Example Delivery Unit

This page brings together the route of DU-001: a barbershop buys a useful advance towards online booking. It shows what the customer receives, how it chooses whether to continue, and how an organized production capacity carries out that choice. Participants, events, rules and amounts are fictional and are not market evidence. The identifiers are explained in the reference model overview.

Intention and chosen scope

The barbershop wants to coordinate appointments by hand less often. It starts by buying discovery, without yet committing to full development. It then receives an offer and chooses to implement booking only. The table separates the initial intention, the commitment and the expected benefit.

Field	In the case
Identity	Delivery Unit DU-001: the barbershop’s booking.
Intention	Let customers choose appointment times online.
First commitment	C0: clarify the rules and produce a prototype, an architecture and information for continuing.
Continuation chosen	C1: implement, validate and deploy booking.
Functional result	Booking and cancellation according to availability, with no overlap for the same professional.
Outside the main commitment	Reminders, advance payment for services, and unlimited ongoing operation.
Expected benefit	Less manual coordination; it depends on use and on later observation.
Dependencies	The rules, access to the calendar service, evaluation, and permission to deploy.

Recording the intention does not authorize construction.

A purchased entry point with useful outputs

C0 is priced at 20 and covers S1, clarification, and S2, design. It delivers the rules, a prototype, an architecture, the dependencies, and conditions for discussing what comes next. The artifacts and their references are delivered to the barbershop with their permitted use stated, including use to guide a purchase from someone else under the agreed conditions.

This entry point has a result of its own. If discovery shows that the solution is not worth it, or the barbershop decides to postpone the investment, C0 can end fulfilled. There is no implicit authorization for S3 to S5, and no obligation to buy the next offer in order to justify the work already received.

The price does not reveal what discovery cost the provider internally, and it does not record any money received: billing and payment follow their own conditions. Whether entry points that end without construction are useful and sustainable is a research question, not a conclusion this example demonstrates.

To see what the bet is about, consider the work needed before C1 can be offered: asking questions, clarifying rules, producing and reconciling a prototype and an architecture, examining dependencies and preparing commercial terms. If every step required extensive manual coordination and elaboration, the entry price the barbershop would accept might not sustain C0. Cycles of AI could change that relationship by reducing the effort of producing and evaluating what is learned. The example assigns no historical cost and does not show that 20 would be enough; it identifies the mechanism that would have to sustain this first purchase.

Booking rules and clarification

In S1, the business authority defines the conditions or delegates their definition. To keep the example checkable, the service considered lasts 30 minutes. Each booking identifies a professional, a start and an end, and must fall entirely within the working hours registered in advance.

Confirmed bookings for the same professional cannot overlap. A booking from 10:00 to 10:30 allows another starting at 10:30, but prevents one from 10:15 to 10:45 for that professional. Different professionals can work at the same time. A confirmed cancellation frees the interval while keeping a record of the operation.

Simultaneous requests must not produce two incompatible confirmations. Repeating the same booking request, identified as such, does not create a second appointment. If the result of an earlier attempt is unknown, it has to be checked before creation is tried again. These criteria guide the interface, the storage and the evaluation.

A1 records the rules and their limits, and Q1 accepts the clarification to the extent the design needs. A choice that nobody has authority to make is not settled just because an agent wrote a plausible hypothesis. With sufficient delegation, a system can decide and record its basis without an additional human approval.

Prototype, architecture and the forming of an offer

S2 develops the prototype and the architecture in parallel where possible. The interface shows the choice of professional and time, confirmation, unavailability and cancellation. The architecture identifies the calendar service that stores bookings and cancellations, and how the rules will hold up under simultaneous requests.

The two outputs are reconciled in A2. A technical constraint can change the prototype; a question about the experience can send the work back to S1. The case does not require separate product, UX and architecture roles, nor does it assume design is a single activity without iteration.

Q2 accepts the technical basis for its purpose. Progressive Estimation uses what was learned, and the provider decides to offer implementation, validation and deployment for 80. Reminders would add 15, making an alternative at 95. Advance payment for services remains outside the offers.

The information allows the provider to take on bounded conditions and risks without guaranteeing the exact cost of construction. The provider remains responsible for its offer; the buyer does not unilaterally decide how the provider allocates resources or absorbs costs.

Partial approval and an economic mandate

The barbershop chooses booking at 80. The selection can be delivered because it does not depend on reminders. In the delegated variant, a system can make the same decision under an additional ceiling of 90, conditional on scope, the identified integration, the criteria and compatible commercial terms.

