25Remake Work
Two hundred and twenty-five points of light arranged into the numeral 25, drawn as a wireframe network.

The system was broken.
Not you.

Four out of five people are no longer engaged at work, and a century of fixes has not moved the number. This book argues it never will.

One · the unit we forgot

Work starts with one whole person.

Not a headcount. Not a resource. A particular human being with a way of thinking, a reason to care, and a direction they are growing in. Every org chart ever drawn forgets this by the second row.

Twenty-five · the unit that works

Around twenty-five, something changes.

Small enough that one person can genuinely know everyone. Large enough to ship something that matters. Neuroscience and computer science arrive at nearly the same number from opposite directions — which means it is not a management preference. It is a constraint.

The network · connected, not stacked

Then you connect them sparsely.

Not one company of two hundred and twenty-five. Nine autonomous teams, thin high-value connections between them, and no head office deciding what everyone does on Tuesday. Every intelligence we know of is organised this way, because it is the only way that scales.

The architecture

This is the shape.

Large organisations exist because coordination is hard — not because large is better. AI is removing that cost. What is left standing is a structure you could actually design around people.

01The diagnosis

AI won't take your job.
It will take your org chart.

Every company on earth is racing to bolt AI onto a structure that was designed in 1911.

That is the mistake. The modern organisation was engineered for a world where coordination was expensive and human beings were the only processors available. AI made coordination nearly free. Which means the pyramid, the middle layers, the reporting lines, the endless alignment meetings — none of it is load-bearing anymore. It is legacy hardware running on a world that already changed underneath it.

25 REMAKE WORK is the first serious blueprint for the AI-native organisation. Not AI added to a company. A company designed around what AI now makes possible: twenty-five people, with an artificial intelligence nervous system underneath them, doing what used to require five hundred.

The number is not a slogan. Neuroscience and information theory converge on it independently, and the book shows the derivation.

79% of people are no longer engaged at work. Four out of five. Figures as cited in the book
1911 the year the operating system we still run on was published — Taylor's stopwatch, then Ford's line. The Principles of Scientific Management
0 fixes that worked. Not flatter structures, not unlimited PTO, not another engagement survey. A century of attempts

This is not an AI doom book and it is not an AI hype book. It is a pro-human argument that happens to be the most aggressive case for AI you will read this year: the machines take the coordination, the bureaucracy, the performance theatre, the eighteen-hour day. The humans take back the work.

Written by an operator with twenty years inside the architecture this book takes apart — for founders, executives, and anyone who has sat in a parking lot at 6:47 in the morning unable to open the car door.

She is not lazy. She is not ungrateful. She has the degree, the title, the compensation package that would make her twenty-five-year-old self weep with disbelief. And she cannot bring herself to open the car door.

The system was broken. Not you.

Chapter 1 — The Silent Epidemic

We have tried everything except the architecture.

Every serious attempt to fix modern work has been an attempt to run the same structure better. That is why the results keep looking the same.

Attempt 01

Better managers

Train them, coach them, give them frameworks. It helps — right up to the point where a good manager's main job is still absorbing the friction the structure creates. You have made the shock absorber better. The road is unchanged.

Attempt 02

More tools

Every tool of the last twenty years made coordination cheaper and then instantly spent the savings on more coordination. More channels, more dashboards, more status. The work did not get bigger. The overhead did.

Attempt 03

More culture

Values on the wall, offsites, a manifesto. Culture is real, but it cannot outvote structure. If the structure rewards visibility over contribution, culture becomes the thing people perform on the way to being promoted.

Each of these is a patch on a load-bearing wall. The argument of this book is that the wall has to come out.

02The idea

What if a company were shaped like an intelligence?

Not a pyramid. Not a matrix. The shape that every system we know of that actually thinks — brains, neural networks — converges on: many small autonomous units, sparsely connected, sharing signal without a boss in the middle of every path.

Piece 01

Teams of about twenty-five

One chief and twenty-four people. No VPs, no directors, no dotted lines. The chief's job is not to approve things — it is to develop people. Twenty-five is the largest group in which one person can hold everyone in mind, and the smallest that can own something real end to end.

Piece 02

AI as the nervous system

Not another tool bolted onto the org — the layer underneath it. The reason big companies exist is that coordination is expensive. Hand that cost to a system that routes information, matches people to work and carries signal without political filtering, and the case for being big quietly collapses.

Piece 03

Growth as the metric

Stop asking whether someone meets the standard. Ask what they are becoming. One question treats people as resources to be judged; the other treats them as trajectories. Organisations that optimise for the second will outperform — not because it is kinder, but because it is smarter.


