Imagine the world as a long-running game of Dungeons & Dragons.

Every company at the table has a character sheet. The attributes down the left-hand side are familiar to anyone who has ever sat through a strategy offsite: scale, brand, capital, distribution, data, licences, relationships, systems, people. Next to each attribute, a score — earned over decades, priced by the market, understood by every other player at the table. Everyone knows the incumbents. Everyone knows everyone's strengths and weaknesses. Everyone knows how battles are fought, because the same battles have been fought, with minor variations, for forty years.

Campaigns under these rules are long. There are house rules — regulation, industry bodies, the way bids work in your sector. Upsets happen, but they happen inside understood physics: a challenger outmanoeuvres an incumbent, a new entrant finds an unserved niche, someone's acquisition goes wrong. Strategy, in this world, means playing your sheet well. You strengthen an attribute a point at a time. You pick the battles where your modifiers are highest. You avoid the players whose modifiers beat yours. It is a game of known quantities, and the winners are the ones who know the quantities best.

Now imagine that between one session and the next, every character sheet at the table is re-rolled.

Not yours. Not your strongest rival's. Everyone's. The scores that made the strongest player strong have been re-priced. Items that sat at the bottom of the inventory for years turn out to be worth more than the flagship equipment. Some of the flagship equipment turns to lead. And nobody at the table — not the oldest incumbent, not the sharpest challenger — has finished re-reading their own sheet, let alone anyone else's.

Everything is up for grabs again.

"AI did not add a new weapon to the old game. It changed the rules, redrew the map and re-rolled every character sheet."

One boundary on the metaphor before it does any work, because the boundary is where the strategy lives. A re-roll is a repricing, not an erasure. The sheet in front of you still lists the same items: the brand is still on it, the customer relationships are still on it, the licences and the data and the twenty years of operating history are all still on it. What changed is the modifier next to each one. Some scores went up. Some went down. Some went to zero. The company that reads its sheet as blank will throw away assets that just became more valuable. The company that reads its sheet as unchanged will defend assets that just became worthless. Both misreadings are common, and both are expensive.

What re-rolled the sheets

The cause can be stated once, briefly, and then it can stop being the story — because the story of this book is not that AI arrived; it is what the arrival did to the scoreboard.

Cognition became abundant. Software became generative. Expertise became callable. Coordination became cheaper. Products that once required departments could be built by small teams. Customers acquired their own intelligence. Competitors became cheaper to create. Entire business architectures — the kind that used to need a platform division and a nine-figure budget — became affordable to attempt.

The scale of the input-price collapse is worth one hard look, because it is the physical driver under every chapter that follows. Epoch AI measured something more useful than "how much does the best model cost?" — they measured the price of buying a fixed level of capability over time. The price of matching GPT-4-level performance on PhD-level science questions fell by roughly forty times per year¹. Across all the milestones they measured, the annual decline ranged from 9× to 900× — and Epoch themselves flag that the fastest drops came most recently, so it is less clear those rates will persist. Looking forward, Gartner projects that inference on a trillion-parameter model will cost providers over 90 per cent less in 2030 than it did in 2025² — a forecast rather than a measurement, and used here only as a direction of travel.

Read the range rather than the headline. Nine-to-nine-hundred is not a tide coming in evenly along a coastline. It is a series of specific tasks becoming nearly free at different moments, for reasons that have to do with benchmark structure and model architecture rather than with anyone's business plan. Which means no company gets to schedule this. "We'll deal with it when it arrives" fails as a plan for one simple reason: it has already arrived somewhere in your value chain, and not somewhere else, and nobody sent a notification. The full economics of what happens when thinking gets cheap — for you and, more importantly, for everyone around you — is territory we have mapped in its own right.

When an input to nearly every business model gets cheaper at that rate, every attribute that monetised the old scarcity gets a new modifier. That is the re-roll. The interesting question — the one the rest of this chapter answers — is what the new modifiers look like.

The six fates of an inherited asset

Scott's own framing of the re-roll carries the honest complication that most AI commentary flattens: "Your past and history does give you some assets. Some items can be taken into the future. Some will increase in value and some will decrease." The re-roll is not a uniform devaluation of the past. It is a repricing — and repricings have winners on both sides of the ledger. Every inherited asset on the character sheet meets one of six fates.

"The past still matters — but it no longer carries its old score automatically."

Hold that sentence against the two misreadings from the top of the chapter, because it kills both. "History is worthless now" — fates one, two and five say otherwise: some history appreciates, some converts, some turns from liability to feedstock. "Our history protects us" — fates three, four and six say otherwise: some history strands, some inverts, and some actively arms the other side. The honest posture is neither confidence nor panic. It is re-reading the sheet, item by item, with the new modifiers — which is precisely the exercise this book will equip you to run by its end.

When a moat becomes a map

The sixth fate deserves unpacking because it is the one the old mental model cannot see at all. A moat, in the classical framing, is a documented advantage: the integration depth, the branch network, the configuration expertise, the switching costs. Documentation is legibility. And under abundant cognition, legibility cuts both ways.

Walk the inversion in three moves. First: your moat describes what your customers value — that is why it holds them. To an attacker who can now afford to build, that description is a requirements document, free of charge. Second: your moat's defensive cost structure — the branches, the seats, the integrations, the headcount that maintain it — is exactly the part the attacker declines to replicate. They take the requirements and skip the costs, because the costs were artefacts of the old price of cognition and coordination, not of the customer's need. Third: your switching costs price the attack for them. Everything below the switching threshold is free territory; the attacker knows precisely how good they must be, because your moat published the number.

The more legible the advantage, the better the map. This is not an argument for hiding — opacity is not a strategy either. It is an argument for knowing which of your defences are still defences.

The tell: advantages stopped translating

How would you know your industry is in a genuine re-roll rather than another technology cycle? The previous waves — ERP, CRM, cloud — all favoured incumbents in the end. New technology arrived; distribution, relationships and balance sheets absorbed it; the character sheets survived with minor edits. A re-roll has a different empirical signature: entrenched advantages stop translating.

That signature is now measurable. Menlo Ventures' enterprise data found it at the AI application layer, and their own commentary carries the astonishment better than any paraphrase could:

"At the AI application layer, startups have pulled decisively ahead. This year, according to our data, they captured nearly $2 in revenue for every $1 earned by incumbents — 63% of the market, up from 36% last year when enterprises still held the lead. On paper, this shouldn't be happening. Incumbents have entrenched distribution, data moats, deep enterprise relationships, scaled sales teams, and massive balance sheets."

Read the middle sentence slowly. On paper, this shouldn't be happening. On paper — meaning, on the old character sheet. The distribution is real. The data moats are real. The balance sheets are real. Every one of those advantages still exists; what stopped working is the modifier each one carries into the new game. And note the speed: thirty-six per cent to sixty-three per cent in a single year. Repricing happens at market speed, not at forecast speed.

All at once

Everything so far could be read as a story about your company: your sheet, your fates, your moat. That reading misses the strategic fact of the era, and Scott's own summary of the re-roll refuses to allow it: "And that's happening to all companies, in all industries, all at once."

Simultaneity changes the category of the problem. When one player's sheet is repriced — a nationalised competitor, a firm caught by a regulation change — that player adapts, and the game absorbs it. This is normal competitive dynamics; every board knows how to think about it. But when every sheet at the table changes at once, three things break together. The relative positions become unknown to everyone simultaneously — including the winners, who can no longer be sure which of their strengths still score. The informational basis of the old strategy — known strengths, known battle patterns, known physics — is gone, and it was that shared knowledge, not any single advantage, that made the old game playable. And nobody has finished re-reading even their own sheet, which means the market's current behaviour reflects the old scores while the new ones are still being discovered.

Which yields the conclusion this chapter exists to establish: the repricing is a property of the table, not of any player. Analysing your own sheet — however honestly, however rigorously — is analysis of the wrong object.

So the instinct that follows a re-roll — study our own sheet harder, audit our own assets, accelerate our own adoption — is understandable, diligent, and aimed at the wrong altitude. If the board is in motion, what exactly should a company be analysing?

That question has a precise answer, and it is where the unit of strategy has quietly moved.

Here is what studying your own sheet harder looks like in practice. Three questions, asked in some form in nearly every boardroom this year:

How should our company use AI?
How should our people become more productive?
Which of our workflows should be automated?

