Aerial black-and-white view of a survey craft leaving a directional wake
Field report 01 / Review operations

Prepare the evidence. Keep the judgment human.

A field report from a governed template-review workflow: objective checks run in bounded lanes, evidence arrives in one packet, and reviewers retain every consequential decision.

Inspect the result All Field Reports
  • Evidence yield 98%
  • External writes 0
  • Promotion Blocked
Business hypothesis

Prepare the evidence before asking for judgment.

Precomputing objective evidence before reviewer judgment can increase review capacity while qualitative decisions and risky writes remain human-owned.

01 / Signal Prepare objective evidence.

Context, published-site checks, screenshots, and policy state arrive before judgment.

02 / Decision Keep uncertainty visible.

Subjective quality, exceptions, and consequential writes remain reviewer-owned.

03 / Proof Measure the boundary.

Evidence yield, false decisions, escalation, and missing recall stay attached to the run.

Operating boundary

Parallel evidence is useful only when uncertainty stays visible.

The workflow was designed as a coordinator with bounded lanes—not one agent pretending to own the entire review.

Operator intervention

The human reviewer owns the decision.

The system can gather, normalize, compare, and draft. It cannot silently approve, reject, contact a creator, or promote policy.

01 / Signal Submission context

Asset version, published surface, prior state, policy snapshot.

02AObjective validationDeterministic
02BRendered evidenceSandboxed
02CContext comparisonReview assist
03 / Proof Normalized packet

Findings, screenshots, provenance, escalation, and missing evidence.

04 / Decision Human reviewer

Confirm, override, request more evidence, or stop with a reason.

Measured result / Evidence collection

49 of 50 cases produced usable evidence.

The balanced calibration covered five outcome strata and eight reviewer buckets. It produced 98 screenshots—two per selected case—and made no external writes.

Usable packets
49 / 50
Measured
Evidence yield
98%
Derived / 98% from 49 / 50
Screenshots
98
Measured
Reviewer buckets
8
Measured
External writes
0
Measured
What changed for the team

Capacity begins with a better starting point.

The system does not remove judgment. It changes what is ready when judgment begins.

Reviewer
Start from an evidence packet.

Objective checks arrive organized so attention can move to visual quality, context, and the final decision.

Review lead
See where the system stops.

Escalations, overrides, failed gates, and policy questions become inspectable calibration work.

Technical owner
Replace lanes without hiding state.

Deterministic tools, Dify orchestration, MCP contracts, and model-assisted judgment remain separable.

Policy owner
Keep authority with named people.

Approvals, rejections, creator communication, and policy promotion cannot bypass a human gate.

Promotion blocked 2 / 2 missed Historical exceptional examples
Failed boundary / Judgment

The broader reviewer was not ready.

The eight-case multimodal run stayed clear of false approvals and false rejections in its small sample, but it missed both historical exceptional examples. The promotion gate was tightened and the reviewer remained blocked. The evidence collector stayed useful; the quality judge did not inherit authority it had not earned.

Evidence basis03 records

The claims stay attached to dated records.

These repo-owned artifacts contain the calibration shape, measured outputs, failed gate, and unresolved business measurement.

  1. 01 Calibration recordBalanced 50-case multimodal calibrationbalanced-50-multimodal-calibration-2026-05-27.md verified#TR-2026-01May 27, 2026
  2. 02 Calibration recordEight-case multimodal shadow evaluationmultimodal-8case-shadow-eval-2026-05-27.md verified#TR-2026-02May 27, 2026
  3. 03 Delivery reportSubmission quality loop report2026-06-05-submission-quality-loop-report.md review#TR-2026-03June 5, 2026
Measurement plan

Reviewer time savings are not measured.

The workflow demonstrates evidence capacity. Reviewer active-time savings still require a before-and-after pilot measurement.

Capacity calculation eligible submissions × (manual objective-check minutes − reviewer verification minutes)

Report the sample size, submission mix, and quality measures beside any result.

  1. 01

    Capture active minutes spent on objective checks before assisted review.

  2. 02

    Capture reviewer verification minutes after the evidence packet is available.

  3. 03

    Compare matched submission types and report the sample size with the result.

  4. 04

    Track false positives, missed objective issues, escalations, and reviewer overrides beside time.

System hierarchy

One method. One map. Three operating surfaces.

Performance Lab supplies the method. The mapping engagement names the boundary. Atlas holds the map. Signal, Decision, and Proof operate it. Substrate keeps the state and receipts owned.

MethodPerformance Lab

Train and test the workflow.

EntryMapping engagement

Name the operating boundary.

MapAtlas

Systems, owners, gates, and evidence.

OperateSignal / Decision / Proof

Watch, route, and record.

OwnSubstrate

State, policy, access, and receipts.

Apply the learning

Map the objective work before measuring the savings.

Start with one repeated workflow, a named decision owner, and a baseline the team can measure without exposing secrets.

Map one workflow Choose a mapping session
Owner
Workflow + decision owner
Authority
Human approval
Proof
Baseline + map + measurement plan
State
ready
  1. 01 / Map Separate objective work from judgment.

    Identify what can be prepared and what must stay human.

  2. 02 / Pilot Run the smallest controlled path.

    Collect evidence without expanding authority.

  3. 03 / Measure Compare active time and quality.

    Publish the result only after the sample exists.