CONCLUSION

The argument of this manual begins with a practical commitment: the work performed by many participants should remain directed toward a result that someone has reason to want and grounds to accept. Agents enlarge the amount of work a person can commission, compare, and bring into a project. They also enlarge the set of contributions that must be understood in relation to one another. Project management gives that activity a purpose, an organisation, and a course through change.

The approach developed here is deliberately opinionated. It favours an accepted product basis before production, identifiable Packages and Deliverables, explicit dependency relationships, a small fixed repertoire of agent roles, and a bounded documentation/governance closeout after each undertaking's intended implementation and evidence integration. It leaves considerable freedom in the assignments, methods, execution resources, and organisation-specific procedures through which those commitments are carried out. The reason for preferring this arrangement is the work it enables people to undertake responsibly.

7.1Agency requires a common purpose

An individual contribution can be accomplished with skill and still fit the project poorly. Its author may have solved a narrower problem than the one intended, used an assumption another contributor does not share, or completed a result for which nobody has arranged integration. More execution capacity can multiply such contributions. The management problem is to give their authors a sound basis and retain responsibility for how their results meet.

Treating an agent as an other makes this relationship explicit. A request begins a contribution whose preparation the requester has not performed personally. The agent can interpret material, select an approach, use tools, and return a result that deserves serious consideration. The human must then examine the work to the amount and level appropriate for its intended reliance. The same responsibility extends to the meaning of the request, the consequences discovered during execution, and the combined result received afterward.

This view supports substantial agent initiative. A good brief leaves room to solve the assigned problem. A capable manager follows findings and rearranges authorised work. Suitable tools remove repeated operations from the agent’s attention. Those freedoms are useful because the purpose and boundaries remain understandable. Requiring the human to approve every routine act would consume the attention needed for choices that determine the product and its consequences.

The human’s participation also shapes the undertaking before there is a settled specification. An example offered by the design partner can disclose a consequence the human had not considered. A correction can reveal that the proposed categories miss the purpose. Conception develops through that exchange. The PRD gives the position reached a form that other participants can examine and use. Its required depth follows the project and the consequences of leaving particular questions open.

The humanist position is therefore a positive account of collaboration. The agent widens what can be investigated and prepared. The human judges what is worth pursuing, what must be preserved, and which results they are prepared to rely upon. Better reckoning can improve that judgment’s basis while leaving its responsibility with the person.

7.2Progress changes the work that management must do

The phases give development a direction without requiring every part to move at the same speed. Each phase establishes something that changes the useful organisation of subsequent work. Conceptual development establishes the intended outcome and its basis. FEED gives that basis a division into Packages and Deliverables and prepares their working context. The work toward 30% resolves enough of the production relationships to construct a usable DAG. Detailed development then establishes the interfaces and design needed for longer completion undertakings.

This account makes early preparation consequential. A useful definition gives several later contributors the same object to work toward. A resolved interface lets them proceed separately. A scope boundary identifies who must carry an obligation and who depends on its result. The preparation earns its place when participants can use those relationships to act, examine a return, or diagnose a discrepancy.

The account also limits what preparation can establish. An accepted PRD can contain open design questions. A constructed DAG can contain pending inputs. A produced Deliverable can await a particular review or human acceptance. These are ordinary conditions of development when their extent and consequences remain visible. A single completion percentage would obscure much of the information the manager needs.

The formation of the first DAG gives the transition into detailed development a comparatively definite result. The end of 60% requires a broader judgment about the remaining route. Further structural changes are no longer anticipated, so substantial work can be organised over longer task and time horizons. That expectation can be overturned by evidence. The appropriate response is a controlled return to the affected question and the consequences of its answer.

I regard this treatment of return as one of the manual’s most useful features. It allows the project to retain what remains valid while revising what no longer serves. A finding can reopen one design decision without erasing unrelated progress. An amendment can preserve the reason for the original choice and explain why its replacement is now preferable. The project’s continuity resides partly in that intelligible relationship between successive positions.

Steering remains important after the structure exists. The graph can expose possible routes, but it does not supply the owner’s priorities or decide which difficulty deserves attention now. HELP_HUMAN and WORKING_ITEMS keep the route connected to the human’s objectives while the team works through the details. This becomes especially valuable when an undertaking is long enough to outlast several sessions or executors.