Give these questions their due before anything else, because they deserve it. They are well-formed. They are answerable. They are asked by diligent people, and answering them produces real value — whole transformation programmes are built from them, and some of those programmes genuinely work. Nothing in this chapter says the questions are foolish.

It says they are aimed at the wrong altitude. All three take the firm as the unit of analysis at the precise moment the repricing is happening to the table. AI is operating one level higher than the questions are being asked — a mismatch we have called the wrong-altitude problem in our board-level work on AI capital allocation.

An industry is a bundle

To see what the higher altitude looks at, start with what an industry actually is. An industry is a bundle — of capabilities, scarcities, institutions, agreements and customer behaviours. "Law firm", "insurer", "SaaS vendor", "consultancy" are not natural kinds; they are equilibria. Each one is the shape work settled into when thinking and coordinating cost what they used to cost. At those prices, holding certain capabilities inside one organisation was cheaper than sourcing them, composing them or generating them — so the bundle held, and the bundle got a name, and the name got an industry association and a conference.

Abundant cognition attacks the glue, not the name. Watch the mechanism in three verbs.

Unbundle. The cost structures that made capabilities cheaper to hold together dissolve. The analysis department, the software platform, the distribution arm, the advisory desk — each was inside the firm because integration was cheaper than the alternative. When cognition and coordination get cheap, the alternative gets cheap, and the components come loose.

Reprice. Each freed component gets its own new price. This is Chapter 1's six fates applied at component level: some components appreciate on their own, some strand the moment they are separated from the bundle that hid their decline.

Recombine. The loose, repriced pieces become available for new configurations that were uneconomic a year earlier — including configurations no incumbent would choose to imagine, because the imagining is now affordable to people with no stake in the old bundle.

Scott's version of this is shorter: "AI is changing how the world is put together — how it can be combined and recombined. And it's changing how industries work: combining and recombining."

What the pieces become

Unbundling sounds abstract until you watch what the freed components turn into. Eight transformations, each already underway somewhere:

The fourth transformation is the one with public numbers attached, and they are worth pausing on. Retool's build-versus-buy research found that 35 per cent of teams have already replaced at least one SaaS tool with a custom build, and 78 per cent expect to build more custom internal tools in 2026⁴. Read that as an industry event, not a procurement trend: a category being partially unbundled by its own customers. No competitor did that. The bundle's glue dissolved, and the customers walked through the gap.

The eighth transformation is the punchline of the whole mechanism:

"A company that once owned the entire value chain may discover that it owns the least valuable part of it."

The mechanism under the line: owning the chain was valuable when integration was the scarce act — when stitching the layers together was the hard, expensive, defensible work. When composition gets cheap, ownership of the chain and ownership of the scarce layer come apart. You can hold every link and find that the value settled in the one link everyone can now reach without you.

The altitude claim

Scott, reviewing his own body of strategy work, put the correction plainly: "A lot of this I've looked at one company at a time. You can take all of it to the industry level."

That sentence is this book's licence to exist, so it is worth being precise about the lineage. Our Terminal Value Doctrine holds the board method — name the scarcity your model depends on, push it to a boundary, ask what still deserves to exist — and its Reshape variant runs that method at industry scale as an exercise. What the canon did not hold — what nothing in it held until now — is the claim Chapter 1 established: every competitor's sheet is re-rolled too, simultaneously. Once that claim lands, the method's own logic forces the conclusion:

"The old unit of strategy was the firm. The new unit is the moving value system around the firm."

"Moving value system" is doing precise work in that sentence, so pin it down before it can blur into consultancy vapour. It means: the set of value layers around your current position — the layers compressing, the layers expanding, the destinations value is migrating toward — plus the actors repricing them. Those actors come in three kinds: customers, competitors and constructors. For now they are a list; the next part of this book is about what each of them does to you. The point here is narrower: the value system moves, and it moves whether or not your firm does.

The board question

If the unit of strategy has moved, the question at the top of the strategy conversation has to move with it. The old question — how do we make this company better? — is not wrong. It is subordinate. It can only be answered correctly downstream of a bigger question, and most boards have never asked the bigger one in writing:

Every clause earns its place. When this industry is recombined — not if; Chapter 1's simultaneity claim removed the if. Where will value live — the clause assumes migration rather than destruction, an assumption this book will justify in full when it maps where value goes. Who will control it — deliberately open, because the controlling entity may not exist yet; incumbency makes you a candidate, not a default. What must we become — the hardest clause, because becoming may mean constructing something your current organisation would not naturally build. Arrive there first — the time clause; windows close, and the second half of this book runs on that clock.

There is a hidden premise inside the board question, and it is the reason the next chapter exists. To say where value will live, you have to know what value is now made of — what became cheap, what stayed dear, and what that did to the scarcities every business model quietly rests on. Before a board can read its industry's future, it has to read the new economics underneath it.

"The task of creating companies is collapsing. Possibility exploding. The scarce resource is knowing what should and will exist. Judgement."

Four beats, in Scott's own compression. The previous chapter ended by asking what value is now made of; those four beats are the answer in miniature, and this chapter decompresses them. Because the era is routinely described in one loose sentence — intelligence became cheap — and that sentence is half true. The false half is where strategies die.

Three things got cheap. Three did not.

What collapsed is precise. What it did not collapse is the whole game.

Answers and truth. Anyone can now produce a hundred plausible answers before lunch — market entries, pricing schemes, product architectures, each fluent, internally consistent, formatted for the board pack. What none of that production buys is knowing which answer survives contact with reality. Truth still costs what it always cost: evidence, testing, someone bearing the consequence of being wrong. The flood has a side effect worth naming: plausibility inflation. When every option arrives polished, polish stops being a signal of correctness — and organisations that used polish as a proxy for quality lose their proxy.

Generation and elimination. Options, drafts, strategies, scenarios multiply at near-zero marginal cost. Killing one still requires a test against the world — a customer who says no, a price that doesn't hold, a regulator who won't wear it. Our work on strategic fog found firms discovering this the hard way: a partnership that generated fifty-plus argued futures in its best strategic year and formally eliminated none of them — generation got cheap, elimination did not, and nobody moved the instrument. A year of magnificent thinking can leave a firm holding triple the possibilities and none of the closure.

Possibility and judgement. The possibility space expands for everyone at the same time — which means the discriminating act, choosing which possibility deserves to become real, gets harder and more valuable simultaneously. This is the economic inversion we have written about since the beginning of this wave: when production gets cheap, scarcity moves to the judgement, intent and context that decide what should exist — to trust, taste, accountability and relationships.

The advantage is no longer answer 101

Follow the shift in three steps, because each step strands a different business model.

When producing an answer required weeks of human effort, the capacity to produce answers was the business. The analyst pyramid, the research house, the agency studio — all of them monetised the gap between a question asked and an answer delivered, and the gap was wide enough to build careers inside.

At answer-abundance, differentiation by production collapses. "When anyone can produce a hundred plausible answers before lunch, the advantage is no longer answer number 101. The advantage is choosing the question worth answering." What discriminates now: recognising which possibility deserves to become real; rejecting attractive but structurally weak futures; making a consequential decision while the opportunity to act still exists.

And the scarce resource has a fuller name than "judgement." The manifesto's own definition carries the two clauses that the rest of this book will spend a part each on: "The scarce resource is judgement: knowing what should exist next, where value is moving and how quickly the opportunity window will close." Where value is moving — Part II's first half. How quickly the window closes — Part II's second. Keep the definition whole; most firms hold the first clause and drop the other two.

One caution from our cheap-thinking work, compressed to a sentence because it decides who this chapter applies to: the exposed work is not the hard work — it is the work with a known shape before it starts, and reproducibility, not difficulty, is the test. Plenty of genuinely difficult work has a known shape. All of it is on the floor.

The token premise — stated once

Our earliest thesis on this economics was deliberately blunt: whoever spends the most tokens wins. It came out of the agent-era observation that machine cognition is R&D — that token spend, bounded by clear objectives and validated outcomes, compounds while rationed cognition stands still. The instinct remains right, and nothing in this book retreats from it: cognition should be funded, not rationed. A company minimising token consumption while its competitors run machine intelligence as continuous research and development is optimising the wrong resource — the CFO instinct to treat the one input that just got cheap as the place to save money is precisely backwards.

