Most organisations already have some form of specification.
It may not be called that.
It might be the way an experienced manager explains a decision. A folder of process documents. A prompt someone carefully maintains. A calculator the team regularly overrides. A checklist that protects the important parts of a handoff.
The question is not whether the organisation has information about how it works.
The question is whether people and AI can depend on that information while doing the work.
That is what this maturity model is designed to show.
This is not a score for the whole company
Maturity models often become consultant theatre.
The organisation is assessed. A coloured chart is produced. Leadership learns it is “Level 2.4” and begins a transformation programme intended to move everything to Level 5.
That is not the purpose here.
A business can be highly mature in one area and entirely dependent on memory in another. Its payroll process may be operational and dependable while its proposal process changes according to whoever is selling the work. Its core product may have a clear specification while a newer service still lives in the founder's head.
That is normal.
Use the model on one important product, service, agent or repeated way of working.
Then make the smallest move that improves it.
Do not try to make the whole organisation mature at once. Make one valuable part more dependable.
Stage 1: Remembered
At the Remembered stage, the way the business works lives mainly in people.
An experienced person knows what good looks like. They understand the customer, the exceptions and the history behind the rules. When something unusual happens, everyone knows who to ask.
What it feels like
- Work moves well when the right person is available.
- New starters learn through observation and repeated correction.
- Similar situations are handled differently by different people.
- Teams ask for approval because the boundaries are unclear.
- A leader's absence slows work or increases risk.
- AI outputs look plausible but repeatedly miss unwritten expectations.
What is valuable here
Remembered knowledge is not bad knowledge.
It often contains years of hard-earned judgement. The organisation may be flexible because capable people can respond to nuance without waiting for a formal process.
The problem is reach.
The judgement cannot be reused without involving the person who holds it. As the business adds people and AI agents, more work returns to the same experts for interpretation.
The risk
The person becomes the undocumented specification.
If they leave, become overloaded or make inconsistent calls under pressure, the organisation has no dependable basis for action.
The next move
Do not begin by documenting everything the expert knows.
Choose one recurring decision and capture:
- the outcome it is meant to protect;
- the rules that nearly always hold;
- the room other people have to use judgement;
- the point at which the expert still needs to be involved; and
- the questions the business has not decided.
The first gain is simply making one important piece of judgement reusable.
Stage 2: Documented
At the Documented stage, information about the work exists outside people's heads.
There may be process maps, policies, standard operating procedures, templates, prompts, task instructions and wiki pages.
The organisation has invested in explaining how things should happen.
What it feels like
- People can usually find something relevant if they know where to look.
- Several documents describe overlapping parts of the same work.
- Nobody is completely sure which version is current.
- The official process and the accepted workaround have drifted apart.
- AI can retrieve plenty of information but encounters contradictions.
- Documents are updated after a project, then slowly become stale.
What is valuable here
The business has begun separating its knowledge from the people who created it.
Important language, decisions and procedures are visible. New people have a starting point. AI has material it can search and summarise.
This stage is necessary. It is just not sufficient.
The risk
Documentation can create the appearance of clarity without forcing the organisation to decide what is authoritative.
More pages may produce more possible answers.
When reality and the document diverge, people follow reality. The work changes immediately while the documented version remains behind.
The next move
Choose one durable truth that appears in several places.
Decide:
- which version the organisation genuinely intends to follow;
- where its authoritative home should be;
- who owns it; and
- which other documents should link to it rather than repeat it.
The shift from Documented to Owned begins with one home per durable truth.
Stage 3: Owned
At the Owned stage, the important part of the business has a living specification with a clear human owner.
The organisation can distinguish current commitments from supporting information, temporary work and unresolved questions.
What it feels like
- People know where to find the current purpose, goals and boundaries.
- Important terms have an agreed meaning.
- Rules have owners rather than anonymous authors.
- Contradictions are brought to the person who can resolve them.
- Open decisions are recorded honestly instead of filled with assumptions.
- A prompt or task can refer to the specification without becoming another policy source.
What is valuable here
The business has created a dependable basis for delegation.
A capable person or agent can begin with the same organisational understanding instead of reconstructing it from messages, files and memory.
The owner remains connected to meaning and outcomes. AI can help draft, compare and maintain the material, but it does not quietly become the authority behind it.
The risk
The specification can still become a well-maintained document that sits beside the work.
People may agree with it in a workshop and ignore it under pressure. Agents may use copied fragments that drift from the source. Completed work may change reality without changing the specification.
Ownership is necessary, but ownership without use does not change how the organisation operates.
The next move
Connect the specification to one real piece of work.
Load it before the work starts. Use its goals and guardrails while the work happens. Route a genuine open decision to its owner. At completion, ask whether the work changed any durable truth.
The specification becomes operational when work depends on it.
Stage 4: Operational
At the Operational stage, the specification participates in execution.
People, AI agents and systems use it to carry out a repeated Loop. The work has a clear start, goal, definition of done, handoff and boundaries. Genuine decisions reach the people with authority; routine judgement remains with the people and agents doing the work.
What it feels like
- Work begins with the context it needs rather than another briefing meeting.
- People and agents can act within clear boundaries without repeated approval.
- Exceptions arrive with options, trade-offs and the relevant evidence.
- Handoffs are judged by whether the result is ready, not whether every worker followed the same recipe.
- Managers receive fewer repeated questions and more meaningful decisions.
- The defined path is often faster than creating a workaround.
What is valuable here
The organisation is no longer relying on people to find, remember and manually apply a separate document.
One clear decision can guide many future actions. Human judgement is amplified without removing human authority.
