Field notes on building growth that holds

AI has crossed from answering to acting. What an agent may do alone, what it must show before acting, who answers when it fails, how it is monitored, how cascading errors are contained and how a bad run is undone: six decisions, four autonomy levels and a seven-step plan for the next two quarters.

On 28 July 2026 the Model Context Protocol dropped sessions and became a plain HTTP workload. What MCP is, what the 2026-07-28 specification changed, six planning implications for the operating model, and the seven steps to take in the next two quarters, written for a board and a technical founder on the same page.

Why strategy fragments between departments, and what closes the gap: unowned handoffs, conflicting incentives, undefined decision rights and unbudgeted capacity, each with the mechanism that addresses it.

What automation can and cannot do for strategy execution: the administrative work it genuinely removes, the judgement it must not replace, what has to exist before it is worth attempting, and how it fails.

Game theory as a habit of thought rather than a set of equations: dominant strategies, why a stable outcome can be bad for everyone, credible commitment, repeated games, and the limits of using any of it as prediction.

Go-to-market settled by economics rather than by preference: what contract value lets you afford, why most named growth models are channels rather than motions, and the signals that say you are running the wrong one.

What actually determines whether a raise closes: the evidence each stage requires, how the instruments differ, the terms that matter more than valuation, and why running a process beats working through a queue.

The people side of change treated as a discipline: why transition runs on its own schedule, what resistance is actually made of, how to diagnose a stalled adoption instead of exhorting harder, and why sponsorship outranks communication.

Space logistics assessed on demand rather than on need: which segments have paying customers today, why operators often prefer replacement to servicing, how regulation rather than willingness to pay is creating the debris market, and the six questions a canvas should actually answer.

Account-based engagement for a market that is a list rather than a funnel: how to build the account set, what to research before contact, how to translate capability into language a buyer can defend internally, and how to tell an account that is moving from one that is consuming effort.

The B2G model as a set of internal capabilities rather than a sales channel: capture before the tender publishes, a bid decision with written criteria, pricing models that decide who carries risk, teaming terms worth reading, and the compliance function a fifteen-person company actually needs.

The implementation sequence: build the value model before choosing any technology, allocate cost to serve because it usually reverses the ranking, segment on current and potential value, and change three real decisions before calling it a programme.

Personalisation infrastructure assessed on economics rather than ambition: where the return actually sits on the ladder, why identity resolution is the real cost, how to measure incrementality honestly, and the EU rules that make personalised pricing a poor idea.

Which engagement flows actually return something and in what order to build them, why a process should work manually before it is automated, and the consent, deliverability and frequency problems that quietly ruin otherwise sensible programmes.

Customer centricity as Peter Fader defined it: a resource allocation discipline resting on real differences in customer lifetime value, why it is not the same as excellent service, and the three conditions that decide whether it applies at all.

Where the revenue in the space economy actually sits, why launch attracts attention out of proportion to its size, what export control means for a European supplier, and which segments a new entrant can realistically address.

Marketing in a market with a few hundred possible buyers, where customers cannot be named, technical detail can be export-controlled and public claims resurface during bid evaluation. What to publish, what to measure, and which parts of the commercial playbook actively cost contracts.

How defence procurement actually works for a small European supplier: who really decides, why pilots so rarely become programmes, which routes to market exist, and what clearance, export control and IP terms determine before commercial strategy does.

A seven-step structure for a business case that holds up in front of a board: objectives and success criteria, costs against benefits, the conditions behind the numbers, and an executive summary that carries the argument.

Issue trees as practitioners actually use them: diagnostic against lever trees, arithmetic decomposition that is provably complete, and the discipline of sizing branches before analysing them.

EOS assessed rather than summarised: the six components, the Visionary and Integrator split that matters more than the tools, where the system fits by size and shape of company, and the strategy question it deliberately leaves untouched.

Teece's framework beyond the three headline words: the distinction between ordinary and dynamic capabilities, the routines that produce each, why seizing rather than sensing is where companies fail, and when the whole framework is not worth its cost.

Building a strategy function that changes decisions rather than producing documents: the mandate that comes before the org chart, the reporting line that determines what the function becomes, and the hiring sequence that works.

The strategy map explained as what it actually is: a one-page causal model across four perspectives, where the arrows are testable claims rather than decoration. How to build it, instrument it and revise it.

How OKRs connect strategy to the operating quarter: the difference between objectives, key results and initiatives, why strict cascading and bonus-linked targets break the method, and the cadence that keeps it alive.

The four components of a positioning statement, the frame of reference decision that comes before them, the three tests a differentiator has to pass, and how to validate a position before it reaches the market.

TAM, SAM and SOM explained from both an academic and a practical angle, with worked top-down and bottom-up calculations and the place each measure holds in the strategy lifecycle.

How corporate, business unit and functional strategy stack into one system, and why treating strategy as a standing function rather than an annual event is what makes the cascade hold.

The five phases that connect ambition to delivered results, from definition and market intelligence through formulation, execution and realisation, and what changes when strategy becomes a standing function rather than an annual event.

A five-stage blueprint for facilitating change, from formulation and planning through implementation, transition and reinforcement, with the failure statistics everyone quotes examined rather than repeated.

The five connected choices that turn ambition into decisions people can act on: aspiration, where to play, how to win, the capabilities that deliver and the systems that hold it together.