But this era forced a correction, and the three pairs above explain why: token burn alone is not an advantage. Undirected cognition generates more possibilities and thickens the fog — it buys generation, which was already cheap, and none of the elimination, which was not. The mature law reads: "Whoever converts the most tokens into better questions, faster evidence, stronger decisions and compounding capability wins." That law — how the conversion works, how it is governed, what it does to the original manifesto — gets its full treatment in its own piece; here it is stated once and holds the rest of this book honest.

"Spend machine cognition lavishly. Spend human attention ruthlessly."

What the four beats assemble into

Put the three pairs back together and the strategic consequence follows. Abundance is symmetric: your cheap thinking is also your customer's, your competitor's, your constructor's. The goal, therefore, is not more thinking — it is consequential thinking: cognition that changes a decision while the decision still matters. Maximise consequential cognition; everything else is fog production at your own expense.

The manifesto compresses the whole machinery into four lines, and they are the book's table of contents in disguise:

Direction and the clock that governs it come next. Capital is the response after that. Evidence — the part most strategy books never reach — is where this one ends up.

First, though, a harder question than any economics. Everything in this chapter described capabilities: cheap generation, expensive judgement, expanding possibility. Capabilities in whose hands? The same price collapse that funded your pilots funded everyone else's moves — and the pressure it creates does not arrive through one door. It arrives through three, at the same time.

The quarterly pack, at a company doing everything right. Page eleven is the AI adoption dashboard, and it is green: Copilots deployed across the business, four pilots shipped to production, staff training complete ahead of schedule, an internal innovation award, a case study the vendor wants to publish. Two pages later, the commercial section: pricing pressure in the core line, a renewal that came back smaller, a prospect who paused the engagement because — the note is verbatim from the account team — "they did the first pass themselves."

Nobody in the room connects the two pages. Why would they? They sit in different sections of the pack because they live in different mental categories: one is the AI programme, going well; the other is market conditions, tightening. When the pattern repeats for a third quarter, the diagnosis writes itself: execution failure. Wrong tools, insufficient training, change management too slow. Do adoption harder.

The diagnosis is wrong, and it is wrong at the category level, which is why doing adoption harder will not touch it.

Adoption is something your firm does inside its walls: tools, training, workflows, dashboards. Pressure is something the market does to your industry — and it arrives whether or not you adopt anything. We first drew this distinction for professional-services firms, where the demand side made it impossible to ignore; the re-roll makes it general. The pack puts the two on different pages because the prevailing mental model does. This chapter is about what happens when you put them on the same page.

Three doors at once

Pressure arrives through three channels simultaneously — the routes by which a fall in the price of thought becomes a rise in the number of plausible futures, as our cheap-thinking work mapped them: customers, competitors, constructors. Each deserves its own look, because each is noticed at a different speed — and the most dangerous one is noticed last.

Customers: the door nobody logs

The mechanism in one sentence: your customer's cost of thinking fell by the same order of magnitude as yours, and they spent the windfall on the part of your work they most resented paying for.

"They do not need to change suppliers to disintermediate an incumbent. They can simply buy less." That is why the customer channel is the one incumbents notice last — no competitor appears, no deal is formally lost, nothing shows up in the win/loss analysis. The revenue just arrives smaller, engagement by engagement, and every individual instance has a plausible local explanation.

Where the unit of sale is most visible, the repricing is already measured. In legal services, research collected by Clio found 67 per cent of corporate legal departments and 55 per cent of law firms expecting AI to change how hours are billed, and 71 per cent of buyers already preferring a flat fee for an entire matter⁵ — and, in the sentence that should be read aloud at every partners' meeting in every industry: "clients have moved faster than the industry has." The demand side repriced before the supply side decided to. The pressure arrives first as a pricing conversation, which is exactly why it gets filed as a commercial issue instead of a strategic one.

There is a second-order effect that compounds quietly. A customer who has produced their own first pass has also produced an opinion — about scope, about duration, about what this work should cost. You are no longer proposing into a blank page. You are negotiating against a document you did not write.

Competitors: mutation, not efficiency

"Competitors do not merely become more efficient. They become cheaper at changing what they are."

The distinction matters because boards instinctively model the first threat and miss the second. A more efficient competitor plays the same game with better numbers — uncomfortable, but familiar. A competitor with a collapsed cost of attempting structural change is a different object. Designing an offer, modelling its economics, testing it against a segment, specifying what would have to be built: that work used to consume a strategy function for a quarter, which rationed it to once every few years. When attempting gets cheap, the rate of attempts rises — and every attempt is another plausible future you now have to hold in mind.

Public markets have started pricing what that does to incumbents. Sapphire Ventures' software indices fell 20 per cent — pure SaaS 23 per cent — through 18 February 2026, with the sector decoupling from an essentially flat Nasdaq (their broader software index, IGV, down 32 per cent)⁶. They name the drivers directly — budgets shifting to AI, pricing moving off seats toward measurable value, continuous model improvement raising displacement risk — and their summary line deserves its full weight: "risk is up, and terminal value assumptions are down (for now at least)." Carry their hedge with the quote, because stripping it would be the kind of citation this book refuses: Sapphire themselves allow that investors may be "selling first and asking questions later." The argument does not need the re-rating to be permanent. Even discounted, a public-market repricing is a price being put on a set of futures — the fastest-moving evidence available that the pressure is real.

Constructors: the door that includes you

This channel is usually argued badly — "there are more startups now" — so it is worth arguing against its own most convenient evidence. The startup data does not straightforwardly support the lazy version. Crunchbase recorded $300 billion invested across 6,000 startups globally in Q1 2026, an all-time record, with AI taking $242 billion — 80 per cent of global venture funding, against 55 per cent a year earlier. But seed deal counts fell 30 per cent year on year, to 3,800; the dollar increase came entirely from larger rounds⁷. Capital is concentrating, not scattering. If the constructor argument rested on a rising count of funded entrants, it would be in trouble.

It rests on something else: the falling cost of construction available to everyone. The customer who decides to build rather than buy. The mid-sized competitor who can now specify and ship something that would previously have required a platform team. The people who understand one workflow deeply and no longer need permission or capital to act on it.

The asymmetry

Now hold the board scene from the top of this chapter against the three channels, and the paradox dissolves into arithmetic. Why can a company lead its sector in internal AI adoption and still lose ground?

"Its cognition compounds inside the walls at the speed of its change programme. Everyone else's cognition compounds across the market at once."

Unpack it, because this is the sentence the whole book was bought for. Inside the walls: adoption compounds through training calendars, rollout waves, governance gates, the sequential machinery of organisational change. The gains are real. They are also serialised — each wave waits for the one before it, because that is what a change programme is.

Across the market: pressure compounds through every customer, every competitor and every constructor discovering cheap cognition in parallel, without coordination, without a programme, without waiting for each other. The market does not hold steering meetings.

These are different exponents, not different levels of effort. The firm's curve is bounded by how fast one organisation can change; the market's curve is bounded by how fast thousands of independent actors can each make one move. No amount of internal excellence changes the outside exponent — which is why the honest name for the situation is not "execution gap" but "altitude error."

And why can't productivity close it? "Productivity does not close that gap by itself. Productivity applied to a depreciating commercial unit simply produces the old value faster." If the unit you sell is being repriced by the three channels, being faster at producing it is not a defence — it is a more efficient way of harvesting a declining asset without noticing. This is the trap our board-level work named early: optimising the current workflow because the mental model says AI is a cheaper staff member.

The scoreboard at market scale

If adoption-without-altitude were merely a theoretical risk, the market data would look random. It does not. BCG's research finds only about 5 per cent of companies achieving substantial financial gains from AI, while roughly 60 per cent see little or no material return despite heavy investment — and the 5 per cent show roughly four times the three-year total shareholder returns of AI laggards⁸. McKinsey's State of AI research adds the differentiating trait: the high-performer cluster is 3.6 times more likely than peers to be pursuing transformative change rather than incremental improvement⁹.

Read the scoreboard correctly. The 5 per cent are not adopting harder — heavy investment is exactly what the 60 per cent are doing. The leaders are repositioning against where value is going, and the reward for that altitude difference is now measured in shareholder returns. The differentiator is not spend. It is which question the spend is answering.

