Part 1The Reset
The world has reset.
Four forces are reshaping the global operating system simultaneously. Each, on its own, would be a once-in-a-generation structural shift. Together, they create a new world that breaks apart the frameworks and models we have used to operate in it.
The geopolitical reset
We are moving from a globalised world to a distributed world.
For three decades, the operating premise was open. Capital, goods, technology, talent, and data moved across borders with progressively fewer restrictions, and every multinational strategy of the post-Cold-War era assumed that this openness would continue. That premise is over. The same flows still cross borders, but now they cross with friction and with political conditionality. Every major economy is weaponising what it controls and fortifying what it depends on: financial systems, semiconductor supply, energy, food, rare earths, digital infrastructure. Trade is sorting into blocs. Alliances are reorganising around strategic interest rather than ideology. The category we once called the "neutral global corporation" — a firm that could operate everywhere without taking sides — is becoming a historical category.
Geography, sovereignty, and political alignment are once again first-order constraints on every strategic decision.
The technological reset
For the first time, intelligence is being manufactured.
Of the four resets, this is the most consequential — and the one most institutions are still misreading. Previous technology waves — mainframes, personal computing, the internet, mobile, cloud — built better tools for human minds to work with. Artificial intelligence is different in kind. It produces intelligence directly, as a renewable, scalable, on-demand resource flowing from data centres at industrial scale.
The cognitive capacity available to civilisation is rising in ways without historical precedent. Problems that required decades of accumulated expertise are now being solved in hours or days. Capabilities that justified entire layers of the global services economy are being produced, at the marginal cost of the systems that produce them, in a way the prevailing economic models did not anticipate and cannot absorb without being substantially rewritten.
This is not a productivity wave. It is a substrate change, and it challenges every institution to confront a critical question: is the value proposition on which this business was built still valid in the world that is now arriving?
The economic and intellectual foundations on which the last two centuries of progress rested are being rewritten — and at a pace that exceeds any prior technology transition.
The financial reset
The financial architecture of the open world is no longer safe to operate on.
Inflation has shifted from a temporary disruption to a structural feature of the global economy. Sovereign debt trajectories in the major reserve issuers — including the United States — have crossed thresholds that change the calculus of every fixed-income allocation. Interest rate regimes that had compressed for decades are normalising and oscillating in ways that re-price every asset class. Forex volatility is moving at speeds and amplitudes not seen in a generation. The dollar-denominated assumption that has underwritten global finance since Bretton Woods is itself being contested.
The causes are structural, not cyclical. Inflation is structural because supply chains are being deliberately rebuilt from efficiency to resilience — and resilience is inherently more expensive. Sovereign debt is under structural pressure because the major reserve issuers have crossed debt-to-GDP levels at which the easy monetary lever — quantitative easing without consequence — is no longer available. And the geopolitical reset is bleeding directly into the financial reset: when reserves are weaponised, the holders of those reserves diversify away, which structurally reduces demand for the issuer's sovereign debt, which changes yields, which re-prices every asset class globally.
What the old system absorbed quietly, the new one transmits violently.
The risk reset
The character of risk in the world has fundamentally changed.
The statistical distributions that risk frameworks rested on — relatively stable correlations, bounded volatility, infrequent extreme events — are no longer descriptive of the world's actual behaviour. Black swan events that classical models priced as once-in-a-century are now occurring multiple times per decade. The variance of outcomes has widened in both directions. More structurally, the world has moved from a stationary statistical regime, in which the past is a reasonable guide to the future, to a non-stationary regime, in which the distribution itself shifts faster than risk models can recalibrate. The challenge is no longer to manage risk within a known distribution. The challenge is to operate when the distribution itself is unstable.
This is not a matter of more risk. It is risk that behaves differently in kind.
The cascade
Each of these four shifts would have been consequential on its own. The reason this moment is unlike any in living memory is that they are running simultaneously — and, more importantly, they are cascading into one another. A move in any one dimension is triggering moves in the others, in a system-level feedback loop that the strategic frameworks of the departing era were never designed to absorb.
Four specific cascades are worth naming.