Why twenty-five, and not eight, or eighty

This is the part of the book that was hardest to write, because the honest answer is uncomfortable: the right size of a working group is not a question of management philosophy. It is a question of information theory. Ask how many connections a unit can carry before signal starts degrading, and biology and computer science answer the same question independently — and land close to the same number.

It came out of three days at a whiteboard trying to design an organisation that kept failing in two directions. Either the centre held too much control and the design was a holding company with extra steps, or the small circles were too disconnected and it was a set of separate companies wearing a shared logo. The middle felt arbitrary until it stopped being arbitrary.

“They both point at a number uncomfortably close to twenty-five. This is not a metaphor. It is a structural finding.”

Chapter 5 — The Architecture of Intelligence

03The book

248 pages. One argument, carried all the way through.

It is not a book you agree with. It is a book you argue with — a diagnosis, a thesis about human nature, a structural argument, a full blueprint, and an honest reckoning with the economics. In that order, because that is the order the argument has to be made in.

  • Kindle also on Kindle Unlimited $9.99
  • Paperback 6 × 9 in $28.99
  • Hardcover library binding $37.75
  • 248 pages
  • English
  • Published 29 Apr 2026
  • ISBN 978-9699299896

Prices shown are Amazon US at time of writing and may change.

Inside

  1. Structure The real disruption is structural, not job-by-job Why “which jobs will AI replace?” is the wrong question, and what the right one costs you if you get to it late.
  2. Coordination The nervous system How AI replaces middle management as the coordination layer, and why that is what finally lets small teams operate at scale.
  3. People The Spectrum Seven dimensions of a human being that AI can finally see. The resume captured maybe four percent of a person. That constraint is now gone.
  4. Teams Intelligent team composition Using AI to answer the oldest unanswered question in management: who should work with whom.
  5. Economics Stress tested, not asserted $500,000 of revenue per employee, with the math shown and the objections taken seriously.
  6. The build The full blueprint The anatomy of a 25 org, the network that connects them, the 25OS technology stack, and what the first hundred would actually require.

Seven parts, seventeen chapters and a technical appendix. See the full contents and read the opening of every chapter.

04Reception

It is a new book.

It came out in April 2026. There are no cover blurbs on this page and no star rating, because I would rather show you none than show you borrowed ones. The only endorsement that means anything for a book like this is someone finishing it and then changing something about how their own organisation is built.

If that ends up being you, say so on Amazon. For a self-published book with no publisher behind it, that is the entire marketing department.

“I have spent twenty years inside the architecture this book is about to take apart, and I have come to believe, with the kind of certainty that only failure and long observation produce, that the architecture must be replaced. Not improved. Not reformed. Replaced.”

From the open letter, between the diagnosis and the design

05Talks for organisations

Bring the argument into the room.

A book gets read by individuals. A structural argument only goes anywhere when a leadership team hears it together and starts arguing in front of each other. That is what these sessions are for.

  • 45–60 minutes
  • English or Hebrew
  • Online or in person
  • Anywhere in the world
  • Q&A included

The Diagnosis

Why a century of engagement fixes has not moved the number, and what it means that the failure is structural rather than managerial. The most uncomfortable of the three, and the one that changes the conversation fastest.

  • All-hands, leadership offsites
  • No prior reading needed

The Architecture

The 25 model end to end — why the number is forced rather than chosen, what one node actually looks like on a Tuesday, how nodes connect, and what the economics have to clear for any of it to be real.

  • Founders, exec teams, boards
  • Works as a working session

The AI Inflection

The question is not how many jobs AI removes. It is what kind of organisation becomes possible once coordination is nearly free — and which parts of your current structure exist only to pay a cost that is disappearing.

  • Leadership teams mid-transformation
  • Pairs well with a follow-up workshop

Ask for a quote

There is no price list — what a session costs depends on the room, the format and the travel. Tell me the shape of it and I will come back with a number. Usually the same day.

Goes straight to Matan. No list, no sequence, no follow-up software.

06The author

Matan Elmalam

I have spent twenty years inside the architecture this book takes apart — building teams, running hiring, and watching brilliant people slowly stop bringing the best part of themselves to work.

Two things shaped the argument more than anything else. The first was Better Place, the Israeli company that set out to make gasoline obsolete. It did not survive, but for a few incandescent years I saw what happens when the quality of the aspiration determines the quality of the people it attracts. The second was studying what Steve Jobs understood better than almost anyone: that greatness comes from refusing to separate the pieces.

The thing I am actually good at is seeing people. Not their resume — the real person, what drives them, and what kind of team would let them do the best work of their life. I wrote this book over two years to turn that instinct into something structural, so it does not depend on me being in the room.

The technology exists. The need is urgent. The only missing ingredient is conviction.