One boundary, honestly drawn, before this chapter closes. None of this says adoption is wasted. Adoption done well lowers delivery cost, builds the muscle the next move will need, and funds the transition — it buys runway. What adoption cannot do is act as a defence, because the defence has to happen at the altitude where the pressure operates. An operating asset was being asked to do a strategy's job.

Later in this book, this paradox runs end to end on a worked incumbent — an adoption leader whose industry gets re-rolled around it on the page. Before that can mean anything, though, one reorientation has to land, and it is the sentence this chapter has been building toward:

"The task is not to accelerate whatever already exists. The task is to discover what still deserves to exist."

Discovery needs a map. Pressure moves things — so the first mapping question is where the value goes when it leaves.

"AI does not only automate work. It can remove the reason the work existed."

That is a stronger and stranger claim than the automation story, and the difference between the two is worth the whole chapter. The standard boardroom framing of AI disruption is displacement: a competitor will do what we do, cheaper or faster, so we must get cheaper and faster. Sometimes that happens. But the dominant pattern of this era is migration: the category itself moves to a different value architecture, leaving the incumbent's operating model intact, well-run — and increasingly irrelevant. Nobody beats you at your game; the game stops being where the value is. The pattern is older than AI — every collapsed scarcity in commercial history has relocated its value somewhere — and our board-level work has mapped it as the value-migration lens.

The conservation law

Why is "where did the value go?" always an answerable question? Three steps.

First: value in a business is rent on a scarcity. Every margin line, traced far enough down, rests on something the customer could not cheaply do or get elsewhere — expertise, coordination, software, access, trust. Second: when AI dissolves the scarcity, the rent stops — but the customer need that priced the rent did not vanish. People still need the analysis, the software, the advice; they just stop needing you to be the expensive path to it. Third: the value therefore migrates to whatever is now the binding constraint on serving that need. It does not move evenly, and it does not wait for the incumbent to approve the new destination.

What cheapens is the generic layer, and the manifesto's anaphora names it plainly:

If migration is invisible to most boards, it is because of an accounting artefact, not a failure of intelligence. The P&L records the rent falling — that shows up as pricing pressure, shrinking scope, compressed margin. No internal statement records where the value re-accrued, because it re-accrues outside the firm's chart of accounts: in the customer's internal capability, in a new layer of the chain, in a category that has no line item yet. The destruction is visible; the destination is not. Boards read the visible half and conclude the value died.

Where value moves to

Value moves toward what remains difficult to reproduce. That sentence is easy to nod at and hard to use, so here is the destination list read carefully — each entry with the reason it resists reproduction, because the reasons are what make the list a map rather than a mood board.

Notice something about this table before moving on: every destination is an attribute a character sheet can carry. Inherited assets can sit on these squares — a brand can hold trust, an archive can hold proprietary context, a licence is authority. Whether yours do, and what it costs to move them there, is a capital question this book takes up in Part III.

Destruction at one layer is creation at another

"Destruction at one layer is creation at another. Compression in one part of an industry creates expansion somewhere else."

The double movement is the part of migration that doom commentary misses entirely. An old bundle may disappear while a new set of specialised transactions multiplies around it — the compressed layer's spend does not evaporate; it re-forms as demand for the layers that make the now-cheap output usable. Where does the expansion show up? Generically, in five places: verification of abundant claims (someone must say which of the hundred answers is true); governance of machine action (someone must be answerable for what the systems do); composition and integration of generated parts (someone must make the pieces hold together); the interfaces where agents transact (someone must own the surface where machine buyers meet machine sellers); and the consequence-bearing signature at the end of the chain (someone must sign).

The two movements have different visibility, and the difference explains most of the era's bad strategy. Destruction is legible: it has line items, and they are going down. Creation is illegible early: new categories have no line items yet, no benchmark, no conference. Boards therefore over-perceive threat and under-perceive destination — they can see what is dying in their own accounts and cannot see what is being born outside them.

A map is not a strategy yet

One honest boundary, stated before the next chapter makes it unnecessary. Everything in this chapter is directional: which layers compress, which expand, where the destinations sit. None of it is a timetable. This book will not tell you how fast your industry's value is moving, because no honest source has that number — and a book that invented it would be doing plausibility inflation of its own.

But the boundary cuts both ways, and this is the uncomfortable half: two firms holding identical migration maps can end in different places — one arrives, one watches. What separates them is not insight. It is speed relative to the window. The map answers where. Strategy also needs how fast — and those two questions have traditionally lived in different rooms of the same building, one with the strategists, one with the finance team.

The instrument that forces them into the same room has a name, and it is the next chapter's subject.

"Time is a big player. Velocity matters because it's about time and urgency — the horizon to change, the speed of change accelerating or reducing. Time is being compressed."

That is Scott, thinking aloud about why the instrument this chapter names has velocity in it rather than just value. And it names the structural problem directly: in most companies, direction and time live in different rooms. The strategy function produces direction — where the market is going, what the future looks like. The finance function produces time — when returns arrive, how long the cash holds. The two meet once a year, in a planning cycle, and the era this book describes punishes that arrangement without mercy. The last chapter ended with two firms holding identical maps and ending in different places; the difference was made in the room these functions don't share.

The instrument

Both components need defending against their weaker cousins, because each has a familiar impostor.

Direction is not trend commentary. It is the migration map of the last chapter made specific to your industry: which layers are compressing, which are expanding, where the destination scarcities sit, which of your inherited attributes can occupy them. Trend commentary tells you AI is important. Direction tells you which layer of your value chain the rent is leaving and which layer it is re-accruing in.

Speed is not urgency theatre. It is a claim about the window — how long the current configuration stays open to action before positions harden. Urgency theatre says "move fast" about everything equally, which is the same as saying nothing. A speed reading says: this scarcity is dissolving now, this layer's recombination has not yet settled, and the difference between those two clocks is where your time actually is.

The instrument asks two questions that traditional strategy asks separately, if at all:

"Direction without time is not strategy. Arriving at the right destination after the opportunity has closed has no economic value."

The inverse holds too, and completes the definition: time without direction is not strategy either — it is motion. The firm that responds to urgency by launching initiatives in every direction is spending its window rather than using it. A velocity is only a velocity when it has both coordinates.

What compressed time actually means

"AI does not merely make work faster. It brings future work states into the present." That sentence sounds like rhetoric until it is cashed out in mechanisms, so here are four, each concrete enough to check against your own operation:

Research that once took months informs a decision this week. The evidence lag that used to define planning cycles — commission the study, wait a quarter, decide next cycle — collapses to days. The bottleneck moves from producing the evidence to being organisationally capable of acting on it while it is fresh.

Multiple product architectures can be explored in parallel. Sequential option-testing — build one, learn, build the next — becomes concurrent. The calendar stops serialising your learning; three possible futures can be probed in the time one used to take.

A proposition can meet a customer while the original thought is still alive. The idea-to-contact interval drops below the attention span of the idea's author. What used to go stale in the build queue can now be priced, shown and refused while the insight that produced it is still loaded.

Evidence can return before the people involved have cognitively moved on. The loop closes while context is still in working memory — which changes not just speed but the quality of the update, because nobody is reconstructing what they meant three months ago.

And underneath all four, the era's strangest property: "Each cycle does not only improve the next cycle. Each cycle can make the next cycle happen sooner." Compounding cadence, not just compounding output. The book will not pretend to quantify that second derivative — no honest source does — but its direction alone breaks the planning assumption most firms still run on.

Because here is the scissor those mechanisms open. The credible planning horizon is shrinking — capability changes quarter to quarter break long-range assumptions — while the number of possible moves is expanding, through every channel Chapter 4 walked. The market branches faster, while many companies still collect evidence at the speed of an annual planning cycle. Both blades move at once; that is what makes it a scissor and not a squeeze.

One integrity note, before the clocks. Compressed time does not mean rushing authority or faking evidence. It means compressing every avoidable delay between question, construction, contact with reality and decision — the delays that are habit rather than physics: approval queues for reversible probes, sequential work that could run parallel, evidence waiting for the next quarterly meeting to be looked at. The irreducible delays — real evidence takes real contact; real authority takes real accountability — stay. That distinction is what separates velocity from recklessness.