After acceptance	Fictional amount	Meaning
C0 fulfilled	20	Price of the discovery received.
C1 authorized	80	Price of the continuation assumed, still to be carried out.
Aggregate contracted	100	The sum of 20 and 80; not a confirmed payment.
Additional mandate	90	Ceiling for continuing after discovery.
Uncommitted balance	10	The difference between the additional mandate and C1.
Aggregate ceiling	110	Discovery plus the additional mandate.
Alternative with reminders	95	Exceeds the additional mandate by 5; not contracted.

The balance of 10 does not authorize undefined features, and it does not make the reminders, priced at 15, partially bought. A new decision would require examining the current offer and its dependencies. If the buyer had asked for an adjustment before accepting, the proposal would be under revision and construction would not yet be authorized.

Hybrid and autonomous configurations

The same scope admits different policies. In a hybrid configuration, people take part in the decisions they have kept for themselves while agents produce, investigate and check parts of the work. In an autonomous configuration, systems carry out delegated clarification, design, the offer, the choice, production, evaluation, correction and deployment within the relevant mandates.

Autonomy combines separate authorities. The buyer can delegate purchasing and business choices; the provider can delegate forming offers, allocation and technical acceptance; the operator of the environment can delegate permission to deploy a revision that is ready. None of these delegations grants the others automatically.

The Execution Policy, Routing and Execution Pools make the configuration operational. Engineering defines the criteria, context, access and recovery, and Quality Gates can operate as automated decisions built into the cycle. The case does not require a human stop at each stage. Missing evidence, or a condition outside the mandate, blocks only the actions that depend on it until it is dealt with properly.

Implementation, waiting and correction

S3 produces the first revision of A3. Local checks support Q3 without declaring the whole integration finished. S4 then waits for access to the calendar service’s validation environment; the demand stays active with that dependency stated explicitly.

Once access is granted, evaluation sends simultaneous requests for the same professional and finds two overlapping confirmations. A4 records the unfavourable result and Q4 is not granted. The problem breaks a rule that is in force, so A3 must be corrected; the rule is not redefined to accept the defect.

C1 covers the correction and the re-evaluation of this behaviour. The flow returns to S3, records another revision, repeats the relevant checks and returns to S4. The earlier local decision stays in the history and is not transferred automatically to the new implementation.

Meanwhile, eligible resources can serve other demands without losing the context of this one. Resuming requires the references, decisions and evidence to be available, and permissions kept separate between clients. Reusing capacity does not remove its cost of preparation, waiting and coordination.

Evidence for accepting the integration

Criterion	Information relevant to the evaluation
Duration and hours	A 30-minute appointment entirely within the registered availability.
Overlap	An incompatible interval for the same professional is refused, including under simultaneous requests.
Boundary between bookings	A booking may start exactly when the previous one ends.
Different professionals	Simultaneous bookings are possible without conflict between their calendars.
Cancellation	The interval becomes available again after a confirmed cancellation.
Repeated request	The same request is recognized without duplicating the booking or its effect.
Technical correspondence	The revision and environment are identified for the recorded results.

Q4 recognizes criteria met with sufficient Evidence. Verbal agreement between agents does not replace observing the relevant behaviour. Acceptance does not show that administrative work was reduced, and it does not on its own grant permission to change production.

Deployment and closing

S5 uses the accepted revision and a valid operational authorization. The action can happen under a delegated policy; if it loses confirmation, the flow first establishes the state of the environment. Repeating the action without knowing its effects is not an attempt that is automatically allowed.

A5 records the deployment and the agreed checks, and Q5 accepts the stage. Closing checks the commitment’s obligations, references and pending items before marking the unit complete, with booking deployed as its result.

The final check that was included is not permanent monitoring. Support, later corrections and recurring operation depend on the scope agreed. Benefits of use are observed separately and are not presumed from functional acceptance.

Legitimate alternative paths

Choice or event	How the result is treated
Close after C0 is accepted, without C1	Discovery concluded; construction was not contracted and is outside the terminal scope.
Defer the C1 offer	The offer is pending or expires according to its conditions; no construction is authorized.
Ask for an adjustment	The solution or the offer is revised under its own authorization; no purchase is presumed.
Pause an active C1	Confirm that work stopped and its effects, keep the context, and reassess conditions on resuming.
Close an incomplete C1	Record the interruption of the commitment and deal with the remaining obligations.
Buy reminders later	Examine a new offer and authorization, linked to the context without silently reopening the completed delivery.

Boundaries

The example is a proposed application of the model. It does not prove that AI is superior, that the model scales, that it lowers cost, that buyers will adopt it, or that it brings any economic benefit. It shows conceptual relations, not a forecast of margin. An empirical evaluation would have to include discoveries that end without construction, failed attempts and shared capacity, as well as the routes that run to completion. The five stages and the identifiers are illustrative conventions, to be adapted to the service contracted.