Layer 04 · Prove · AI Business Value Engineering
Northbeam · Aurora Methodology
Outcome Against Projection
L01 projected. L03 delivered.
L04 measures what actually happened.
Outcome against projection. Variance classified. Four-party attested. No reverse-fit narratives.
L01 · Projected
$3.8M ARR at risk · 7 workstreams ranked
→
L02 · Specified
Charter locked · 17 AI safety decisions
→
L03 · Built
$2.94M retained · 26 saves
→
L04 · Measured
Faithful readout against baseline
Baseline · Locked at L01 Signoff
The projection L04 measures against — captured, hash-signed, immutable.

L04_BASELINE · v1

Captured at L01 signoff · 2026-05-23T01:30:00Z
Hash
sha256:demo-l04-baseline-…7f3c1b9e
Upstream L01 dossier
sha256:e6518d6e…
Upstream L02 charter
sha256:dcb07909…
Success metric
2pt churn reduction · Q4 FY26
Pre-Aurora baseline pool
$3.8M ARR at risk
Structural Rule 1 · Baseline body unchanged across readouts. Revision requires documented change order + 4-party re-attestation.
Pilot Readout · 2-week window · 2026-06-02 → 06-13
First measurement window. Mechanism live.
41
At-risk identified
canonical signal
36
Playbooks dispatched
5 auto-suppressed
26
Saves confirmed
63.7% save rate
$2.94M
ARR retained
of $4.62M cohort
Watermarked PILOT — first window, not yet steady-state. Annualization deliberately deferred.
Variance · 7-category rubric · Mutual exclusion enforced
Three variances. All positive direction.
workflow
0
no operating workflow changes
spec
0
charter acceptance criteria satisfied as written
build
+1
v-03 · ADR-016 type-level enforcement exceeded spec
model
+1
v-02 · AI loop exceeded conservative L01 attribution · critical magnitude
adoption
0
rolled into model variance per mutual exclusion
assumption
+1
v-01 · at-risk pool $3.8M → $4.62M (+21.6%)
macro
0
no macro variances
Structural Rule 2 · Variance Explained AGAINST Projection
+$2.84M positive variance. Classified MODEL — not "L01 sandbagged."
v-02 · model · critical magnitude
+$2.84M pilot pro-rata

The AI driver-attribution + dispatch agents combined performed at the upper edge of L01's conservative attribution split.

The L01 projection deliberately distributed value across workstreams conservatively. When WS-001 (canonical foundation) + WS-002 (attribution) + WS-003 (dispatch) combined in production, they captured value the per-workstream attribution had distributed cross-workstream.

The realized number is not the projection rewritten. The projection's conservatism is the explanation for why the realized is higher.

✓ AUDITOR · auditor-pilot-v1-clean · 0 reverse-fit narratives detected
Structural Rule 3 · Four-Party Joint Attestation
Four parties. One readout. One hash.
✓
Lin Zhao
CFO · Commercial economics
2026-05-23T02:25Z
✓
Diego Martinez
CRO · Commercial GTM
2026-05-23T02:26Z
✓
Maya Chen
CEO · Strategic
2026-05-23T02:27Z
✓
Bill Tennant
Northbeam · Methodology
2026-05-23T02:28Z
Readout hash · sha256:demo-l04-pilot-readout-v1-…9e8c7b6a · 4 of 4 signers attached · readout body locked
The Discipline Beat
What the attestation refuses to claim.
✕
NOTthe pilot save rate (63.7%) will sustain across subsequent cohorts
✕
NOTthe $2.94M ARR retained is annualizable
✕
NOTthe 5 deferred measurements have been measured (they're explicitly deferred)
✕
NOTthe 2pt churn-reduction success metric has been hit (measured at Q4 FY26 close)
✓
Attesting:the mechanism is live, the variances are classified faithfully, the trajectory framing for the board is grounded in measurement.
Aurora Dependency Watch · Armed at Pilot Attestation
Continuous structural observability.

7 ADR-compliance checks

✓ PASSADR-001 · no parallel definition references
✓ PASSADR-003 · disclosure protocol before publication
✓ PASSADR-006 · canonical YAML hash matches runtime
✓ PASSADR-008 · no opaque-ML imports
✓ PASSADR-015 · no bare-LLM in narrative
✓ PASSADR-016 · human-confirmation on external actions
✓ PASSADR-017 · no validation-gate bypass

Shadow-IT re-emergence + routing

✓ PASSRiya's Notion read-only + redirect
✓ PASSDiego SF renewal model deprecated
⏸ DEFERDevon suppression retirement (deploy+30d)
Alert routing · 4-tier
info
watch
material
critical
ADR violations + shadow-IT re-emergence = always material+, regardless of variance threshold
Net AI Business Value Engineering · Pilot window
$3.0M net pilot value at $63K/yr AI cost overhead.
$3.0M
Net pilot-window value
$3,002,923
Gross $3,007,769 − overhead pro-rata $4,846
$63K/yr AI cost overheadFully-loaded · disclosed transparently
Tokens
Infra
Ops
Maint.
Compl.
Tokens
$6K · 10%
Infra
$24K · 38%
Ops
$15K · 24%
Maintenance
$10K · 16%
Compliance
$8K · 13%
Aurora Chain · Fully Sealed
Four layers. Four hashes. One audit-grade chain.
L01 Discover
Dossier signed
sha256:e6518d6e…
→
L02 Specify
Charter locked
sha256:dcb07909…
→
L03 Build
Wave 1 verified
sha256:a8e3c120…
→
L04 Prove
Baseline + readout attested
sha256:demo-l04-baseline-…7f3c1b9e
Engagement now lives under continuous ADW observability with scheduled quarterly readouts. Every readout produces a fresh scorecard, fresh variance classification, fresh four-party attestation — or surfaces an ADW alert if drift is detected between scheduled windows.
01 / 10