The two clocks

Speed relative to what, though? The window has a structure, and our fog-and-clocks work holds it in full: strategic fog thickens when the market's branching rate outruns the firm's evidence-backed elimination rate — the outside clock generates possible futures; only the inside clock retires them. For this book's purpose, the structure compresses to three sentences:

Now run the arithmetic on waiting, because every board runs it eventually and most run it wrong. Waiting shortens the runway: the old unit keeps depreciating while you deliberate — the outside compounding does not pause for your process. And waiting does not start the proof clock: no evidence accumulates from deliberation; the successor is exactly as unproven after a year of consideration as before it. Which yields the line that ends the "wait for clarity" argument on arithmetic rather than temperament:

"Delay is the only move that worsens both clocks at once."

Every other move — probing, harvesting, converting, even a failed construction — improves at least one clock. A probe that dies produces evidence; a harvest funds; a conversion builds. Delay alone spends runway and banks nothing. "The market does not preserve an opportunity because an incumbent needs another quarter to become comfortable."

The recombination window

The thesis of this book has carried a time clause since Chapter 1, and it can now be defined properly. The recombination window is the period during which an industry's re-rolled pieces remain unclaimed — while inherited assets can still be attached to the new value layer, and the roles in the new configuration are still open. It closes not by announcement but by accumulation: early recombinations win customers, set interfaces, absorb the migrating value, and position by position, the new game's sheets fill in.

Two properties, honestly bounded. First: windows differ by industry — which is precisely why TVV is an instrument rather than a forecast. This book cannot tell you your window; it can tell you how to read it, and reading it per industry is the point. Second: windows close asymmetrically. For observers they close slowly — nothing visible changes for quarters at a time. For entrants they close suddenly — the interface is set, the customers are committed, the destination squares are occupied. The visible calm of a closing window is the last chapter's visibility asymmetry wearing its time face: destruction you can see, destination you can't, and a clock on both.

Part II is now complete as an analysis: pressure explains why adoption alone loses; the migration map says where value goes; velocity says how fast the window moves. And all of it together changes precisely nothing — because analysis moves no money. Maps and clocks change nothing until the budget believes them, and the budget is the one document in a company that cannot lie about its strategy.

That document is where this book goes next.

"Capital allocation is strategy made visible. Every dollar, every hire, every acquisition, every product roadmap and every quarter of executive attention is a vote for a particular future."

Strategy decks can say anything. They are cheap to produce — cheaper than ever, as Part I established — and they carry no consequence for being wrong, which is exactly why they can afford to be visionary. The budget is different. The budget is falsifiable. Read any company's true strategy off four allocations: where the dollars went last quarter, where the senior hires went, what occupied the roadmap slots, and what the executive team actually spent its attention on. The fourth is the tell most analyses skip — attention is the scarcest allocation in the building and the least audited. A board can claim its industry is being recombined and spend forty-eight of fifty-two weekly executive meetings on the current model's operations. The deck says transformation. The calendar says harvest-by-neglect. The calendar is telling the truth.

Scott's framing of his own venture work applies to every company: "This is about capital investment — capital allocation." Not tools. Not adoption. Where the money and the attention go, under re-rolled conditions. This chapter is the grammar for that.

Reclassify before you allocate

The first act of honest allocation is reclassification, for a blunt reason: you cannot vote for a future while your asset register still speaks the old language. The register says "flagship platform", "market-leading service line", "strategic partnership" — labels priced by the old game. Under re-rolled conditions, assets sort by one question: how does this asset's value behave as AI improves? Our board-level work formalised the answer as three classes — stranded assets, whose returns depend on a scarcity AI is dissolving; convertible assets, whose latent value can travel if transformed; and compounding assets, which appreciate as AI improves. Chapter 1's six fates were this taxonomy seen from the character sheet: the appreciating and raw-material fates point at compounding, the travels-if-converted fate at convertible, and the stranded, inverted and moat-to-map fates at the class whose name says it.

What the classes feel like inside a board pack, because that is where you will meet them: a stranded asset looks like healthy cash flow with a quietly deteriorating multiple — the yield is real and the terminal value is falling. A convertible asset looks like nothing at all — archives, methods, histories that appear on no register because conversion has no cost code. A compounding asset looks too small to defend — a line item some efficiency review keeps trying to cut, because its returns accrue to other lines.

The capital grammar

Reclassification done, the response grammar is three verbs at asset level and three motions at company level:

One mechanical point before the motions are walked, and it is the one that decides whether the grammar works at all: these are parallel motions, not phases. The failure mode is sequencing them — harvest now, migrate next year, construct when certain — which converts the grammar back into waiting, and waiting worsens both clocks at once. The three-motions-in-parallel discipline comes from our self-disintermediation doctrine, which the next chapter develops in full; here it is the grammar's operating condition.

Harvest

Run the declining unit for cash, deliberately. Margin discipline. No deepening investment — you do not add seats to a stranded licence model or hire into a collapsing pyramid. An explicit wind-down horizon, held by the board rather than discovered by events. What harvest is not: denial, and equally, not vandalism. The instinct to either defend the old unit as if it were the future, or to torch it in a fit of transformation enthusiasm, both destroy the same thing — the funding source.

"The old business may still produce excellent cash. Harvesting it is not denial. It is how the transition is funded. But operational excellence buys runway. It does not choose the destination."

Note what that last pair of sentences does to Chapter 4's argument: it completes it. Productivity cannot defend the firm — but it can fund the firm's authorship. The same operational excellence that is strategically insufficient is financially essential. That is not a contradiction; it is a division of labour between the motions.

Migrate

Identify what can travel, and convert it — history, data, methods, relationships transformed into forms the new value layer can use. This is the motion the register's second lie hides: the convertible assets are invisible, so no one is assigned to move them, so they sit.

Scott's raw framing carries the stakes: "Your past and history does give you some assets. Some items can be taken into the future." And the manifesto's formulation carries the deadline:

"Unconverted history is not a moat. It is memory that eventually walks out the door."

Every quarter the archives sit unconverted, the people who can interpret them get closer to retirement, resignation or a competitor's offer — the asset's carriers are mortal even when the asset is not. What, precisely, makes an asset able to travel — what conversion actually consists of, and how to test whether a given asset can make the crossing — is its own doctrine, treated in its own right elsewhere. For this book's purpose the motion is the point: conversion is a funded allocation with an owner, or it is not happening.

Construct

Fund the successor as a real allocation: a separate unit, real capital, senior people who are measured on the new model's evidence rather than the old model's comfort — and a proof clock attached, in the last chapter's sense. The successor exists to produce an answer while the runway can still fund the next move. What construct is not: a pilot, a committee, an innovation lab whose output is decks, or a Copilot rollout wearing a transformation badge. The evidence standard that separates construction from performance gets its own chapter shortly.

The sidelines

Which leaves the posture this chapter has been circling: the board that accepts the analysis, commissions further study, and allocates nothing. It feels like neutrality. It presents as rigour — we take this seriously; we are studying it. Scott has no patience for the posture, and his raw versions built the line this book is named alongside: "Standing on the sidelines and being left right out is not really a strategy." And: "Standing on the sidelines is still a decision. AI, change and time won't wait for you."

The finished formulation deserves its full weight:

"Standing on the sidelines is still a capital-allocation decision. It allocates the future to somebody else."

This is arithmetic, not rhetoric. Capital not moved to the new layer is capital defending the old one — there is no third place for it to stand; every dollar and every senior hour is somewhere. Every quarter of "no decision" is a quarter of runway spent and proof clock unstarted. And the somebody else is not an abstraction: it is the challenger with less to defend, the customer's internal team, the two leavers with a clear view of one workflow — the attacker list from Chapter 4, all of whom are allocating while the study is being commissioned.

The grammar is complete: reclassify, then harvest, migrate and construct in parallel, with the budget telling the truth throughout. But something is still missing from it, and the gap is not mechanical. Two companies can run identical motions — one as reluctant defence, dragged through each step by its advisers, one as deliberate authorship — and they will end in different industries, because the motions were the same and the company running them was not. The grammar says what to do with capital. It does not say who you are willing to become.

That question has exactly three answers, and no company gets to skip choosing one.

"No company gets to choose whether its industry will be recombined. It chooses only the role it will play. It can become raw material. It can become a spectator. Or it can become an author."

Three roles. They deserve precise definitions, because a board should be able to classify its own current behaviour against them — and most boards, classified honestly, are not in the role they believe they are in.