The geopolitical reset is driving the financial reset: reserve weaponisation has forced central banks to diversify away from the dollar, which is restructuring sovereign debt demand, yields, and asset prices globally.
The technological reset is driving the geopolitical reset: the race for sovereign AI has made compute, models, and data into strategic resources, which is deepening the trade-blocs sorting.
The financial and technological resets together are reshaping the risk regime: structural inflation plus AI-driven labour displacement creates a world in which neither monetary policy nor expected income trajectories can be reliably modelled, which changes what risk fundamentally is.
And the risk reset feeds back into all three: when risk becomes non-stationary, institutions, investors and individuals all pull back into resilience, which deepens supply chain fragmentation, which deepens geopolitical sorting, which deepens financial re-pricing.
This is the central insight I have been working with for two years, and it is the foundation of everything I am building. The reset is not an event the institution can wait out. It is not a phase that will end. It is the new operating condition. Inside that condition, the strategic frameworks of the era we have just exited — set a destination, execute toward it, refresh every three to five years — are no longer fit for purpose. What is required instead is a discipline most institutions have not yet built: the capacity to continuously read the world and adapt to it. Continuously. Not as a one-time transformation programme, but as an ongoing institutional competence.
And inside that continuous discipline, the most consequential work is the work of deliberate discrimination — telling apart three things, continuously:
What in this institution is durable — what should be preserved because it is what makes us who we are, what we are uniquely good at, what carries forward through any future?
What was built for a world that is now gone — what should be evolved, or reconstructed from first principles, because the conditions that made it work no longer hold?
And what has no place in what is coming — what should be deliberately transitioned out of, even while it still produces revenue, because holding on to it will eventually cost more than letting it go?
The discipline is in telling these apart — continuously, deliberately, and with the conviction to act on the discrimination as the world keeps moving.
The moment
While the institution navigates this transition itself, it has to recognise that its customers, its suppliers, and its employees are equally inside the same reset — pursuing their own journeys with the same uncertainty, the same anxiety, the same need to find footing in a world whose rules are no longer the rules they trained for. The work is to support them as they navigate their own version of this transition, while the institution navigates its own.
Don't mistake the reset for a threat to be handled. It is the greatest opportunity ever available to us. The reset is creating a level playing field in ways that have not existed in living memory. The advantages an established institution accumulated over decades — scale, distribution, brand, capital, expertise — are being repriced by the four resets at the same time as new entrants are gaining access to capabilities that were unimaginable a few years ago. The advantage an experienced professional built over twenty years is being repriced by AI, and a new hire can access the same capabilities and more in months. The advantage a capital-rich company held over a capital-scarce one is being repriced as new financing structures and AI-native business models are making capital itself less of a moat than it was.
The institutions and individuals that embrace this reset will have a head start and will compound generational advantage over the next decade.
Let me now look closely at the reset that is most on everyone's mind — and, as I have said, the most consequential of the four: how AI is reshaping the world, and the institutions inside it.
Part 2AI and the Architecture Problem
Most institutions are getting AI wrong — not because they are moving too slowly, but because they have misunderstood what kind of opportunity it actually is.
I argued earlier that the ground beneath our institutions has shifted, and that the technological reset forces a question most boards have not yet asked out loud: is the value proposition on which this business was built still valid? Artificial intelligence is where that question stops being abstract. It is also where, right now, the gap between activity and outcome is widest.
The AI buzz is everywhere. Almost every institution I speak with talks about AI, but few have turned it into real value for the business. The work is being done; the outcomes are not arriving. And the reason is structural, not technical.
Three ways institutions are engaging AI
When I look across the organisations I encounter, AI engagement falls into three broad patterns.
The first treats AI as a conversation. The board discusses it, leaders read about it, and there is broad agreement that it matters — but the engagement stays at the level of awareness. It is about AI in the abstract rather than what AI does to this business. Most institutions are here.
The second treats AI as a set of technology-led pilots. Functions are encouraged to experiment, the CTO sponsors initiatives, and teams automate existing processes within their own walls. The work is real, but its ceiling is low — a process gets faster, a chatbot gets deployed, and the organisation calls that AI adoption. Some institutions have reached here.
