Source: narration-scripts.js (canonical Watch-track narration data)

What this source is

A 16,905-byte JavaScript file at public/js/narration-scripts.js that is the canonical narration cue data for every narrated video in the SuiteCentral 2.0 Watch track. The player at public/squire-v2-media-demo/watch/videos/player.html loads this file at runtime and uses the cue data to drive Web Speech synthesis — reading the text string at the start timestamp for each cue, so users hear narration through their browser’s TTS voice rather than pre-recorded audio.

This is how the wiki finally gets text content for the Watch track videos. Without this file, each video’s text would have to be extracted via STT (the approach we used for the AI Governance Layer hook video — see ai-governance-layer-video). For the other ~25 videos in the Watch track, this file already contains the narration text directly.

Notable absences:

• ai-governance-layer — has embedded audio narration, handled separately via STT
• installer-central — in the asset manifest (scripts/demo-asset-manifest.ts) but NOT in this file. The video exists at public/media/demos/installer-central-proximity.webm but has no narration text here. Reason unknown.

What IS in the file: narration for 8 scene/reel videos (executive-reel, scene1-problem, scene2-intro, storyboard-overview, ai-field-mapping (aka scene 3), scene4-governance, scene6-action-gate, scene7-opportunity) plus 2 Context Sidecar videos (context-sidecar + context-sidecar-highlight) plus 15 module demo videos.

Key claims

Seven-scene storyboard (per storyboard-overview and executive-reel)

1. The SuiteCentral 2.0 storyboard has exactly seven named scenes in a pre-defined narrative order. Per executive-reel cues, the mapping is: Scene 1 = Problem, Scene 2 = Intro, Scene 3 = AI Field Mapping, Scene 4 = Governance/Compliance, Scene 5 = Context Sidecar (the killer feature, NetSuite-native AP workflow), Scene 6 = NL Action Gate, Scene 7 = Opportunity. → three-review-paths and (NEW) demo-site seven-scene section.

Tier 1 / Tier 2 evidence taxonomy (per scene2-intro)

2. NEW taxonomy — the scene2 narration introduces a two-tier evidence framework:
   • Tier 1 evidence (governance): reasoning traces, compliance exports, competitive proof
   • Tier 2 evidence (execution): sidecar intelligence, action gating, integration readiness → suitecentral-2-overview (new section) — this is the first ingested source that explicitly names and categorizes the evidence types.

Context Sidecar — major new detail (per context-sidecar and context-sidecar-highlight)

3. NetSuite AP workflow is the canonical demo context: “Context Sidecar starts inside a NetSuite AP workflow, not a separate integration console.” → context-sidecar
4. Auto-refresh on focus change: “When invoice or vendor focus changes, context refreshes automatically with no tab switching.” → context-sidecar
5. Vendor context surfaces: risk score, compliance alerts, action-ready recommendations. First specific enumeration of what the sidecar presents in vendor context. → context-sidecar
6. Customer and invoice contexts show the same pattern (account health, receivables). → context-sidecar
7. Purchase order context surfaces “delay exposure before downstream schedules are impacted.” → context-sidecar
8. Debug panel confirms message flow for reviewer verification. → context-sidecar
9. Pause-payments mitigation is a named high-risk control — “High-risk states expose stronger controls, including pause-payments mitigation.” The highlight video adds: “High-risk conditions expose one-click controls like pause payments.” This is a concrete control action the sidecar can execute. → context-sidecar
10. Data provenance is explicit: service-backed responses, fallback behavior. → context-sidecar
11. Action dispatch is captured with operator intent and auditability. → context-sidecar
12. AI insights summarize trend signals: lead-time movement, open purchase exposure. → context-sidecar
13. Same governed interaction model runs across NetSuite and Business Central (second-source confirmation of dual-ERP claim). → context-sidecar and suitecentral-2-overview
14. MCP discovery and call paths are policy-scoped (confirms MCP architecture from ai-governance-layer-video). → suitecentral-2-overview
15. Full cut vs highlight cut framing: “Use this full cut for due diligence, and the highlight cut for fast executive review.” → demo-site — clarifies the two-video structure and who each version is for.

