier

A bottle costs $80 at the store. The restaurant charges $400.

It still sells. People go back for more. The sommelier tells them a story about the vineyard — sixth-generation family, mineral soil, a certain slope. Some of it might even be true.

It doesn't matter. The wine tastes better now.

Scroll

This was already odd before AI. Anyone with a smartphone could pull up the same tasting notes the sommelier had memorized. Vivino, Wine Spectator, GPT-4: they all know more about wine than any single human ever could.

By the logic of "information will get commoditized, so jobs that sell information will disappear" — the sommelier should have been extinct by 2015.

So why isn't the role gone?

Scroll

The opposite happened. Sommelier income is higher in 2024 than it was in 2014. Master Sommelier certifications have multiplied. Restaurants pay six figures for top wine directors.

This is the puzzle. And solving it is the entire playbook for thinking about reskilling in the AI era.

The sommelier didn't survive by knowing more wine facts than the phone. The sommelier survived by not selling wine facts in the first place.

Scroll

Sangeet's framework: every job in a system exists to resolve a constraint. There are three kinds.

Scarcity — something essential is hard to access. The 20th-century sommelier sold this: wine info was scarce. Risk — actions carry consequences, and someone must own them. Coordination — many parts must align for the system to work.

When a constraint disappears — scarcity, in this case — the role that managed it loses value. Unless the role can migrate to a different constraint.

Scroll

That's exactly what sommeliers did, without anyone announcing it.

The scarcity-based premium evaporated. But two new constraints rose to take its place. Risk — the diner sweating over a $400 wine list, terrified of picking the wrong one in front of clients. Coordination — pairing the wine with the food, the table, the occasion, the price comfort of the host.

The wine information stopped being the product. The confidence in a moment of uncertainty became the product.

Scroll

And the sommelier rebundled their entire role around these new constraints. Curation: a 50-bottle list of pre-vetted choices instead of 5,000. Storytelling: every bottle now arrives with a 90-second narrative. Authority: institutions like the Court of Master Sommeliers manufactured a credential whose value is *exactly* its scarcity, restoring a different kind of premium.

This is the move every knowledge worker has to make: identify the constraint your role used to resolve, notice when it disappears, and migrate to a new one.

Scroll
Value isn't in giving people more information. It's in giving them confidence in a moment of uncertainty.

Reskilling — learning new tasks AI can't yet do — is a treadmill. The flywheel keeps catching up.

Rebundling — identifying the new constraint your system has and migrating your role around it — is the actual move. The sommelier saw it 30 years ago. Most knowledge workers haven't seen it yet.

Sangeet on this in Chapter 5 ↗

Scroll
▍ Read another

More from the series

One new idea, first weekend of every month. The next few you haven't seen:

▍ All caught up

You've read every idea in the series so far. New ones drop the first weekend of every month.

/00 · ~10 min Read

Reshuffle — The Companion

Who wins when AI restacks the knowledge economy

/01 · ~5 min Read

The Intelligence Distraction

Why "how smart is AI?" is the wrong question

/02 · ~6 min Read

The Map Redraws Power

Every map that describes reality is also reshaping it

/03 · ~6 min Read

Designing for Indecision

AI changes what humans choose by changing what they're asked

/04 · ~5 min Read

Unintelligent AI Matters

Even dumb AI restructures systems

/05 · ~6 min Read

The Tool Integration Trap

Why buying 17 AI tools is worse than buying none

/06 · ~7 min Read

Why Incumbents Always Lose the Reshuffle

Kodak, Blockbuster, Sears — and the AI version playing out now

/07 · ~6 min Read

The Aggregator Playbook

Google, Facebook, Amazon — and the AI version playing out now

/08 · ~6 min Read

The New Chokepoints

Where the power actually lives in the AI stack

/09 · ~6 min Read

The Skill Premium Collapse

Why your expertise stopped paying — and keeps stopping

/11 · ~6 min Read

Coordination Beats Talent

The Galácticos paradox — why structure is the new advantage

/12 · ~5 min Read

The $125 Million Coordination Bug

Mars Climate Orbiter, 1999 — and every AI rollout, 2026

/13 · ~6 min Read

The Building Blocks Economy

MrBeast launched 300 restaurants in a day. He owned one block.

/14 · ~5 min Read

Algorithmic Awareness

Michael Smith made $10M streaming AI music to his bot accounts

/15 · ~7 min Read

The Five Levers of Power

How the British Empire (and Walmart) controlled what they didn't own

/16 · ~7 min Read

Where to Play, How to Win

Chegg's collapse, Singapore's bet, and the book's closing keystone

/17 · ~6 min Read

First, Second, Third Order

Most companies stop at the first-order win. The wealth moves later.