Raw material. The company's assets — customers, data, trained staff, methods, even its accumulated trust — get repriced and absorbed into other people's recombinations. Its value chain is quarried. The tell is that departures start looking like supply: the talent leaves toward the recombiners, the clients drift toward the new configurations, the methods walk out in the heads of the people who built them. Nothing dramatic happens to the company itself. It becomes an input to someone else's output.

Spectator. The company watches its value proposition be unbundled around it while running an excellent version of the old model. It adopts tools — often enthusiastically. It defends existing lines. It allocates nothing to authorship. The tell is in the portfolio: every AI initiative optimises the current unit. The spectator is frequently the best-run company in its category, which is what makes the role so comfortable and so terminal — excellence at the wrong altitude, dashboards green to the end.

Author. The company treats the recombination as a design problem it is entitled to work on. It uses its inherited assets as construction material rather than fortification. It writes some of the new configuration itself.

Scott's framing puts the choice in its commercial register: "The question is whether they take a passive part in the recombining of their industry, or an active part — exploring new products, new companies, new ventures. It's about how value is being created and redistributed."

The authored path

What does authorship actually consist of? One move, stated in the manifesto with the incumbent's full arsenal attached:

"An active company asks how an AI-native attacker would rebuild the industry, then uses its capital, customers, trust, data, distribution and domain knowledge to build that attacker itself."

The doctrine underneath has a canonical statement in our board-level work: "If an AI-native competitor could destroy part of your business, build that competitor inside your own company first." The term is chosen over "cannibalisation" for a reason worth two sentences: cannibalisation implies eating your own customers. Self-disintermediation is structurally different — you are removing the intermediary your current model places between the customer and the value: the product, the channel, the pricing model, the operational layer. Ideally with the value the company captures still intact, relocated to the new configuration.

The instrument for the move is a question, asked seriously and answered structurally:

And notice who is best placed to answer it. The list of things the attacker would need — capital, customer access, trust, data, distribution, domain knowledge — is the incumbent's inventory. Chapter 4 ended its attacker list with "and it may be you," and this is where that line stops being a threat and becomes an invitation: the best-positioned constructor of your industry's successor is frequently the company reading this sentence, because the assets are already on your sheet. If they can be made to travel.

Why incumbents fail at this naturally

If authorship is available and rational, why is it rare? The answer matters because getting it wrong produces theatre. The failure is structural, not moral — and not intellectual either. It is not that incumbents cannot see the attacker's design. It is that the organisation is built so that nobody's job is to build it.

Walk the org chart: the product manager's job is to grow this product, not to ask whether it should exist. The sales leader's job is to make the number, not to question the distribution model. Every role scales yesterday's decisions — that is what an org chart is for. Organisations mechanise previous decisions; the doctrine demands deliberate divergence from plan in selected categories, and there is no box on the chart for that.

"A company cannot ask the horse division to invent the car and then be surprised when it requests a better saddle."

The horse division is not stupid. It is doing its job — keeping the horses healthy, the saddles fresh, the carts maintained. Asking it to design its own obsolescence is a category error of organisational design, and the error explains a pattern the market data now makes measurable: when incumbents respond to AI-native attackers by asking the existing product team for a competitive response, what comes back is a feature. A Copilot in the product. A chatbot on the old model. It demos well, and its fate is visible in the attach rates: Activant Capital reports fewer than 4 per cent of Salesforce customers paying for Agentforce — the incumbent's flagship AI agent product, inside its own installed base¹⁰.

Read that number for what it is rather than as a jeer. It is not evidence that incumbent AI is doomed. It is evidence that an AI feature on an old model is not a successor — and that customers can tell the difference even when the vendor cannot. A faster horse with a chatbot is still a horse. This is why Construct is a separate allocation with separate people (the last chapter's motion), not a memo to the existing roadmap.

One more structural honesty, briefly, because boards will meet it: shareholders compound the problem. Continuity is what the register rewards — twelve per cent on last year, not a deliberate dip while value migrates. The framing that survives shareholder contact is the truthful one: we are migrating the value this company creates — not "we are disrupting ourselves," which frightens capital, and not silence, which forfeits the story to the attackers.

Incumbency, conditional

All of which sharpens this book's thesis to its final form. The re-roll did not abolish the incumbent's advantages; Chapter 1 established that some inherited assets appreciate. But the advantages have a condition attached now, and the condition is the whole game:

"Incumbency is not automatically an advantage. It becomes an advantage only when the assets of the incumbent are attached to the future value layer rather than used to defend the old one. History matters only when it can travel."

Hold the two sheets side by side one more time. The incumbent holds capital, customers, trust, data, distribution, domain knowledge — every input authorship needs. The challenger's only structural advantage is having nothing to defend, which buys it mutation speed. If the incumbent authors, the challenger's one advantage evaporates: mobility means little against a player with the same freedom and ten times the inventory. If the incumbent defends, its inventory becomes the attacker's map — Chapter 1's sixth fate, now with a mechanism under it.

The three build orders

The authored path resolves into instructions — the manifesto's most direct passage, and the point where this book's argument becomes a to-do list:

Each order attacks a different layer of the incumbent's inertia. The first attacks the product's necessity, not its quality — the question is not whether customers like it but whether the need it answers can now be met without it. The second attacks the margin structure, not the revenue line — the successor's economics should be the ones an attacker would choose, which means they will look wrong by the old model's metrics. The third attacks the incentive map — the business model nobody inside proposes because every existing bonus, target and career path points away from it; this is the one the horse division can never see. And the fourth line is not an order but a clock: the window in which authorship is fundable is the runway, and the runway is being spent either way.

One caution closes the chapter, because it opens the next. Three build orders can be performed as theatre too. A venture lab with a lease and a logo. A press release about the future. A pilot that tours the conference circuit. The difference between construction and performance is not visible in the activity — both look busy, both produce artefacts, both photograph well. What separates them is a definition of success that cannot be gamed.

That definition exists. It has a derivative in it.

Every vendor deck now claims it. "AI-native" attaches to chat features, to model wrappers, to cloud migrations with a copilot bolted on, to anything shipped after 2023. The label has been claimed by so many things that it describes nothing — and a term that can describe anything constrains nothing. A board funding a successor under the last chapter's build orders needs something stricter: a definition that can fail. A definition you cannot claim your way into. A definition with a derivative in it.

The definition

Read it clause by clause, because each one excludes a familiar pretender. "Reorganised its assets" — the last two chapters' work actually executed: compounding assets built, convertible assets converted, stranded assets in managed decline. A company with an unreclassified register has not reorganised anything, whatever its tooling. "And commercial model" — the unit of sale moved off the compressible layer, so that what the customer pays for is no longer the thing AI makes free. This clause is where most claimants fail quietly: the tooling is new, the invoice is old. "Advances in AI increase its enterprise value" — the sign of the derivative. For the reorganised company, model releases arrive as dividends; for the legacy company, as threats.

The definition generalises what we first established for professional services as the sign flip — a firm is AI-native when improving AI increases the value of its productive assets rather than increasing its exposure. The re-roll makes it the general definition for any company, because the mechanism was never about services: it is about which side of the repricing your assets sit on. The positive sign even has a name in our work — the Model Dividend: a casual user gets a better chatbot out of each model release, while an organisation with compiled context, delivery machinery and evidence systems gets an uplift to its entire production system.

The test

Walk both truthful answers. If the honest reading is exposure up — the release lets customers replace more of what you sell, makes more of your margin contestable, commoditises more of what differentiated you — then you are operating the legacy model, regardless of your AI budget, your council, your adoption awards. If the honest reading is stronger — your proprietary context just became more useful, your delivery machinery just improved, your evidence systems just got more powerful, your products just became more valuable — the sign has flipped. The same force that is crushing the legacy configuration is compounding yours.

"That derivative cannot be performed."

This is why the definition is built on it. Adoption metrics can be staged. AI councils convene. Copilot seats get bought. Even revenue-from-AI can be theatre — a relabelled line item, a bundled feature. But the derivative is a fact about your asset structure, and a model release computes it for you, on its own schedule, without asking permission. Every frontier release is an involuntary audit of every company in the economy. The only choice is whether anyone in the room reads the result.