The third treats AI as a strategic question owned at the board and executive level — a question about the business itself, not its tooling. This is where the value is. It is also, in my experience, rare.
The distance between the second pattern and the third is the distance between activity and advantage. And the evidence that most institutions have not crossed it is now hard to dismiss.
The evidence is no longer anecdotal
In July 2025, MIT's Project NANDA published The GenAI Divide: State of AI in Business 2025. Drawing on 150 executive interviews, a survey of 350 employees, and analysis of 300 public deployments, it found that despite an estimated $30–40 billion of enterprise investment, roughly 95% of generative-AI initiatives were delivering no measurable impact to the bottom line. The researchers were explicit that the failure was not one of model quality — it sat in integration. High adoption, low transformation. The tools were being used; the businesses were not being changed.
The most instructive case is Klarna, because it is the complete arc in a single story. Between 2022 and 2024 the company replaced roughly 700 customer-service roles with an AI assistant, reported around $40 million in projected annual profit improvement — about $10 million of it in direct savings — and announced that within a month of launch the system was handling some 2.3 million conversations, roughly two-thirds of its customer chats. For a while the numbers looked like vindication. Then, by mid-2025, customer satisfaction had fallen, service quality had become inconsistent, and the company began rehiring human agents — its chief executive acknowledging that the firm had weighted cost too heavily and underestimated what human service protected. The point is not that Klarna was reckless; it moved earlier and more boldly than most. The point is what the arc reveals: a short-term operational win that became a long-term strategic loss. The savings were modelled; the cost of unwinding the strategy was not.
If Klarna is the strategic-loss case, the cost case is arriving in real time. Through late 2025 and into 2026, several of the most technically sophisticated organisations in the world found their own AI spend running ahead of any business return — not because adoption was too low, but because it was high and unstructured. Microsoft moved to wind down most of its internal advanced-coding-AI licenses after usage-based costs consumed the budget within months of rollout. Uber, by its own chief technology officer's account, burned through its entire 2026 AI coding budget — some $3.4 billion — in four months, as per-engineer spend climbed to between $500 and $2,000 a month across roughly 5,000 engineers: "the budget I thought I would need is blown away already." Commentators began describing a moment in which, for some tasks, using the technology was costing more than the people it was meant to make more productive. These are not stories of institutions that were timid. They are stories of what happens when a powerful tool is poured into the organisation without an architecture to govern where the value is supposed to come from — spend compounds quickly, and outcome does not follow.
Two failure modes recur often enough to name. The first is technology-first reversal — beginning with what the technology can do and working backwards toward a use case, rather than beginning with the business and asking what it now needs. The second is AI bolted onto legacy — layering new capability onto operating models built for a different era, so the capability is throttled by the very architecture it was meant to transform. Klarna is, in effect, both at once; the runaway-cost cases are the first one at scale.
And the concerns leaders themselves raise map onto this precisely. Asked what holds them back, executives are consistent: skills are the first constraint, not technology; there is genuine unease about scaling without control; there is real uncertainty about how much of the current investment is durable and how much is a bubble; and there is concern about dependence on a small number of providers. These are not the worries of people who lack ambition. They are the worries of people who can feel that something in the approach is off, without yet having the frame to name it.
I find this clarifying rather than discouraging. The diagnosis is not mine to invent — it is the conversation institutions are already having with themselves. What is missing is the frame that makes sense of it.
The Frame
AI is being treated as a deployment problem when it is, in fact, a business-architecture problem.
A deployment problem asks: how do we get this technology into the organisation? A business-architecture problem asks: given what this technology now makes possible, what should this business become? The first question produces pilots, runaway spend, and the 95%. The second produces advantage. Almost everyone is answering the first.
The difference is visible in how far an institution has actually moved, and there are three levels — only the higher ones representing architecture-thinking rather than deployment-thinking.
AI is applied to existing processes and functions — a faster workflow, an automated task, productivity inside a single team. This is the deployment level, and it is where the great majority sit. It is not nothing, but it changes nothing structural; the business that emerges is the same business, slightly faster. Most of the 95% are stuck precisely here, mistaking motion for transformation.