7.3Keep the organisational choices few and the assignments precise

The four-role repertoire provides a durable set of responsibilities. HELP_HUMAN maintains the relationship to the human and coordinates undertakings. HELPS_HUMANS develops conception and design. WORKING_ITEMS carries an understood undertaking through execution and owns its integrated return. TASK performs a bounded contribution. Their use changes with the work while their responsibilities remain recognisable.

The preference for a fixed repertoire is a choice about where complexity should live. Technical specialisation belongs in the assignment, selected method, context, and tools. The managerial relationship concerns who prepares that assignment, supervises its progress, receives its findings, and integrates the result. Keeping those matters distinct avoids creating a permanent organisational identity for every kind of work encountered.

The distinction also improves the question asked about models. The relevant question concerns the task’s demands and the means available to examine its result. A difficult bounded review may justify more capability and reasoning effort than a routine coordination assignment. A repeatable comparison may be best carried principally by a tool. Role names and Type numbers cannot settle those choices. Experience of comparable work can inform them, provided the comparison includes preparation, review, correction, and integration.

The work graph and delegation hierarchy need one another, but they need not have the same shape. A graph records which contribution requires another. The hierarchy records who assigns and carries responsibility for the return. Several graph nodes may belong to one manager’s undertaking. A dependency may cross two managers and require coordination through HELP_HUMAN. Giving those relationships an explicit treatment is more useful than reproducing every graph node as an agent.

I would preserve this simplicity even as the tools and models change. It gives participants a stable account of what their collaboration requires and leaves execution free to improve. Its adequacy should still be judged in use. Where a proposed new role solves no responsibility left unassigned by the existing repertoire, it adds another definition to maintain without an established benefit. Where the work exposes a missing responsibility, examine that problem before choosing the remedy.

7.4Make the record useful to thought and action

A record is useful when it helps someone establish what applies, what exists, what has been examined, and what should happen next. This manual gives those questions different homes. The accepted basis carries commitments. The implementation and its observations establish actual behaviour. The local graph carries the current route through authorised work. Evidence and decision records preserve the grounds and acts on which later participants may rely.

The distinctions matter most when the records disagree. A source change may be implemented while its ruling remains unapplied to a Deliverable. A reported test may concern an earlier candidate. A current note may describe a proposal as an accepted decision. Reconciliation must determine the kind of difference before choosing a repair. Making the files agree with one another is insufficient when their agreement would conceal an unmet requirement or misattribute the human’s direction.

The claim-granularity rule provides a particularly useful discipline. A production contract should preserve the commitments, interfaces, and verifiable conditions that matter across implementation changes. Repeating incidental mechanisms, revision snapshots, and transient counts gives the record many opportunities to become stale without any change to the obligation. Keeping those details in supporting records can reduce repeated reconciliation while preserving the evidence needed to understand the current implementation.

This discipline demands precision rather than vagueness. A broad statement that cannot be examined gives little help to an executor or reviewer. A stable claim must still identify a meaningful result and the grounds by which its fulfilment can be considered. A specifically adopted mechanism must retain its governing force. The human judges disputed cases; a general preference for higher-level claims cannot settle their consequences automatically.

A short local run row makes the relevant evidence findable without distributing decision authority among Deliverables. Keep detailed run records where evidence or required execution provenance needs them, and link them from the PR and row. Existing graph results may already provide that record. A second local narrative for every action adds no necessary coverage.

The record remains an aid to a situated person. Two people can read the same material and notice different implications. A successor can discover a defect in an accepted account. A human may later revise a judgment made in good faith. The project should preserve enough identity, provenance, and history for those corrections to be made intelligently. It cannot guarantee that every participant will arrive at the same understanding.

There is a corresponding danger in the method. A project can become occupied with maintaining its forms while the product and its use receive too little attention. The remedy begins by examining what each record or control contributes to the work. A repeated report that serves no decision, dependency, recovery, or examination should be questioned. A necessary record that is costly to maintain may warrant a better tool or a simpler representation. The responsible person must continue looking through the records toward the result they are meant to support.

