PUBLIC SYNTHETIC EXAMPLE: no real participant or pilot. The private label below describes the report default; this constructed example contains no confidential participant record.

Return to the guide

H.A.R.D. Protocol 0.3 · Public Preview

Human-centered AI Readiness and Decision Protocol

Evidence supports a bounded engineering step

PRIVATE working record. Check permissions and identifying content before sharing.

SYNTHETIC-LOW-001 · Fictional advisor intake · 0.3-preview.1

Engineering recommendation only. Owner authorization and deployment evaluation are separate.

Review mode: engineering_commitment; artifact population: runtime_ai. Population remains separate from evaluator provenance.

DecisionProfile neededRiskFirst stop
PROCEED_TO_ENGINEERINGQUICK6lowNo gate stop

The software checks structure and decision rules. Recorded human evidence-quality review: RECORDED_PASS. The software does not independently verify evidence authenticity or certify criterion conformance.

Next action

Does the named owner authorize this exact bounded engineering step?

Owner: engineering lead. Record the owner's separate dated authorization or refusal and conditions. Do not infer authorization from a gate pass.

Six gates - no combined score

Gate and criteriaResultWhat is missing or failed
Understand today's work (HCAI-1.2)PASSRequired checks passed
Name the need and success condition (HCAI-1.1, HCAI-1.3, HCAI-1.4)PASSRequired checks passed
Model system truth, state and recovery (HCAI-3.1, HCAI-3.2)PASSRequired checks passed
Check the requirement and tested revision (HCAI-2.1, HCAI-2.2, HCAI-2.3)PASSRequired checks passed
Estimate the remaining human work (HCAI-3.3)PASSRequired checks passed
Bound the engineering decision (HCAI-4.1, HCAI-4.2, HCAI-4.3, HCAI-4.4, HCAI-4.5)PASSRequired checks passed

PASS means the structural rules and supplied review judgments satisfy this gate. It does not prove the evidence is true or sufficient in practice. MISSING means gather evidence; FAIL means repair a known problem; NOT_EVALUATED means no conclusion was drawn.

Effort, costs and performance

ResultValueInterpretation
Gross labor minutes saved / case15.0Before AI-output oversight
Oversight minutes / case5.0Review + correction + escalation + rework
Net labor minutes saved / case10.0Projected; not measured performance
Net operating benefit / period470.0USD / month; PROJECTED
Preparation elapsed minutes0.0Preparation time outside the review session
Review session elapsed minutes12.0Includes capture/reporting
Total evaluation person-minutes24.0Disjoint preparation + evaluator + participant + adjudication
Protocol evaluation cost12.0USD
Operational system performancenot_collectedNo deployment decision is made by this instrument

Missing values are not zero. Money and time are scenario estimates unless separately observed. The cost of this review is reported separately from workflow operating cost.

Current workflow - observed versus reported

StepRole and actionNext / endpointBasis
C1intake advisor: Read and check intakeC2, C4observed
C2intake advisor: Ask for correctionC3reported
C3intake advisor: Correct saved fields or retain cancelled recordC4reported
C4intake lead: Confirm or close intakeEndpointobserved

People and consequences

Affected roles: intake advisor, requester

Review areaApplicability and reasonRequirements / owner
access_usabilityapplicable: Stipulated synthetic case: readable editable drafts, equal correction access, local fictional data, advisor control. Not empirical evidence.R1 / intake lead
privacy_securityapplicable: Stipulated synthetic case: readable editable drafts, equal correction access, local fictional data, advisor control. Not empirical evidence.R1 / intake lead
unequal_effectsapplicable: Stipulated synthetic case: readable editable drafts, equal correction access, local fictional data, advisor control. Not empirical evidence.R1 / intake lead
human_agencyapplicable: Stipulated synthetic case: readable editable drafts, equal correction access, local fictional data, advisor control. Not empirical evidence.R1 / intake lead

Proposed workflow - connected paths

StateBehaviorNext / endpoint
S1Review requestS2, S4
S2Flag missing fieldS3
S3Retain corrected or cancelled intakeS4
S4Close with retained recordEndpoint

Consequential choices and accountable review

This ledger examines choices visible in the artifact against its purpose and criteria. Historical alternatives need source records. Proposed alternatives and new rationale belong to the current review. A justified existing choice can remain; replacement is not a prerequisite.

