Skip to content

Example: create and launch a product

This is an illustrative end-to-end workflow. It shows the five steps and the shared business records fitting together when a real outcome runs through them. Each numbered item below names the step it belongs to. Availability of each step depends on installed capabilities, connected services, permissions, and verified implementation.

The setting. Code Pineapple is a children’s book imprint. Its brand agent is Scout. The owner has noticed parents asking for something to keep six-to-ten-year-olds busy over a school holiday and wonders whether an activity-and-puzzle resource would sell.

  1. Research the opportunitystep 2 · talk to your agent

    The owner opens a conversation with Scout: “Is there a real audience for a holiday activity book, and what would it need to be?” Scout already has Code Pineapple’s positioning, past titles, and customer language in context. It runs a research flow against authorised sources, and comes back with findings, the assumptions it made, and a proposed business goal. The owner corrects one assumption in chat; that correction is recorded as a decision, not lost in the thread.

  2. Shape the product and offerstep 2

    Together they define the resource (a printable activity pack), the buyer (a parent, not the child), the promise, the price point, and the acceptance criteria — what “done and good” means before anything is made. A product record and an offer record now exist and link to each other.

  3. Plan the workstep 3 · it does the work

    Scout creates tasks under a project, selects the flow that fits a printable product launch, and lays out milestones. The plan names the points where a human decision is required (cover approval, price confirmation, publish). The owner approves the plan; it becomes version zero of the task record.

  4. Create the deliverablesstep 3

    Agents write, design, and assemble: the activity pages, a cover, a sales page draft. Each piece is produced by a named skill in a scoped working environment, and each run records what was invoked with what inputs. Image generation goes through the one media gateway, so every image has a receipt and a cost.

  5. Review and approvestep 4 · you review and publish

    The owner opens the task in the Workbench. Deliverables appear as cards; the cover opens in the same viewer as the sales page and the PDF. Feedback attaches to the specific version it was given on — “warmer palette on v2” — and Scout produces v3 without losing v2’s history. Approval is on a version, not on a name.

  6. Publish and distributestep 4

    The approved pack and sales page are published through supported connections. The publication record captures the external identifiers that came back — the listing, the page — so that results can be matched to what was actually shipped.

  7. Measure the outcomestep 5 · you measure

    Visits, inquiries, sales, cost, and feedback appear with their source and freshness visible. The dashboard says what it can prove (“forty-one visits from the sales page, six purchases”) and what it cannot (“attribution to the original research assumption is not observable”).

  8. Improve deliberatelystep 5 · it learns

    Scout proposes one product change and one capability change, each with the evidence behind it. The capability change — a tweak to the cover-design skill — is tested against the current version before anyone promotes it. The accepted version and the reasoning are kept; see The improvement contract.

flowchart LR
  AU[Audience need] --> P[Product] --> O[Offer] --> T[Task] --> R[Run] --> V[Artifact version] --> PB[Publication] --> RS[Results]
  RS -. next proposal .-> T

audience → product → offer → task → run → artifact version → publication → results. The model preserves these links where the underlying systems provide evidence, and states the gap where they do not.

  • The owner never chose a tool. They stated an intent, answered questions, and approved versions. Scout chose the flow and the skills, and said which ones it chose.
  • Every “yes” was on a specific thing. A version, a price, a plan — never “the cover” in the abstract.
  • The loop closed. Results fed the next proposal. Nothing was rewritten because one number looked good.

direction This example illustrates the intended workflow. Individual steps — research flows, the Workbench, versioned review, the media gateway, publishing connections — are at different stages; the release scope page is authoritative.

Sources: input/WilsonOS_Overview.html §04 · Last verified 2026-09-15