7.5Optimise the whole course to an examined result

The clearest opportunity for scale appears when enough of the design and dependencies have been established to release independent work. Completion toward 90% can then proceed through substantial concurrent branches. Local graphs retain the links among their production, review, integration, and reconciliation. The opportunity is considerable precisely because the preparatory work has made the contributions separable and their returns assessable.

For the adopted development arrangement, this route ends in one bounded documentation/governance closeout, terse local run entries and the final PR. Its merge confirms the completed work graph after the required review, checks and human decisions. An earlier component merge or an unmerged final candidate is an intermediate result.

Useful optimisation examines the whole route. Increasing production while review is congested creates more unfinished work. Combining too much into one candidate can make its examination and recovery difficult. Requiring every branch to wait for an unrelated return imposes an unnecessary dependency. The manager should locate the actual constraint, whether it is an input, a decision, an execution resource, a shared environment, or an integration responsibility.

Tools can remove a different class of recurring cost. A defined comparison, inventory, check selection, or evidence compilation need not be improvised through repeated model actions. A suitable tool can perform it consistently and return facts that the agent and human can examine. Qualification, failure treatment, interpretation, and later upkeep remain part of its cost. A reliable operation applied to the wrong source or criterion will reliably answer the wrong question.

Calibration and review are productive parts of this arrangement. Examining early returns can prevent a mistaken convention from being multiplied across a corpus. Independent review can expose an error that every structural check permits. Deterministic validation can establish populations and relationships too tedious to reproduce reliably in prose. Each contributes something different to the result. Removing duplication is justified when its coverage is understood and preserved; removing an inconvenient examination changes the basis on which the result can be trusted.

The human’s time also belongs in the comparison. A well-prepared decision can spare the owner repeated reconstruction. It can bring a common question forward once while preserving the cases that still differ. The purpose is to make necessary judgment practicable. A reduction in questions is beneficial only when it reflects sound preparation and suitable discretion, rather than suppressed uncertainty.

These are reasoned expectations to examine in use. The mechanisms described do not establish a fixed productivity gain, a universal saving from every workflow, or a model-independent staffing ratio. A team should retain enough observation to determine whether its methods are producing useful, examined results with less total effort and fewer unaccounted consequences.

7.6Judge the project by the product and the responsibility it leaves

A project concludes with more than a record of completed activity. It leaves a product that people may use, an account of the scope actually achieved, and responsibilities that continue after publication. Its success must be considered in relation to those outcomes, the constraints under which they were pursued, and the consequences that remain.

This is where the four perspectives meet. Ontology identifies the things and relationships for which an account is being given. Epistemology examines the grounds offered for the claims about them. Praxeology follows the work and its effects through the people, agents, and tools that performed it. Axiology concerns the purposes and obligations that make one result worth accepting and another unacceptable. Their value lies in the questions they expose about this particular project. They provide no mechanical certificate of coherence.

A useful result may justify the effort devoted to it. A failed attempt may leave knowledge of constraints and a better basis for another undertaking. A reduced project may meet a worthwhile revised objective while falling short of the original one. The concluding account should allow those judgments to be made without recoding the history to make the outcome appear inevitable or wholly successful.

In my view, the manual’s strongest argument is for preserving the conditions under which capable work can be received responsibly. The method gives agents substantial scope to act, yet keeps their contributions connected to human purposes, identifiable obligations, and examination. It gives change an explicit route without requiring the project to pretend that its initial understanding was complete. It gives subsequent participants a usable starting point while leaving them free to discover that the earlier account needs correction.

That is a demanding standard for the method itself. Its structures should continue to earn their cost in the projects they serve. Its records should help participants understand and act. Its controls should preserve consequential distinctions without multiplying empty ceremony. Its improvements should be supported by the work they actually improve.

The reason for this manual is to help people carry greater agency through that discipline. A project can then grow in size, duration, and concurrent activity while retaining an intelligible relationship between what was intended, what was done, and what someone is prepared to stand behind. The final responsibility remains human. The practical achievement is a collaboration that gives that responsibility useful work, sound grounds, and a product worth delivering.