Declared consequential choice IDs: CHOICE-SYNTHETIC-1

CHOICE-SYNTHETIC-1: Synthetic fixture: preserve advisor control and contact details

Decision criteria: Synthetic criterion: preserve the defined fields on correction and cancellation; Synthetic criterion: require advisor approval before changes

Observed choice: Synthetic fixture: retain editable intake drafts instead of sending automatically

Observed evidence locations: examples/evidence/synthetic-workflow.md (synthetic artifact only)

Alternative categoryDescriptionSource evidence
Proposed in this reviewSynthetic current-review alternative: a non-AI form with field checksProposal; not historical evidence

Alternatives review: Synthetic current comparison; no claim about a creator's historical choices

Rationale provenance: new; justification: Synthetic stipulated current review: retaining editable drafts meets the preservation and approval criteria; retain this choice for a bounded prototype

Rationale evidence: No historical rationale evidence recorded

Impacts and tradeoffs: Synthetic tradeoff: advisor review adds time but retains control; no live sending

AssumptionStatus and consequence if falseEvidence / next evidence / revisit
A-SYNTHETIC-1: The retained draft remains the authoritative editable record until advisor confirmation.supported | Correction or cancellation could silently operate on stale or conflicting data.examples/evidence/synthetic-workflow.md | revisit: Any persistence, ownership or cancellation-flow change

Affected requirements: R1

Engineering deepening: required; basis: The choice mutates persistent draft state and its correctness depends on preservation and ownership assumptions.

Deepening triggers: persistent_state_mutation

Required decision surfaces: truth, state

Decision surfaceCurrent model and statusEvidence / next evidence
truth / DS-TRUTHThe synthetic intake draft record is authoritative until advisor confirmation. / supportedexamples/evidence/synthetic-workflow.md
state / DS-STATECorrection and cancellation preserve the same synthetic draft record before closure. / supportedexamples/evidence/synthetic-workflow.md
ChallengeCondition and invariantResult / evidence
CH-1Correction or cancellation interrupts the normal path | The same defined fields remain available after recoverypass / walkthrough | SYNTHETIC T1: stipulated walkthrough, not actual execution

Next coherent slice: Implement one disposable correction/cancellation path through persistence and recheck field preservation.

Human disposition: accepted; accountable owner: Fictional human reviewer (synthetic; no actual approval)

Human disposition evidence: SYNTHETIC owner decision: stipulated acceptance for software tests only

Verification evidence: SYNTHETIC T1: stipulated walkthrough, not actual execution

Follow-up: Synthetic plan: recheck field preservation after the disposable prototype changes; owner: Fictional intake lead (synthetic)

Unresolved gaps: No recorded structural gaps

Inspect requirement -> artifact -> check

R1: Correct or cancel without data loss

Success condition: All three paths preserve the defined fields

Behavior shown: simulated; this label is not inferred from visual fidelity.

Form: Review form with editable fields | Fit: Advisor intake workflow and email dependency | Function: Preserve details through correction/cancel

Reference records: E1

Artifact E1 · version fixture-1 · SHA-256 87a35fd15180404eedfd92f7c23b7afc6e35d8309ff1af4581ec356baaf2a5c7

Check T1: Stipulated three-path walkthrough | walkthrough | pass | tested revision matches: True | requirement/context matches: True

Engineering scope and owner decision

Owner: engineering lead

Engineering step: Build a disposable intake prototype

Resource cap: One engineer-day; no live sending

Revisit trigger: Review after prototype walkthrough

Owner authorization: pending separate dated record. A recommendation alone does not permit deployment.

Provenance and limits

Versions: protocol 0.3-preview.1; mcp 0.3.0rc1; skill 0.3.0-rc.1; contract 0.3.0-rc.1

Input SHA-256: 4ce86e99149cb20e072937002e5ed6010c85413aaf28bd0e819adbd2de971d0c

Previous run: Not recorded; revision: Not recorded

The report checks the structure of supplied evidence digests without fetching or authenticating the files. Practitioner correspondence and software tests do not establish empirical effectiveness.