A definition this compressed needs to be seen running before a board can use it. So run it — twice, on two companies chosen to be as unlike each other as possible, because the test's claim to generality rests on surviving the contrast. Both are composites, drawn structurally from our industry work; no named companies, no invented figures — directions, not magnitudes.

The test, walked: a mid-market workflow SaaS vendor

The composite: a workflow-software vendor, per-seat licence economics, deep custom configuration as the historical moat, a decade of integration history, a customer-success operation that walks customers through implementation.

The legacy sign, at a release. A frontier model ships with better code generation. Follow the behaviour, step by step: more customers regenerate internal tools instead of renewing — the build-versus-buy behaviour Chapter 2 measured, now a little cheaper again. Renewal conversations arrive with smaller seat counts. The configuration depth that was a moat converts, customer by customer, into switching-cost resentment — a map, in Chapter 1's sixth fate. Margin contests arrive as procurement events, not as competitive losses. Meanwhile — and this is the part the board room feels as paradox — the vendor's own adoption dashboard can be green the whole time: its Copilot features ship, its internal productivity rises. Δ capability up, exposure up. Both curves real; different exponents.

The flipped sign, at the same release. Now the reorganised version of the same company — the one that ran the last two chapters. It converted its integration history into interface specifications, its configuration playbooks into workflow patterns, its product into an outcome-priced generation-and-governance engine: customers regenerate their stacks with the vendor's patterns and governance baked in, paying for outcomes rather than seats. The same release now lands differently: the generation engine generates better, so the product improves overnight without an engineering sprint. The patterns apply more widely. The governance layer becomes more valuable, because every customer's generated stack multiplies the surface that needs governing. Δ capability up, productive asset value up. Same company, same release, opposite sign — the difference is entirely in what got reorganised.

The test, walked: an asset-heavy operator

The second composite is deliberately as far from software as the economy goes: an industrial equipment distributor and service operator. Fleet and inventory, workshops, licensed technicians, decades of service history, a quoting-and-advisory desk that customers ring before they buy.

The legacy sign. A release improves technical question-answering and scheduling. Customers self-serve the advisory layer — spec answers, model comparisons, first-pass fault diagnosis — before they ever ring the desk. The quoting desk's margin compresses; the advisory relationship that fed the sales funnel thins. The firm's differentiation collapses toward "we own the assets" — true, but the cognitive shell around the assets, the utilisation planning and pricing and service judgement that the organisation was actually paid to exercise, is being repriced toward zero. The physical asset is left carrying the overhead of an organisation that used to be paid to think around it. Δ capability up, exposure up — and not a line of software in the business model.

The flipped sign. The reorganised operator made its physical capacity legible to the new value layer: machine state, configuration, service history and operating constraints held in machine-usable form, with explicit authority over who may do what to which asset. Its commercial model moved from margin-on-transactions toward outcomes — uptime, availability, throughput. Now the same release lands as a dividend: coordination gets cheaper and better, so the fleet serves more customers at higher utilisation; the service promise gets stronger, because the thinking around the assets is now machinery that improves with every model. And the scarcity anchors — the physical capacity itself, the licences, the consequence-bearing service authority — appreciate, because abundant cognition makes genuinely scarce things more valuable, not less. How a physical asset acquires that machine-legible counterpart is its own conversion doctrine, treated elsewhere; the point here is the sign. Δ capability up, productive asset value up.

One sentence generalises both walks, and it is the chapter's yield: the test is substrate-independent. Software firm or steel firm, the derivative reads the same two things — what the commercial unit is denominated in, and whether the assets are organised to receive releases as dividends.

Proof appears in behaviour

The derivative test tells you which way a company faces. For the successor being constructed under the last chapter's build orders, there is a companion standard — what counts as evidence that the construction is working. The manifesto's list is exact:

"Activity is not evidence. A polished future is not a proved future."

Those two sentences are the standard the rest of this book answers to. Part III is complete: a posture (author), a grammar (harvest, migrate, construct), and a test (the derivative, with behavioural proof behind it). What remains is the part most strategy books skip — showing the whole apparatus actually run. The manifesto's closing instruction points straight at it:

"Build options. Put them into the world. Let reality eliminate the weak ones. Allocate serious capital when evidence earns it."

Doctrine that has never run on a real industry is still theory. The next chapter runs it — four character sheets, one industry, end to end.

Two workshops. Same room, same people, same week, same client.

Workshop A opened with the standard question — how can AI help our business? — and three hours later the wall was covered in Post-its grouped by department: operations, marketing, sales, finance, HR. Sixty ideas. The CEO later picked six. None of the six would move the needle six months out, because sixty ideas grouped by department is what the question was shaped to produce.

Workshop B opened differently:

"Our business depends on at least three scarcities — competent advice, internal coordination cost, and trust. Pick one. Assume AI makes that scarcity near-zero within three years. Walk us through what breaks first, where the value migrates, and what we'd build."

Two hours later, Workshop B had produced one structural risk, one value-migration map, and three capital-allocation decisions.

"Workshop A produced sixty ideas. Workshop B produced three decisions. The difference is the shape of the question."

This chapter runs Workshop B's question at full length — not on a function but on an industry, and not on one character sheet but on all four seats at the table. Everything the previous nine chapters built gets used, and nothing gets re-explained. If a term feels load-bearing, it was defined earlier; this chapter only derives.

The industry is mid-tier management consulting. The incumbent is a composite assembled from our industry work: roughly eighty partners, about nine hundred staff, pyramid leverage around seven to one, revenue in the region of $320 million; a mix of strategy advisory, transformation programmes and managed-service consulting; differentiation resting on brand permission with mid-market boards, sector frameworks that live mostly in partners' heads, and relationships built over decades. One trait is added for this book's purpose, and declared openly: the incumbent is drawn as a sector AI-adoption leader — Copilot rollout complete, internal tooling genuinely good, an adoption dashboard the board is right to be proud of. That trait is what makes the walk a test of Chapter 4's paradox and not just an exercise. Why consulting? Because every reader has bought or sold advice, so every reader can check the derivation against experience — and because the boundary case it needs is the era's most credible: what if competent generic advice becomes free?

The four sheets, written out

Sheet 1 — the incumbent

The firm's best assets survived the re-roll. Its business model did not.

Sheet 2 — the challenger

The rival firm's sheet is the same class of document: smaller brand, shallower relationships, the same stranded pyramid. One line differs, and it is the line that matters. With fewer partners living off the old unit, the challenger has less to defend — which buys it mutation rate. It can re-denominate two service lines while the incumbent's pricing committee is still meeting. The challenger's sheet is not better. It is more mobile. And — worth saying ahead of the motions — if the incumbent chooses authorship, that one advantage evaporates: mobility means little against a player with the same freedom and ten times the inventory.

Sheet 3 — the customer

The sheet that changed most is the one nobody at the old table thought of as a player. The mid-market client CFO's sheet now carries items that did not exist two years ago: an internal first-pass capability — market scans, competitor profiles, first-cut models produced in-house in days; AI-assisted evaluation of advice — the firm's deliverables get machine-read, benchmarked and challenged before the partner's follow-up call; and an opinion, formed before the engagement conversation starts, about scope, duration and price. The demand side of this exact industry repriced first — Chapter 4 carried the measured evidence. What it means at the table: the customer stopped being the demand backdrop and became an actor whose sheet reprices the suppliers' sheets.

Sheet 4 — the constructor

Two probable shapes, both short sheets. Shape one: two senior practitioners who leave with a compiled method, a client list's worth of trust, and no pyramid to feed. Shape two: the client's own strategy function, armed with the tooling and increasingly confident after each internal first pass. Both run the same boundary cases our method generates — "What if competent generic advice becomes free?" and "What if a well-funded AI-native startup rebuilt our industry from scratch with no legacy cost base?" The brevity of the constructor's sheet is the sheet: no stranded assets, no incentive map defending the old unit, total mutation freedom. And its weaknesses are equally legible: no brand permission, no relationships, no licensed authority — the scarcity anchors it must rent, borrow or wait for. Which is exactly the incumbent's authorship window, if the incumbent moves inside it.

The paradox, worked

Now put Sheet 1's adoption programme against Sheets 2, 3 and 4, and run time forward — as ordering and direction, not invented figures, because the shape is the proof and the numbers would be decoration.