This is where the practical promise of a spec-driven organisation becomes visible: greater speed without constant supervision.
The risk
Operational specifications can become rigid if change is difficult, or unstable if every new signal rewrites them.
An approved exception may accidentally become a rule. A local measure may improve while another team absorbs the cost. Agents may continue executing an approach after reality has moved on.
The next move
Create an intentional learning loop around the work.
Collect outcomes and exceptions. Let AI surface patterns and propose improvements. Keep observations, experiments and durable changes distinct. Ask the accountable owner to decide when the organisation has learned something that should change the specification.
The move to Compounding is not automatic self-improvement. It is better evidence connected to better human decisions.
Stage 5: Compounding
At the Compounding stage, completed work improves the context available to the next piece of work.
The organisation does not merely execute from a specification. It notices where the specification helped, where it was incomplete and where results challenge the current approach.
What it feels like
- Agents surface recurring exceptions and meaningful outcome patterns.
- Decisions arrive with evidence, rationale and affected dependencies.
- Experiments can run quickly without immediately becoming permanent rules.
- Accountable people approve durable changes and can explain why.
- Supporting work, agents and materials stay aligned with an intentional change.
- The next person or agent begins with better context than the previous one had.
What is valuable here
Each piece of work leaves the organisation more capable.
A customer conversation can improve the offer. A delivery issue can clarify a proposal boundary. A manager's decision can become reusable context for the whole team. Organisational learning stops being trapped in individual experience or completed tasks.
The result is compounding clarity.
The work uses the specification, and the specification gets better because the work happened.
The risk
Compounding can be confused with constant change.
If every observation updates the specification automatically, people lose the ability to understand and own how the business operates. The organisation becomes responsive to every signal and dependable to none.
Maturity at this stage means a higher quality of change, not simply a higher frequency.
The next move
Improve the quality of the learning system.
Ask whether proposed changes include the relevant evidence, consequences and affected parts of the business. Raise the change bar as a specification becomes more proven and more widely depended upon. Keep the human owner connected to the outcome rather than only approving edits.
There is no final state where the organisation runs itself.
There is an ongoing practice of making human judgement clearer, more reusable and better informed.
One loop through all five stages
Here is what the model looks like applied to a single piece of work: meeting follow-up.
Remembered. The project manager knows who agreed to what. Actions live in their head and their notebook. When they're on leave, follow-up stops.
Documented. The team writes down how follow-up should work: transcript in, actions out, decisions recorded, follow-up sent. The document is useful, but three tools and four people each still interpret it their own way.
Owned. The meeting owner becomes accountable for the specification. One transcript source becomes authoritative. One system becomes the home for actions and decisions. Conflicting interpretations now have somewhere to be resolved.
Operational. The specification becomes a running meeting-to-action loop: an agent starts when the transcript arrives and finishes only when every action has an owner and date, decisions are recorded and follow-up is sent. The owner handles genuine exceptions, not routine runs.
Compounding. Corrections and exceptions from real runs flow back to the owner, who decides which of them should change the specification. The loop gets more reliable every week, and its sub-loops (task creation, decision recording, follow-up) become reusable in other loops.
The same progression works for sales follow-up, onboarding or reporting. The six-question loop specification is the document that carries a piece of work from Documented to Operational.
A quick diagnostic for one piece of work
Choose a repeated activity such as proposals, customer onboarding, campaign approval or hiring.
Then ask:
- Where does the real decision logic live? In a person's head, scattered documents or one owned specification?
- Which source wins when information conflicts? Is that authority explicit?
- Does the specification guide real work? Or does someone still need to translate it manually?
- Can people and agents act without repeated approval? Are the boundaries and open decisions clear?
- Does completed work improve the next run? Who decides whether an observation becomes a durable change?
The earliest “no” usually identifies the current stage.
Do not argue about whether the workflow is really Stage 2 or Stage 3.
Choose one behaviour that would make it more dependable and implement that.
The most useful move at each stage
The model can be reduced to five practical moves:
- Remembered → Documented: capture one recurring piece of expert judgement.
- Documented → Owned: give one durable truth an authoritative home and accountable owner.
- Owned → Operational: connect the specification to one real end-to-end Loop.
- Operational → Compounding: bring outcomes and exceptions back as evidence for deliberate improvement.
- Compounding → more dependable: strengthen human ownership and the quality of change as scale increases.
You do not need to complete every move before creating value.
Each one removes a specific kind of friction.
How BusyWork Dispatch fits
BusyWork Dispatch provides a practical route through the middle of this model.
The Workbook gives durable truth an owned home. Dispatch connects that context to work moving now. A Loop defines a repeated piece of work through its start, goal, definition of done, handoff, resources, guardrails and owner.
As work is completed, durable learning can return to the Workbook rather than disappearing into a closed task or private conversation.
The right place to begin is not a company-wide maturity assessment.
Choose one product, service or recurring piece of work whose current ambiguity is creating real cost. Move it one stage. Use it. Learn from it. Then decide what should come next.
Bottom line
A spec-driven organisation matures as important business knowledge moves from memory, to documentation, to clear human ownership, into the work itself and finally into a deliberate learning loop.
The five stages are:
- Remembered
- Documented
- Owned
- Operational
- Compounding
The goal is not to earn a Level 5 badge.
It is to make one important part of the business clear enough that people and AI can depend on it, then allow completed work to make that shared understanding better.
That is how organisational clarity begins to compound.
BusyWork Dispatch helps organisations move durable knowledge into an owned Workbook and connect it to work that gets built, run and improved. See how BusyWork works or book a call with Ben.