AI is used to re-engineer how work is actually done — not speeding up a process but rebuilding it. JPMorgan is instructive at the boundary: it now runs AI across hundreds of use cases, most of them first-level accelerations, but some genuinely re-engineering how analysts work. AIG is the cleaner case — it has reported compressing parts of its underwriting cycle from weeks to days, which is a redesign of the work, not an acceleration of it. This level is real, and most of the better stories live here. But it is still optimising the existing business rather than reconceiving it.
The rare one. AI is used to build something new. In May 2026, PwC launched its Office of the CFO as the first standalone business unit anchored entirely on a single AI platform, describing it plainly: where many are running pilots, we are running production. It is not a tool dropped into the existing firm; it is a new, AI-native unit conceived from first principles and standing alongside the old one. Very few institutions have done anything comparable — which is exactly why the advantage there is uncontested.
That progression — process, re-engineering, new build — is the real map. The first level is deployment-thinking; only the second and third are architecture-thinking, and the advantage compounds as you climb. The gap the evidence keeps exposing is now plain: almost all of the activity sits at the first level, some of the better work reaches the second, and the advantage lives at the third, where almost no one is.
What this asks of the institution
If the diagnosis is right, then the most consequential decision an institution makes about AI is not which tools to buy or which pilots to run. It is whether the question gets owned at the level where business architecture is actually decided — the board and the executive team — or whether it is delegated down to be solved as a technology rollout. That single choice separates the institutions that will climb from the ones that will spend.
That is the work I have spent this period building — frameworks and models to help institutions navigate these gaps and move deliberately up the levels. More about it in the parts that follow. If there is one point to carry out of this section, it is this: the failure modes institutions are hitting in AI are not random or unlucky — they are the precise and predictable result of treating an architectural opportunity as a deployment one. Which means they are addressable, by anyone willing to ask the harder question first.
The institutions that compound advantage over the next decade will not be the ones deploying AI fastest. They will be the ones that understood, early, what kind of opportunity it actually was — and reorganised themselves around the answer.
Part 3Leaders and People in the Reset
The leaders and people who built the institutions of today are not, in general, the leaders and people those institutions will require tomorrow — and the gap is one of kind, not degree.
The reset does not only change what an institution does; it changes what it asks of everyone inside it. The competencies that earned a leader their position — deep functional expertise, authority drawn from experience, fluency in a playbook that worked — are precisely the things the new environment is revaluing. Authority that rested on having seen it before means less in a world no one has seen before. Functional mastery means less in an organisation reorganising from silos into networks. The leader of tomorrow owns a whole rather than a part, earns credibility through present capability rather than tenure, and treats continuous change not as a programme to be managed but as the condition in which they operate. This is not a matter of topping up skills. For many, it asks them to unlearn the very things that made them successful — which is among the hardest things a capable person can be asked to do.
What is true of leaders is true, in its own form, all the way down. Every person in the institution faces a version of the same journey: the work itself is being redefined, the definition of "the best people" is shifting beneath them, and the path from who they are today to who the institution will need is neither short nor certain. It is a genuine transition — and like every real transition, it carries real cost. People are being asked to commit to a destination that keeps moving, while no one, including their leaders, can promise it will arrive where intended. Some roles will carry forward transformed; some will not. To pretend otherwise is not kindness; it is the absence of leadership.
And this is where the onus falls, squarely and unavoidably, on the Executive team and the Board. The transition of an institution's people cannot be delegated to a function, run as a change programme with a start and an end date, or reduced to a line in the training budget. It is the continuous, human work of carrying an organisation's leaders and its workforce across a divide — supporting them through uncertainty that the leadership itself shares, making hard decisions honestly rather than hiding behind process, and holding the institution's trust intact while almost everything else is in motion. An institution can get its strategy, its technology, and its capital allocation right and still fail the reset entirely if its leaders and its people are left to cross this divide alone. The Board and the Executive team are not spectators to this journey. They are responsible for it.
I have developed considerable thinking on the specific competencies the new leadership demands, the journey leaders and people must travel, and the cultural work this transition requires of an institution. It is more than this letter should carry; I will take it up in the letters that follow. The point to hold here is simpler — that the reset is, in the end, a human transition as much as a strategic one, and the institutions that treat it that way from the top will be the ones whose people carry them through.
