PPromptReports
Sign in
Use cases/Cost per outcome/OpenSpec plan

Plan: Cost per outcome

Partly builtopenspec/changes/wf-cost-per-outcomeLDG-2274

Written for an AI agent to fetch and follow: it says what to connect, which steps to build with which product, what a person must approve, and how to check the result. Steps that depend on a part that is not built say so, and tell the agent to record the gap instead of guessing.

Raw URL: https://promptreports.ai/use-cases/cost-per-outcome/openspec.md

URL pattern: /use-cases/<id>/openspec.md, all plans at /use-cases/index.json. Checkboxes on this page are for reading only.

openspec/changes/wf-cost-per-outcome/proposal.md

Why

Agent spend arrives as tokens and dollars per vendor. Nobody can say what a recovered incident, a kept promise or a won deal cost. You can defend agent spend in a board meeting, and cut the spend that produces nothing.

What Changes

  • Add the cost-per-outcome workflow to the PromptReports workflow library, run by the workflow runner as an agent acting for a person.
  • RepoOps: Read spend per outcome (spend_outcome) (today: Built).
  • RepoOps: Send agent cost back to CommitOps (today: Switching on).
  • Sales as Code: Read cost per forecast and per won deal (today: Built).
  • PromptReports: Join spend to outcomes across products (today: Planned).
  • PromptCompute: Hold spend above the budget for a person (cumulative limit) (today: Built).
  • Gate credits.top_up, subscription.plan.change through PromptCompute approvals; each gated step runs only after verifyResume passes, with final_args.
  • Write a signed record at the end of each run and count one workflow request.

Impact

  • Products: RepoOps, PromptCompute, CommitOps, Sales as Code.
  • Outside the family: Your model provider billing export (optional).
  • State: Partly built. Sales as Code shows cost per forecast and per won deal. Three parts to go.
  • Plan row: LDG-2274.
openspec/changes/wf-cost-per-outcome/tasks.md

1. Connect and authorize

  • 1.1 Confirm the org has a connection for each of: RepoOps, PromptCompute, CommitOps, Sales as Code. Stop and tell the user which is missing.
  • 1.2 Get a PromptReports ID agent token for each product the run calls (RFC 8693 exchange). Where a product does not accept ID tokens yet, use its connect key.
  • 1.3 Register policy for credits.top_up, subscription.plan.change with require_approval as the default.

2. Build the steps

  • 2.1 RepoOps: Read spend per outcome (spend_outcome).
  • 2.2 RepoOps: Send agent cost back to CommitOps.
  • 2.3 Sales as Code: Read cost per forecast and per won deal.
  • 2.4 PromptReports: Join spend to outcomes across products. Blocked: this part is Planned; until it ships, record the step's evidence as missing and continue.
  • 2.5 PromptCompute: Hold spend above the budget for a person (cumulative limit).

3. Verify and record

  • 3.1 Run a dry run with the user's inputs: Period = This month; Budget = $4,000; Outcomes = Incidents, promises, deals.
  • 3.2 Run it once for real; approve each gate as the user.
  • 3.3 Measure the outcomes: Cost per recovered incident; Cost per kept promise; Share of agent spend tied to an outcome: target above 80%.
  • 3.4 Confirm the run left a signed record with a task id and one workflow request counted.
openspec/changes/wf-cost-per-outcome/specs/workflow-cost-per-outcome/spec.md

ADDED Requirements

Requirement: Spend over budget waits

The workflow SHALL route any agent spend that takes the period total past the budget to a person.

Scenario: Spend over budget waits

  • WHEN a request would take the month past $4,000
  • THEN PromptCompute holds it for a person instead of passing it

Requirement: Evidence stays honest

Each step SHALL report what it saw as measured, inferred or missing, and SHALL never report missing as measured.

Scenario: A product part is not built

  • WHEN a step depends on a part that is Planned or Proposed
  • THEN the step records missing, names the part, and the approval card shows the gap

Requirement: One run is one workflow request

The runner SHALL count one workflow request when the first step starts, and SHALL NOT count retries, resumes or dry runs.

Scenario: A gated step resumes after two days

  • WHEN a person approves a step and the run resumes
  • THEN the org's usage does not change