Strategy Dashboard: The Unified
Intelligence View
5 minutes to 1 hour, by freshness need
sub-12s cold-cache via BI Engine
EU AI Act posture, all pillars, one view
The Strategy Dashboard is the platform-layer aggregation surface for The Autonomous Enterprise Platform. It reads from every process pillar currently running on the platform and presents a unified intelligence view for ClaraVis Medical Systems' C-suite. Current scope covers the Quote-to-Cash pillar and Governance Layer modules; Procure-to-Pay and Supply Chain panels are on the roadmap as those pillars deploy.
It writes nothing, decides nothing, and infers nothing. It is only meaningful because the data fabric beneath it — validated by Data Governance and scheduled by GreenOps — is running correctly. Five materialised views, each with its own refresh cadence and freshness SLO, feed a Looker Studio dashboard with drill-through links to the HITL queue that originated each flagged item.
§1 System Context
The Strategy Dashboard depends on all operational modules — and every additional pillar that joins The Autonomous Enterprise Platform adds a new data stream. It is the measure of how complete the platform deployment is, not the capstone of any single pillar. Current Quote-to-Cash scope: 5 panels. Procure-to-Pay and Supply Chain panels are planned for H4 / PI-9 as those pillars enter production.
§2 Panel Source Table
Each panel has a distinct data source combination, refresh cadence, and consumer. HITL counts are presented as read-only indicators with drill-through links to the originating module queue. Each panel carries a declared freshness SLO — the maximum permissible data age before Cloud Monitoring fires a staleness alert. The dashboard takes no action; it surfaces the signal.
| Panel | View | Sources | Refresh | Freshness SLO | Primary Consumer | HITL Drill-through |
|---|---|---|---|---|---|---|
| Pipeline Health | sd_pipeline_health | ae_ccai · ae_contracts · Salesforce REST | 15 min | ≤ 30 min | CTO · Head of Sales | HITL-01 · HITL-02 counts → originating queue |
| RevRec Posture | sd_revrec_posture | ae_financial · ae_finrisk | 5 min | ≤ 10 min | CFO · Finance Controller | HITL-04 queue depth → RevRec AI module |
| Fleet Intelligence | sd_fleet_intelligence | ae_devices + ae_governance quality flag JOIN | 1 hour | ≤ 75 min | CTO · VP Field Service | HITL-06 count → Asset IQ module queue |
| EU AI Act Compliance | sd_compliance_posture | All HITL counts · Vertex AI Model Registry · ae_governance | 1 hour | ≤ 75 min | CCO · Legal · CTO | All 11 HITL indicators → originating module queues |
| ESG / GreenOps | sd_esg_summary | ae_greenops.job_metrics · MTD aggregation | 1 hour | ≤ 90 min | Sustainability team · CFO | None — read-only ESG summary |
§3 Dashboard Mockup
The Looker Studio dashboard renders all five panels from BigQuery BI Engine-accelerated materialised views with a Looker Studio cache TTL of 15 minutes. The figures below reflect ClaraVis Medical Systems' actual Quote-to-Cash pipeline and fleet telemetry at current platform scale. Each panel shows its last refresh timestamp; HITL counts drill through to the originating module queue. Render time is sub-5 seconds on cached views; cold-cache render (after TTL expiry) is sub-12 seconds on BI Engine.
§4 Production Readiness — SLO · IAM · Observability
Every data-serving surface on the AE Platform carries a documented freshness SLO, an IAM access boundary, and a Cloud Monitoring alerting policy. The Strategy Dashboard is no exception. Five views. Five SLOs. One IAM boundary. Zero silent failures.
Each materialised view has a declared freshness SLO. SLO breach triggers a Cloud Monitoring alert policy → PagerDuty → on-call SRE. BigQuery BI Engine slot reservation: 1 GB across all sd_* views. Looker Studio cache TTL: 15 minutes.
IAM is enforced at three layers: (1) BigQuery dataset-level roles scoped to sd_* views only — no access to ae_* source tables. (2) Looker Studio workspace-level access lists per consumer role. (3) Column-level security on regulated fields (HITL audit records, revenue figures) via BigQuery column masking policies.
scoped to:
ae-platform.sd_views.*panels: all 5
scoped to:
ae-platform.sd_views.revrec, sd_views.esgpanels: RevRec Posture · ESG/GreenOps
scoped to:
ae-platform.sd_views.compliance, sd_views.revrecpanels: EU AI Act Compliance · RevRec Posture
roles/bigquery.dataViewer binding on ae-platform.ae_audit.* — no Looker seat required. The Strategy Dashboard compliance panel is a summary view; the primary evidence chain lives in BigQuery and is immutable.Five alert policies — one per materialised view — monitor refresh lag against each panel's freshness SLO. Breach → Cloud Monitoring notification channel → PagerDuty → on-call SRE. Stale views display a visible staleness banner in Looker Studio; they never silently surface outdated data as current.