Part 4The Response
"The world as we have created it is a process of our thinking. It cannot be changed without changing our thinking." — Albert Einstein
With the world reset, it is critical that we change our thinking. I have conceived The R Doctrine specifically with that objective — it is the thinking framework and the models for the new world, and it is continuously evolving. It is built on the recognition at the heart of this letter: that the reset is not an event to wait out but the new operating condition, one that demands the discipline to continuously read the world and adapt to it, and within that the harder discipline of deliberate discrimination — what to preserve, what to rebuild, and what to let go. The R Doctrine is what turns that recognition from instinct into a repeatable institutional capability.
It has grown into five connected bodies of work — one discipline at different apertures, wide when a whole institution is in view, narrow when a single question is:
The full-aperture transformation of an institution across every dimension on which it is being remade, from its business model to its culture, holding the whole rather than fixing the parts.
The response to the failure described earlier: a methodology for moving an institution from treating AI as a deployment problem to treating it as the business-architecture opportunity it is.
Built backward from what the reset-state organisation must be able to do, to the competencies that demands of its leaders: the shift from owning a part to owning a whole, from authority earned through experience to authority earned through present capability.
The human transition itself, treated not as a change programme with an end date but as continuous work: reskilling, the shifting definition of who an institution's best people are, and the holding of trust while everything else moves — including the support structures an institution must extend to its people as they navigate their own journeys.
The same discipline turned from the institution to the individual — because the rules most of us were taught for protecting and growing wealth were written for a world of stable correlations and calm compounding that has reset along with everything else.
Rather than deep-dive on each of these, let me choose, for this letter, to elaborate on The AI Doctrine — given the level of interest in AI in the world today.
The AI Doctrine begins from the claim I made earlier: that AI failure is not a technology failure but an architecture failure — the consequence of handing the most powerful technology in industrial history to the part of the institution least positioned to answer what the institution should become. The Doctrine is built to correct that, and it runs as four stages in a continuous loop.
The discipline here is that the AI conversation should begin where it matters most: at the Board and the Executive team. Before a single tool is chosen, before a single pilot is run, the Board and the Executive team are engaged to build a shared understanding of AI and its possibilities, and are taken through structured frameworks to define the targeted business outcomes — along with the number of tracks they want to run AI on. Preferably three: a business-outcome track, a functional-and-process-outcome track, and a deployment track. The first two are owned by the Executive team and the Business Heads; the last is a more laissez-faire effort, designed to tap into the hinterland innovation that lives across the company.
The stage that sets the organisation up for AI success, and the heaviest, most preparatory of the four. Its real work is to rewire the institutional operating system itself: to remove the legacy gating mechanisms — the approval chains, the risk postures, the functional boundaries — that were built for a slower world and now throttle AI execution. The most consequential move here is turning the support functions — risk, legal, compliance, HR — from gatekeepers that vet what AI produces into enablers that help produce it, which means re-engineering how those functions themselves operate, with an explicit mandate from the top. Only once that groundwork is laid does the conventional scaffolding matter: building shared understanding and direction with the leaders and teams, identifying the pilots across the tracks defined in ReFrame, setting each pilot's success criteria against the intended business outcomes, and naming the executive sponsors, budgets, integrated teams, and technology decisions that ready the organisation to genuinely embrace AI.
The stage of execution, which begins by refining the mental models of the leaders, the pilot teams, and the support functions: build fast, fail fast, rebuild fast. AI can construct complex models and programs at remarkable pace, so the usual software-development and change-request approaches only throttle it. ReBuild ensures all parties agree on the approach and run the pilots against rapid milestones — so that executive sponsors and leaders stay close to the progress, the results, and the critical decisions the rebuild cycle surfaces.
The stage of institutional honesty. As results come from the pilots: what is flowing back, and to whom? Who decides what moves from pilot to production? And, most pointedly, do the pilot's results actually hold in production? React resists both premature celebration and premature abandonment — and it distinguishes the operational wins, which can be banked within the existing business, from the strategic shifts that should redirect it. Then it feeds back into ReFrame for the next wave.
