1. It is easy to get an answer too early
A single request in a ChatGPT or Claude chat can turn a few ideas about a business into a launch plan. You can receive a landing-page outline, a set of messages and a campaign schedule before you have brought your own thinking together.
A founder or small team may have observations about a problem, ideas for a product, possible customers and an ambition for the business. Some of this is written down. Some lives in conversations or in the founder's head. These ideas are the starting material for understanding what the business could become.
Chat makes it easy to move straight into production. Each answer gives you more things to do, and each new request can introduce another plausible direction. You correct one proposal, open another conversation and receive more material. The underlying ideas remain scattered while the work becomes increasingly detailed.
Operating a business with AI requires a coherent understanding of the business behind that work. That understanding gives the agent a context for its choices and gives you a basis for judging the result.
2. Start at the top
Starting at the top gives each part of the business a purpose before the work becomes detailed. The initial thinking may contain a problem worth solving, product ideas, possible customers and convictions about the kind of business the founder wants to build. Bringing that material together reveals how the ideas connect, where they conflict and which questions remain open. AI can help develop that understanding while the founder settles the choices that give it direction.
Decomposition develops this initial understanding into areas such as strategy, branding, messaging, content, product, web interface, site, legal, financials, customer relationships, etc. Each area has a distinct subject within the same business. Strategy develops the venture definition and the connected choices about customers, positioning and offers. Product develops what the business provides. Brand develops its character and expression; messaging develops the meanings and proof its communication carries. Their relationships give narrower work a basis for its decisions.
From the business down
The business’s purpose and initial understanding give context to illustrative areas. Strategy contains the venture definition, customer choices and offers. Brand, Messaging, Product and other areas share the same direction. Branches show scope, not a completion sequence.
Business
Purpose and initial understanding
Ideas, observations and convictions brought together
Gives purpose and context to
Strategy
Brand
Messaging
Product
…
Illustrative areas · shared direction, distinct subjects
For a launch page, those contributions become concrete:
| Definition | What it contributes to the page |
|---|---|
| Strategy | Venture definition, target customers, positioning and the offer, including what the customer receives and its access conditions |
| Brand | The venture's character, voice and visual identity |
| Messaging | The product's value, the messages that explain it and the proof behind them |
| Web copy | The page brief and wording: what the page must explain and which next action it enables |
Each area's document contract makes its boundary explicit: which documents belong, what each contains, which definitions it draws from and how they relate. This gives the agent a defined space in which to work. A page brief can apply the chosen positioning without reopening the target market. Messaging can express the offer without inventing different access conditions.
The same boundary determines what stays outside that space. An abandoned campaign proposal or an unrelated product idea would introduce another source of direction for the page. Adding documents indiscriminately makes the agent resolve conflicts and relevance that the organisation of the work should already have settled.
Granularity matters because a narrower model makes the important distinctions visible. A page's purpose, audience, material and next action are specific enough to guide decisions about its content and interaction. A folder called “marketing” with a collection of notes leaves much more for the agent to infer.
A small business can begin with short definitions, with related material sharing a document where that remains clear. The value comes from knowing which meaning belongs where and which definitions the next piece of work must preserve.
3. Keep the business and its work in one place
You must take care of this context yourself. A chat vendor's memory of you and your earlier conversations is an uncertain place to leave the current definition of your business. The direction you have chosen needs a home you can inspect, organise and revise.
The vendor story is that each new tool brings a little magic. In practice, spreading the business across Obsidian, Linear, Confluence, Notion and other tools also spreads its context. Notes live in one place, decisions in another, tasks somewhere else and implementation in a separate workspace. You become responsible for keeping the pieces connected and making sure AI receives the right versions.
That responsibility grows as the separate spaces fill up. Notes, abandoned proposals and copied decisions accumulate independently. Each space is easy to clutter, and the material relevant to the work becomes harder to distinguish from the noise.
Connecting these spaces to AI through MCP (Model Context Protocol) gives the agent access to their contents. The connection itself establishes no agreement about which definitions are current, which decisions govern the work or how changes stay consistent. The agent can reach several competing accounts of the business without a shared basis for deciding what takes precedence. The fragmentation and clutter now enter its working context, with another layer of connections to maintain.
Keeping this arrangement working adds costs and a maintenance headache of its own: paid subscriptions, integrations, file transfers, duplicated information and another set of places to organise. The founder or small team spends money and attention maintaining the toolset as well as operating the business.
A shared repository gives the business definitions and the work they guide one home. Strategy, messaging, product definitions, written material, interface work and code can live together, with their relationships available as the agent moves from understanding a request to carrying it out. Consolidating this work can save on overlapping subscriptions and reduce the time spent maintaining connections between tools.
With the business definitions, page copy and implementation together, the founder or team can direct the agent to the relevant material, discuss a change, inspect what it edited and preview the result. They do not have to export a messaging document from one service, transfer it to another and remember which copy the implementation used.
An IDE provides the workspace for this: files, search, editing, coding agents and change inspection in the same environment. You can move between a business definition and the code or writing it informs while staying with the work. Learning to use this workspace is valuable for founders and people outside engineering too.
One context to work from
First, illustrative document panels show notes in Obsidian, tasks in Linear, decisions in Confluence, docs and plans in Notion, interface work in a design workspace and source code in a separate repository. Other tools and spaces are indicated by an ellipsis. Second, an illustrative venture/{name}/ repository contains business knowledge—strategy, brand, messaging, product and growth documents—and implementation: site source code and tests. Selected document and source rows show the context for the work.
Distributed context
Several places to maintain
Obsidian
Notes and proposals
Linear
Tasks
Confluence
Decisions
Notion
Docs and plans
Design workspace
Interface work
Separate repository
Source code
...
