Reshuffle∗

The true power of AI lies in coordination.

AI is misread as a tool of automation - a faster way to do the work humans already do.

The real story is different. AI's economic impact comes not from how it performs a task, but from how it restructures the system around the task - letting domains, teams, and tools coordinate where they couldn't before.

The last time a new coordination layer rewired the world economy, the technology looked deceptively simple. A steel box.

What everyone read as automation.

In April 1956, Malcolm McLean loaded 58 standardized steel boxes onto a converted WWII tanker - the Ideal-X - and sent it from Newark to Houston.

Before then, shipping was break-bulk. Armies of dockworkers manually hauled barrels, sacks, and crates onto ships, one piece at a time. Cargo sat idle for days. Sometimes weeks.

The obvious read on McLean's box was an automation story: faster cranes. Bigger ships. Fewer dockworkers. A hardware upgrade for the docks.

True, but this missed the real impact that the container would unfold!

The real unlock wasn't speed. It was coordination.

The real impact of the container played out once every mode of transport - ships, trains, trucks - agreed on the same dimensions of the container, unlocking intermodal logistics.

A box loaded onto a truck in Detroit could land on a flatcar in Chicago, slide into a ship's cell in Long Beach, and lift onto a chassis in Tokyo. The standardized box enabled coordination.

McLean's other innovation - a unified contract across all modes of transport - covered the whole journey. A shipper no longer had to negotiate separately with truckers, rail, and ocean carriers.

Once every link in the chain spoke the same format, the chain became predictable. A box loaded in Detroit on Monday had a known schedule to Tokyo on Friday. The real problem the container solved wasn't speed. It was unreliability.

And then a whole new economy got built on top.

Reliable shipping changed the logic of manufacturing. If you could trust that a box loaded in Shenzhen would arrive in Long Beach on a known date, you no longer needed to manufacture near your customer.

Manufacturers unbundled across borders. An Italian fashion brand could stitch garments in Indonesia. An American electronics company could assemble phones in 14 countries. The factory was no longer a place; it was a coordinated network.

Inventory shrank. Just-in-time replaced just-in-case. Retailers that had been regional went global. Detroit and Manchester emptied out as production reorganized around this new logic of coordination.

None of this was visible in 1956.

We are at the same moment with AI.

Every read of AI you hear right now is an automation read. Faster, better, cheaper. A productivity tool for work that already exists.

But that distracts us from what's truly possible.

AI does to knowledge work what the container did to cargo.

A knowledge job has always been a bundle of tasks. A lawyer reviews contracts, drafts briefs, judges precedent, owns client relationships, and bills hours - all stitched together inside one person.

AI unbundles it. The drafting becomes a component. The lookup becomes a component. The classification becomes a component. Each piece can be performed, audited, or replaced separately - by a tool, an agent, or a human.

Once work is modular, it's available to be reconfigured in fundamentally new ways. What matters now is coordination!

What AI drops to near-zero: translation cost.

Knowledge work has always been siloed because every domain speaks its own language. A lawyer's brief. A trader's spreadsheet. An engineer's spec.

Coordinating across them was a translation problem.

AI drops that translation cost to near zero. A language model can read a doctor's chart and write a discharge summary. Read a legal brief and surface the precedent. Read a spec and write the code. Listen to a sales call and update the CRM.

For the first time, silos can speak to each other. A handshake between any two domains is suddenly cheap.

And here's what's new.

The container needed every port, every carrier, every trucker to agree. ISO 1161. Standard dimensions. A single bill of lading. The format had to be settled before anyone could move.

AI doesn't.

It learns each silo's language separately and translates between them - without forcing any of them to standardize. Customer support keeps writing in their ticket language. Engineering keeps writing in their bug-tracker language. AI translates across these different formats. Coordination starts before anyone agrees.

That move has no historical precedent. Coordination without consensus. It is why the real impact of AI won't show up in cost savings on tasks that already exist. It will show up in domains that have been talking past each other for decades - suddenly working together.

"AI is not a tool of automation. It is a mechanism for coordination."

- Sangeet Paul Choudary, Reshuffle

Sangeet on this in Chapters 1–2 ↗

Why did Kodak lose but Shutterstock win?

Both faced commoditization of their scarcity-based business. Both saw it coming. Only one rebundled in time.

Watch where the value goes - and who follows it.

Kodak's value lived in physical imaging - film, paper, chemistry, prints.

It was a scarcity business. The premium came from owning the consumables nobody else could make at scale.