What I am most willing to stand behind is this: the four stages were not assembled as good practice in the abstract. Each was designed against a specific, documented way AI deployments fail — the technology-first reversal, the pilot-to-production chasm, the data-readiness gap, the runaway costs, the legacy retrofit, the governance bottleneck. The methodology, and the frameworks and models within it, do not merely describe a sound path; they target the precise failures the evidence has already recorded.
And I have not designed The AI Doctrine through the lens of a thinker alone, but of a practitioner. The two multi-agent AI systems I describe below were built by me, as a team of one, with AI as my collaborator. What would once have taken a team of twenty-plus a couple of years to build, I have built personally, with AI, in weeks and months — and having constructed enterprise-grade, highly complex systems that operate on real production and market data, I hold direct evidence, more than any article or video could provide, that AI delivers significant business outcomes when it is structured right. One of the two systems is fully built and working; the other has shown clear results in pilot and is being built out now.
A system for discovering trading strategies from first principles. I gave it five years of market data — Nifty, Nifty Futures, the VIX, and options across every strike, with price and volume at five-minute intervals; AI helped me assemble that dataset in the first place. And I placed one deliberate constraint on it: no technical indicators, and an explicit instruction never to use them. The inherited human playbook was withheld on purpose, so the system would have to derive structure from the raw data itself.
It runs as four agents in sequence. The first begins from a clean slate, exploring the data and forming hypotheses. The second refines each hypothesis toward defined return and risk thresholds — configurable — and tests whether it can become a tradeable signal; this agent actually writes the signal and submits it to the third. The third validates the signal against a control dataset the first two cannot see, stress-testing each surviving hypothesis against varied market conditions; what fails is sent back to the first two agents with the reasons it failed, and only what passes proceeds. The fourth converts a validated hypothesis into a live signal and paper-tests it for a defined period, according to the nature of the signal; once it passes, it joins a repertoire of signals available to execute live trades. The first agent, meanwhile, can either pick up an existing failed signal to work on further, or begin a fresh exploration of its own.
I can say from live experience that this system has surfaced genuinely interesting signals — in option buying, option selling, intraday strategies, and multi-day and multi-week strategies.
The platform those signals feed — a multi-agent system for navigating markets, organised around a simple but powerful idea: that intelligence and execution can each come from either a human or a machine, and the real work is in composing the four combinations deliberately. In one quadrant, the Cognitive Engine's signals are executed autonomously, end to end. In another, I supply a macro, multi-month view from my own market model and the agents execute it — deciding entry and exit, when to book profit and re-enter, and when to scale capital hard into a strategy that is working without taking on proportionate risk, which is where real asymmetry between risk and return is made. In a third, the machine scans a broad universe I define — equities, indices, precious metals, commodities, crypto — and surfaces where I should focus my own judgement. The fourth supports my own trading directly.
Beneath all of them, the strategy agents work only at the level of decision — when to act, how much to commit, when to stop — and hand their order intents to a separate execution backbone, Vajra, which carries them out asynchronously across multiple brokers through a purpose-built adapter layer. The layer I am adding now is a Hedge Fund Manager agent that allocates capital across the strategy agents, draws profit back to a central fund as it is booked, and runs a regime detector that senses shifts in the underlying market and reallocates ahead of them — engineered for stable returns in an unstable world. The platform is still being built; its pilot results have been strong enough to tell me the architecture is right.
As I have said, it is these personal experiences — more than any single article or video — that have shaped how I think about AI and its possibilities. The AI Doctrine I have put together is, in the end, the work of a Practitioner Thinker.
This letter has covered the body of that work: understanding the reset, a discipline for institutions, a methodology for AI success, the leadership and people work the transition demands, a framework for individuals, and the AI-native systems that put the thinking into practice — built end to end by one person, with AI as collaborator. It is far more than a single letter can hold, and I have kept this one to the frame rather than the full mechanics.
If any of it meets where you are — in your institution, on your board, or in your own thinking — I would welcome the conversation.
The work of the next decade will be done by those willing to think differently about it. That is the company I hope to keep.