Competing definitions remain
Subscriptions · synchronisation · transfers
Illustrative arrangement
Organised repository
One context source
venture/{name}/
strategy
venture.md
brand
voice.md
messaging
message-system.md
product
product-definition.md
growth
growth-programme.md
code/site/
src/site/index.ts
tests/
Selected context
Within that shared home, the boundaries developed earlier keep the context precise. The page work can use the strategy, brand, messaging and brief it needs, while the rest of the business remains accessible in the same repository. The agent has a deliberate selection of relevant material to work from.
Version control preserves earlier versions without keeping them alongside the current direction. There is no need to maintain a row of files called “final”, “revised” and “final again”. When the task calls for understanding an earlier decision, the history is there to inspect. Routine execution can begin with what is agreed now.
4. Develop one vertical at a time
A shared understanding of the business can develop into useful work without every area being defined in full at once. A founder or small team can take one vertical far enough to produce a result, developing the necessary depth in each area it passes through. The rest of the business becomes more detailed as its work requires.
Decomposition identifies an area worth developing. Modelling establishes how that area works: the things it deals with, their relationships, the work that changes them and the conditions a result must satisfy. This draws on mathematical modelling: types make distinctions explicit, relations connect them, transformations describe the supported work and rules establish what must remain consistent. The model gives AI a precise professional context in which to develop the business's actual choices.
Strategy provides a concrete example. Its model includes a venture definition, customer segments, target choices, opportunities, positioning, revenue, offers, prices, objectives and priorities. These form a connected account of how the business intends to create value and earn from it. A target refers to a defined customer segment. An offer addresses a need in that target. Its price belongs to a particular offer and agrees with how the customer is charged.
Those relationships give the work its requirements. Choosing a target means developing a reasoned preference among the relevant segments. Developing an offer means connecting the promised result, customer, delivery and commercial exchange. Revising a price means preserving agreement with the offer and its charging basis. The model identifies what the work must consider and which decisions a change may affect.
The document contracts express these meanings in files the agent can read and maintain. Checks can establish that references resolve, units agree and arithmetic holds. Specialist judgment examines whether the choices make sense together and serve the venture's purpose. The founder decides which direction to accept. Each contributes something different to the quality of the result.
Workers can now be defined around coherent parts of this model. In the Strategy group, the venture strategist develops the venture definition, objectives and priorities. The market strategist develops segments, targets and opportunities. The positioning strategist develops differentiation and positioning. The business model designer develops revenue, offers and prices. Their boundaries follow from the decisions they maintain and the expertise those decisions require.
The model gives workers their responsibilities
Strategy declarations with State expanded into fourteen document components. Doc names the document of the specified type; multiplication joins these components into the shared state.
S : Request × Input × State -> Result(Change)
accept : Change × OwnerDecision ⇀ State (Supervision)
Change = { Cand(t) | t is selected from State's document types }, each Cand(t) = (Doc(t), change id, base revision | absent-base)
X_work = X_current ⊕ C_change (every skill reads the working view)
State = Doc(venture-definition) × Doc(objectives) × Doc(bet-portfolio)
× Doc(focus-plan) × Doc(strategy-review) × Doc(segments)
× Doc(target-segments) × Doc(opportunities) × Doc(differentiation)
× Doc(positioning-decision) × Doc(category-definition) × Doc(revenue-model)
× Doc(offer-catalogue) × Doc(price-book)
Each worker carries the methods for its part of the work: how to develop possibilities, compare them, explain a recommendation and revise it when the basis changes. The group also defines what must agree across their results. A change to the target customer can therefore reach the affected positioning and offers through known relationships. The worker has a model, professional methods, defined results and a basis for checking them.
The complete Strategy group is available under MIT, including its worker definitions, document contracts, methods and checks. It makes this construction inspectable and reusable: a founder can bring an actual business question and work through it with their own coding agent. The result is a business definition or strategic choice that remains in the repository and can direct subsequent work.
5. Operating with AI
This is where organising the business becomes operating capacity. A founder can take responsibility for work beyond their own professional training because the agent has the relevant model and methods to work with. The person supplies purpose, material and judgment; AI develops the specialist work within that context. The result gives the business something it can use and continue developing.
The same approach extends across connected areas:
Strategy → Brand → Messaging → Web copy → Web interface → Site.
The accepted strategic choices give Brand and Messaging direction. Their definitions and results then inform the writing, interface and implementation. Each area has its own model and methods, while the shared context lets work carry through into a usable outcome. The founder or team can inspect how a business decision reaches what a customer reads, sees and does.
The work also makes the construction better. A repeated misunderstanding can reveal an unclear concept, a missing relationship or an inadequate method. Correcting that definition gives subsequent work a better basis. Results remain beside the context that produced them, so the operator can examine the reasoning and improve the part responsible. This is how the system develops through the venture's actual work.
For a small team, that changes how much it can carry. The same people can direct strategy, communication, design and implementation with AI doing more of the professional production. Clear definitions reduce repeated explanation; connected results reduce handoffs; one workspace reduces copying and maintenance. More of the business becomes manageable without a separate person and toolset for every responsibility.
That capacity can be developed gradually or supplied in ready-to-use pieces. An existing worker or group brings a model, methods and checks for a defined area. A custom construction develops them around work particular to a venture. In either case, the capability has a clear place in the business and a result the operator can judge.
A coding agent and a working repository are enough to begin. The value grows through the understanding developed inside them: what the business is trying to achieve, how its work is defined and how its parts support one another. Operating with AI becomes a practice the founder can direct, examine and keep improving as the venture grows.
From LeanOS: Strategy · MIT · Ready-to-use tools · Custom builds.