The inside curve. Season one: rollout waves complete, drafting accelerates. Season two: delivery margins improve; the dashboard's usage numbers climb. Season three: an internal award, a case study, genuine cost-per-deliverable gains. Every gain real. Every gain sequential — each wave waits on the one before, because that is what a change programme is.

The outside curve, same seasons, in parallel. The client CFO's team does its first internal market scan and likes it. The challenger re-prices two service lines away from hourly billing. The two leavers sign their first client — a name from the incumbent's own conference table. Another client's board asks, for the first time, why the proposal costs what it costs, attaching a document the incumbent did not write. No programme. No gates. No coordination. Each actor made one move — and there are hundreds of actors.

Where the gap shows first. Not in the win/loss analysis — nothing was formally lost. It shows first in pricing conversations: pushback that reads as commercial weather. Then in scope: engagements shrink to the judgement layer as the first-pass layer quietly evaporates from statements of work. Then — the tell nobody briefs the board on — in the graduate intake question: the pyramid's feeder logic stops making sense, and the firm discovers its business model has changed in a recruitment meeting, quarters before it appears in a strategy paper.

The firm led its sector in adoption. And its cognition compounded inside the walls at the speed of its change programme, while everyone else's compounded across the market at once. Both numbers were true the whole time.

The migration map, derived

With four sheets repriced, the map is not asserted — it falls out. Trace each dissolving scarcity to where its value re-accrues:

The compression of the whole map into one sentence comes from our own analysis of this industry: two ends of the pyramid survive; the middle collapses. And Chapter 2's inversion check lands precisely: the firm owned the entire advisory chain, and the chain's most valuable link turns out to be the one it never itemised — adjudicated judgement with evidence attached.

The motions, staged

The map converts to allocations through the capital grammar. Written as board resolutions for this firm — with each proof-clock trigger stated as structure (what evidence releases what next allocation), never as dates or dollars, because those are the board's to set:

And the counterfactual, because the walk owes the reader the other branch. The same firm, allocating nothing: harvest happens anyway, unmanaged — the pyramid thins by attrition while the pricing erodes. Migrate never starts — the archives sit, and the methodology resigns partner by partner. Construct never starts — and in three seasons the firm discovers it has become the industry's supply: its leavers staff the constructors, its alumni run the clients' internal teams, its methods circulate in other people's engines. Raw material — the first role from Chapter 8, arrived at not through any decision but through the absence of one. Standing on the sidelines allocated the future to somebody else, exactly as promised.

The Re-Roll Sheet

Everything above was one industry. The method is portable, and this is the book's artefact — the one-page protocol a board can run next week:

One paragraph of honesty seals the walk. Nothing in this chapter required the author's venture, a vendor, or a tool. Four sheets, a boundary case, a conservation law, three motions, one test — all of it runnable by your board, with your industry knowledge, on one page. The doctrine survives with its author's commercial interests deleted. That is the standard a doctrine should meet before its author is allowed to show you his own use of it.

Which raises the fair question, and the last one this book owes an answer to: does anyone actually run capital this way — on purpose, with their own money, as the operating model rather than the emergency response?

Does anyone actually run capital this way — on purpose, with their own money, as the operating model rather than the emergency response?

Yes. And the specimen I can vouch for personally is my own. SongbirdX — the venture engine my co-founder and I run — builds AI-native companies on the working assumption that everything in this book is true: that the sheets are re-rolled, that the unit of change is the industry, and that value is moving fast enough to be caught only by whoever is already in motion. What follows is shown, not sold: a specimen is evidence that the doctrine is operable, not proof that it is optimal — and this chapter holds itself to the same standard the last one set, which is why the previous chapter ran entirely without it.

What the engine reads

The engine works at the altitude Part I established — where industries are being unbundled, repriced and recombined. Its raw material is a reading, taken per industry, of five things:

The reading tells us what to build, what to rebuild and where capital should move next. Then the engine runs its sequence: Thesis. Model. Brand. Product. Market. Five stages, each closing one question — what should exist; how it captures value; how it is recognised; what actually contacts the customer; whether anyone pays. Nothing in the sequence is exotic. What is unusual is the clock attached to it.

Thirty days

The engine turns a thesis into a live venture in thirty days. Read wrongly, that is bravado — the kind of number that decorates accelerator landing pages. Read rightly, it is a clock discipline, and the two readings are distinguished by what the thirty days is claimed to contain.

"Not because a complete company can be declared finished in 30 days. Because the first coherent venture must make contact with reality before the insight that created it goes cold and before the market branches again. Thirty days is the first proof cycle."

"Thirty days is not the time needed to invent all the underlying thought. Much of the cognition has already been capitalised. Thirty days is the compile and contact-with-reality period. The old time was spent over decades. The new time is spent invoking, recombining, building and testing it."

That is the honest mechanics of the number, and it explains why the engine is fast without being reckless: the thinking is old. Decades of frameworks, methods and judgement already exist in usable form; the thirty days spends none of it on invention and all of it on contact. And note what the clock is calibrated against — the two decay rates this book named in Part II: the insight's freshness, and the market's branching. The engine holds no legacy model, so only one clock runs — no runway to manage, only a proof clock. Which is exactly why an incumbent cannot copy the number. It can only copy the discipline.

What a proof cycle must produce

The fourth output is the signature. Most venture processes — and nearly all corporate innovation processes — are engineered so that the answer cannot disappoint: success criteria soft enough to always pass, feedback gathered from audiences selected to be kind. The engine inverts that. Each cycle ends with evidence that tells us whether to stop, reshape, or allocate more capital — the elimination loop from Part I, running on a thirty-day period. Activity is not evidence; the cycle is built so it cannot be mistaken for it.

What an incumbent borrows from this

The specimen is a venture engine; most readers run companies. What transfers is not the number — it is four properties of the discipline:

1. The proof cycle has a fixed, short period — set by insight-decay and market-branching, not by comfort or by the quarterly calendar. The right period for a construct unit inside a company is context-dependent; the wrong period is always "when the study completes."

2. Every cycle ends at contact with reality — the four outputs above, scaled to the unit's context. An internal milestone is not contact; a customer who can refuse is.

3. The answer is allowed to disappoint — and disappointment is a funded outcome. A cycle that kills a weak option has retired a possibility at its cheapest price. The elimination is the product.

4. Capital follows evidence, not narrative. The behavioural gates release allocations; polish releases nothing. This is the last chapter's Resolution 3, running as a standing rhythm instead of a special event.

Install those four properties in a construct unit and you have the specimen's engine without its context — which is all the borrowing the doctrine requires.

What the specimen claims, and what it does not

"We do not claim to know the final form of the future. We build a system for participating in its formation. This is not innovation theatre. It is capital allocation made executable."

Both halves matter. Not claiming the final form: the engine is a search procedure, not a prophecy. It expects to be wrong often, cheaply, on schedule — it wins by cycle count and honest elimination, not by being right first time. And capital allocation made executable: the whole arc of this book — the budget as the strategy document, the motions, the gates — compressed into an operating loop where the allocation decision recurs every thirty days with fresh evidence in the room.

The specimen's own summary line is where this chapter ends, because it is the doctrine in the first person:

"We see where value is moving. Then we build into it."

The proof part of this book is now complete: a doctrine derived on an industry that could be anyone's, and a doctrine operated in a venture that happens to be ours. What remains is not more argument. It is the form a doctrine takes when a company decides to be bound by it — commitments, stated in the first person, with the hedges removed.

Doctrines bind behaviour or they are moods.

A board can agree with every chapter of this book and change nothing — agreement is the cheapest asset in the building, and it photographs well in the minutes. Which is why this book does not end with a summary. It ends the way the doctrine was written: in first-person commitments — what we will not do, and what we will. The "we" is an offer. Any board may adopt it, and the adoption is measured the only way this book measures anything: in what the budget and the calendar do next.

What we will not do

What we will do

The standing agenda item

A board that adopts the declaration needs one recurring instrument, and it already has it — asked in writing, answered in writing, re-run at every frontier release alongside the derivative test, on the worksheet the proof chapter supplied:

The choice

The game did not pause while this book was read. The table from the first chapter is in session now: sheets re-rolled, modifiers new, positions being discovered move by move — by your customers, your competitors, your constructors, and the players who have not sat down yet. The only unrolled thing left at the table is the response.

The character sheet is on the table. The pen is in your hand.