Severity: Warning · Notification: Slack #ae-platform-ops
Severity: Critical · Notification: PagerDuty P1 · Finance on-call
Severity: Warning · Notification: Slack #ae-platform-ops
Severity: Critical · Notification: PagerDuty P1 · CCO on-call
Severity: Warning · Notification: Slack #ae-platform-ops
Stale data is never presented as current. The compliance panel showing 11/11 HITL checks compliant when the underlying view is four hours old would be a regulatory risk — the alert plus banner pattern prevents this.
§5 Architecture Decision Records
Three dashboard decisions, each chosen over a real alternative — with cost evidence.
Looker Studio can connect directly to BigQuery tables with no intermediate view layer. Rejected for three reasons. (1) direct connections expose the dashboard to schema migrations in source modules — if Asset IQ changes ae_devices, a direct connection breaks immediately, whereas materialised views provide an abstraction boundary. (2) direct queries run at dashboard load time, generating BigQuery slot consumption on every session; materialised views pre-compute on a schedule, amortising compute cost. (3) the materialised view layer is queryable by any BigQuery principal with read access — the same data the CTO sees in Looker is queryable via SQL, enabling regulatory audit access without a Looker seat.
Trade-off quantified: direct queries at 10 daily CTO sessions × 5 panels × ~500 MB scan each ≈ 25 GB/day on-demand BigQuery cost (~$0.125/day at $5/TB). Materialised view refreshes ≈ 5 scheduled jobs/day at ~200 MB each ≈ 1 GB/day ($0.005/day). The materialised view approach reduces query cost by ~95% at this session volume and eliminates schema-coupling risk entirely.
A single global refresh interval (e.g. 15 minutes for all panels) was considered. Rejected because the five panels have materially different staleness tolerances. The RevRec posture panel needs 5-minute freshness — the Finance Controller checks it multiple times daily and SLA compliance depends on queue visibility. The fleet intelligence panel is acceptable at 1-hour freshness because Asset IQ's HITL-06 notification is the operational alert path, not the dashboard; running fleet queries at 5-minute cadence produces unnecessary BigQuery compute cost with no operational benefit.
Cost delta: sd_fleet_intelligence at 5-minute cadence = 288 refreshes/day × 400 MB scan ≈ 115 GB/day ($0.58/day). At 1-hour cadence = 24 refreshes/day × 400 MB ≈ 9.6 GB/day ($0.048/day). Per-panel cadence saves ~$0.53/day on the fleet view alone — roughly $190/year — with zero operational loss, because HITL-06 is the real-time alert path for critical fleet events.
An alternative proposed surfacing HITL approval/denial buttons directly in the Strategy Dashboard. Rejected on architectural grounds. The Strategy Dashboard is a platform-layer read-only surface — adding write paths to HITL queues would require IAM changes, HITL audit trail modifications, and EU AI Act Article 14 logging currently scoped to the originating module. More practically, HITL decisions require context — the SHAP explanation, the confidence score, the flagged clause — that lives in the originating module's queue UI, not in a dashboard panel. The drill-through pattern is deliberate: it surfaces the signal at the dashboard level and routes the decision to where the context lives.
IAM risk of the alternative: adding write paths from the Strategy Dashboard would require expanding its IAM scope from roles/bigquery.dataViewer to roles/bigquery.dataEditor on ae_audit — a significant privilege escalation for a surface consumed by 20+ C-suite users. The read-only constraint is not a design limitation; it is an access control property.
§6 Stakeholder Rebuttals
Four objections raised during the dashboard design review. Each rebuttal is grounded in architectural evidence, not opinion.
data_quality_flag field from ae_governance. Units with stale features due to quarantine are shown with a grey "data unavailable" indicator — not a green/amber/red status. The panel caption notes "N units: data pending reinstatement" when the quarantine count is non-zero. Stale data is never presented as current. The compliance posture panel also surfaces the quarantine event — the Data Governance section shows the current quarantine count and steward resolution status. The dashboard degrades gracefully with visible flags.§7 Live Demo — Three Events
Watch the materialised views refresh and the panels update. A suite event arrives on the left; the relevant dashboard panel highlights and updates on the right.
| Event | Source | View Refreshed | Panel Updated | HITL Drill-through |
|---|---|---|---|---|
| Deal Closed · Pipeline Update | CCAI Sales Agent | sd_pipeline_health | Pipeline Health | 3 HITL-01 items → CCAI queue |
| RUL Alert · Fleet Event | Asset IQ | sd_fleet_intelligence | Fleet Intelligence | MCH-0042 · HITL-06 open → Asset IQ queue |
| HITL SLA Slip · Compliance Amber | RevRec AI | sd_compliance_posture | EU AI Act Compliance | HITL-04 SLA at 85% → RevRec AI queue |
§8 Adjacent Modules
Three platform modules — the foundation was first. Data Governance (H1 · PI-1) is the prerequisite that makes this dashboard meaningful. GreenOps (H3 · PI-7) contributes the ESG panel. The Strategy Dashboard (H3 · PI-8) closes the platform layer for the Quote-to-Cash pillar. Procure-to-Pay and Supply Chain panels follow in H4 · PI-9. Deployment order: M-08 → M-06 → M-07.