Role Scorecards

Two views of the same model: This page is the executive scoreboard. For the interactive explorer (open a role to see responsibilities and capability dependencies), see Role Replacement Explorer.

Executive scoreboard

Role scorecards

Role scorecards make the CEO test visible: if the CEO did not show up tomorrow, what would still happen?

These are honest maturity states, not measured percentages.

The only receipt-grounded role baseline today is the SDLC Orchestrator at ~40–55% full-role standing (see CURRENT_STATE). A role shows a percentage only when receipts support it; every other role reports a qualitative maturity state, not a number.

~40–55%SDLC Orchestrator

The only receipt-grounded role baseline. Six of seven weighted orchestration streams now run unattended in production.

90%+Target routine work

The target is not full autonomy; it is governed routine orchestration.

ReceiptsMovement standard

A percentage appears only when receipts prove a responsibility stopped requiring the executive.

SDLC Orchestrator

Current Replacement~40-55%

Target90%+

TrendMaterially past <25%; six of seven weighted streams unattended in production.

Expected Burden ReductionMedium - most covered-path orchestration is unattended; gaps in deployment-proof, receipts, and learning loop.

ResponsibilitiesDiscover Ready work, clarify, dispatch, monitor, verify, request review, prepare merge decisions, update receipts.

CapabilitiesIssue Discovery, Issue Clarification, Work Order Ingestion, Hermes Dispatch, PR Review, Merge Readiness, Receipt Generation.

EvidenceCURRENT_STATE, Work Orders D1-D3, Reference Realization receipts, Merge Readiness build evidence.

Remaining GapDeployment/standing-proof (Unknown), dry-run receipts, learning loop unproven, coverage limited to agent:candidate issues.

Context Carrier

Substrate built · not consumed

Current ReplacementSubstrate built · not consumed (no measured percentage)

TargetMost routine context-carrying

TrendPrerequisite substrate built, not yet consumed.

Expected Burden ReductionMedium-high - the executive still repeats architecture, decisions, and prior context.

ResponsibilitiesRetrieve relevant decisions, assemble issue/PR context, preserve corrections, surface contradictions.

CapabilitiesGBrain, Knowledge Ingestion, Context Assembly, Receipt Explorer.

EvidenceGBrain outbox and manual canon.

Remaining GapKnowledge documents do not yet enter reasoning.

Dispatcher

Dry-run only

Current ReplacementDry-run only (no measured percentage)

TargetThe bulk of routine dispatch

TrendMoving as Work Order ingestion and Hermes wiring land.

Expected Burden ReductionHigh - humans still wake and route the loop.

ResponsibilitiesConvert intent to work, select executor, carry authority, track state, follow up on stalls.

CapabilitiesWork Orders, Capability Requests, ExecutionRequests, Hermes Runner, Portfolio Health.

EvidenceD0 bridge, D1 endpoint, scheduler records.

Remaining GapD2-D4 and event substrate.

Status Checker

Prerequisite built · not wired

Current ReplacementPrerequisite built · not wired (no measured percentage)

TargetNearly all routine status checks

TrendImproves only when receipts and health are product-native.

Expected Burden ReductionMedium - humans still poll GitHub, CI, PRs, and agent outcomes.

ResponsibilitiesObserve state, summarize deltas, detect blockers, route health signals.

CapabilitiesOperational Health, Capability Portfolio, Executive Updates, Receipt Explorer.

EvidenceBurden Intelligence scheduler and dry-run projections.

Remaining GapLive health and delta surfaces.

Agent Coordinator

Blocked on execution plane

Current ReplacementBlocked on execution plane (no measured percentage)

TargetMost routine agent coordination

TrendBlocked on Hermes execution plane maturity.

Expected Burden ReductionHigh - humans still spawn and coordinate agents.

ResponsibilitiesPrepare plans, launch agents, monitor progress, request adversarial review, route outputs.

CapabilitiesHermes Dispatch, Planning Checkpoint, Adversarial Review Loop, Receipt Generation.

EvidenceExecution doctrine and Work Order chain.

Remaining GapPersistent runner and review loop receipts.

Memory

Designed · not operational

Current ReplacementDesigned · not operational (no measured percentage)

TargetThe bulk of routine recall

TrendDesigned, not operational.

Expected Burden ReductionMedium - repeated corrections remain human-mediated.

ResponsibilitiesRemember decisions, detect repeated feedback, propose constraints, improve capabilities.

CapabilitiesGBrain, Constraint Proposal, Knowledge Graph, Learning Loop.

EvidenceGlossary, ADRs, manual, receipts model.

Remaining GapConstraint and knowledge feedback into capabilities.