Digital imaging killed film. The smartphone killed the print.

The scarcity Kodak owned stopped being scarce. Photography didn't shrink - but the value stopped living where Kodak was.

Value walked through the layers Kodak didn't own.

First to storage and search - Flickr, Google Images, infinite photos meant finding one became the bottleneck.

Then to sharing and editing - Instagram. When capture was free, expression and circulation became scarce.

Then to attention - Facebook bought Instagram for $1B. By then the value was three layers past Kodak's front door.

Shutterstock started in the same kind of business.

A licensable image library. Curation, licensing, distribution. Cloud made all three cheap.

Same shock Kodak felt - different response.

Instead of defending the library, Shutterstock followed the constraint.

What enterprises struggled with wasn't access to images - it was workflow, rights clearance, compliance, brand governance.

Shutterstock rebundled the business around those layers. The library became one input, not the product.

Then AI commoditized image generation itself.

But Shutterstock wasn't defending an obsolete library. They were already sitting at the constraint that just got more acute: when anyone can generate infinite images, governing them becomes the job.

Variation control, recombination, rights, staying on-brand across channels. The new constraint was the old constraint, only louder.

Kodak owned a scarcity. Shutterstock owned a constraint.

Scarcities erode. Constraints shift but persist - and the firm positioned at the constraint inherits the value as it migrates.

The strategic question isn't "what new skill should we acquire?" It's "where is the new constraint, and what would it take to rebundle the firm around it?"

The framework is in Chapter 5 ↗

This part is about your job.

Pick the role closest to yours in the simulator on the left. The numbers below change. The lesson does not.

A job isn't really a job.

It's a bundle of tasks - drafting, looking up, judging, owning the relationship, verifying. They travel together because, historically, they were cheaper to keep in one person than to split.

Drag the AI capability slider up.

Watch which tasks commoditize first. The first to go aren't the ones you'd guess.

Try it before you scroll ↑

The job didn't disappear. It unbundled and then rebundled - around three new constraints:

Human judgment, where the call is genuinely yours. AI orchestration, a new role that didn't exist five years ago. Risk and verification, because AI can't audit itself.

This is what Sangeet calls the reshuffle.

Now zoom out.

This same unbundle–rebundle dynamic plays out at the organisation level (which functions get pulled into a central platform team, which get pushed out). And at the value-chain level (which firms become the new orchestrators, which get commoditized into suppliers).

It's the universal mechanic.

You don't need an AI strategy. You need a strategy for the conditions AI creates.

- Sangeet Paul Choudary, Reshuffle

The work isn't predicting which jobs AI will replace. The work is identifying where the reshuffle creates new constraints - and positioning yourself there.

Sangeet's full taxonomy is in Chapters 4–7 ↗

The AI-native company

Most companies use AI as a tool. The real prize goes to companies that use it as the engine.

TikTok vs Instagram. Same business, two architectures.

01 · Two architectures

Instagram, Facebook, YouTube used AI to enhance a social-graph architecture. The graph still determined what you saw - AI just ranked smarter.

TikTok used AI as the engine. Its feed was driven by what you watched, not who you knew.

02 · Two graphs

Instagram's content was shaped by your social graph - declared relationships.

TikTok inferred a behavior graph from scratch. Every swipe, pause, and replay trained the recommendation engine.

The graph emerged from behavior, not relationships.

03 · The constraint

TikTok imposed a hard rule: no video could be longer than 60 seconds.

Not a limitation - a positive constraint. It forced creators to deliver value fast, and let users watch dozens in a row.

Shorter videos meant tighter feedback loops. The engine learned faster than any traditional platform.

04 · Distribution

On Instagram, creators needed an audience before they got distribution.

On TikTok, a good video could go viral on its own - regardless of follower count.

Distribution came from the engine, not from your audience.

05 · The competitive shift

Analysts believed Facebook and Instagram had unassailable network effects from the size of their social graphs.

TikTok made the social graph irrelevant.

AI as engine elevates content based on behavior, not relationships.

06 · Catch-up, too late

Every competitor eventually copied the behavior graph.

But not before TikTok had asserted its dominance.

Copying late isn't the same as building around it.

07 · The lesson

With AI as the engine, the business no longer competes on the same old terms. It wins by changing the rules of competition.

AI as a tool makes you smarter at the existing game.

AI as the engine restructures the system around its capabilities. The basis of competition shifts.

Chapter 8 - The tool integration trap ↗