NL Action Gate (Scene 6) — architectural detail (per scene6-action-gate)

16. NL Action Gate architecture is explicit: “Commands are parsed through regex fast-path and LLM fallback for resilient intent handling.” Two parsing strategies: regex (fast, deterministic) for known commands and LLM (flexible, slower) as fallback for ambiguous commands. → (NEW) nl-action-gate
17. Allowlist filtering: “Action parameters are filtered by allowlist before any service dispatch occurs.” This is the security gate — even if the AI parses an action correctly, it cannot execute unless the action + parameters are on an explicit allowlist. → nl-action-gate

AI Field Mapping (Scene 3) — more detail (per ai-field-mapping)

18. Mapping approach: “semantic similarity and structural compatibility signals” — two named techniques the model uses. → suitecentral-2-overview
19. Confidence score + explicit rationale: “Each suggestion carries a confidence score and explicit rationale, not a black-box guess.” Confirms the confidence-scoring claim from read-elevator-pitch Beat 2.
20. High/low-confidence routing: “High-confidence rows move fast, while low-confidence rows are intentionally routed for human review.” This is the practical implementation of the human-in-the-loop gate.

Scene 4 governance — new detail (per scene4-governance)

21. Governance pacing + DLP differentiation (per Oracle comparison note): “The Oracle comparison reinforces explainability, governance pacing, and DLP differentiation.” Second-source confirmation of the Governance Pacer and DLP PII Shield as differentiators vs Oracle. → suitecentral-2-overview
22. SOC 2 criteria map to implemented controls: third-source confirmation of SOC 2 mapping claim (now confirmed in 11-role-brief-cto + read-elevator-pitch + here).

Scene 7 close — new detail (per scene7-opportunity)

23. SuiteApp.AI badge readiness: “tied to concrete evidence endpoints and governance artifacts.” First specific framing of what SuiteApp.AI certification depends on — live endpoints, not just documents. → pilot-30-60-90 (post-pilot path)
24. Decision shape: “approve pilot scope, owner, and timeline.” Three-part decision framing at close. Aligns with 13-pilot-30-60-90’s Day 1-30 “assign sponsor and owner.”

15 module demos (the module library)

25. The Watch track has individual narrated demo videos for 15 modules — contract-central, customer-central, customer-payment-portal, finance-central, inventory-central, mdm-central, payment-central, payout-central, portal-central, quality-central, service-central, supplier-central, sync-central, vendor-portal, workflow-central. Each video is ~12 seconds of narration with a consistent structure: (a) what the module does, (b) who uses it, (c) what operational problem it solves, (d) the outcome. → new concept page module-library catalogs all of them with their narration descriptions.

26. MDM Central — new detail (per mdm-central narration): “golden-record operating model”, “duplicate detection and merge workflows”, “stewardship queues for ownership and remediation.” → mdm-central

27. Missing module: installer-central has a video in the asset manifest (installer-central-proximity.webm) but no narration in this file. Means either (a) the video has no narration, (b) narration lives elsewhere, or (c) the module was added after narration-scripts.js was last updated. → noted on module-library

Pages updated by this ingest

Created (2 new pages):

• module-library — catalogs all 15 modules with narration-derived descriptions
• nl-action-gate — dedicated concept page (previously only mentioned as “one of the four named differentiators” in 01-executive-summary)

Updated (5 existing pages):

• context-sidecar — major update with NetSuite AP workflow detail, vendor/customer/invoice/PO context patterns, pause-payments mitigation, debug panel, AI insights, full-vs-highlight framing
• mdm-central — golden-record operating model, duplicate detection, stewardship queues
• suitecentral-2-overview — Tier 1 / Tier 2 evidence taxonomy section, AI field mapping technique names (semantic similarity + structural compatibility), cross-link to module-library
• demo-site — seven-scene storyboard section, full-vs-highlight sidecar video framing
• preston-test-repo — note that narration-scripts.js is the canonical Watch track narration data

Notable quotes

“Tier one evidence anchors governance: reasoning traces, compliance exports, and competitive proof. Tier two evidence anchors execution: sidecar intelligence, action gating, and integration readiness.” — scene2-intro cues [12]-[15]

“High-confidence rows move fast, while low-confidence rows are intentionally routed for human review.” — ai-field-mapping cue [12]

“High-risk states expose stronger controls, including pause-payments mitigation.” — context-sidecar cue [40]

“Commands are parsed through regex fast-path and LLM fallback for resilient intent handling. Action parameters are filtered by allowlist before any service dispatch occurs.” — scene6-action-gate cues [5]-[10]

“SuiteApp AI badge readiness is tied to concrete evidence endpoints and governance artifacts.” — scene7-opportunity cue [4]

“Bottom line: governance is embedded in execution, not bolted on during audits.” — scene4-governance cue [21]

Cross-references / contradictions found

• Confirms 16-module footprint from a new angle. Previous sources (01-executive-summary slide 3) claim 16 modules as 12 core + 4 extension/platform. This source enumerates 15 modules with narration + 1 module (installer-central) without narration = 16 total modules. The 12-core vs 4-extension split is still not mapped to specific modules.
• Confirms dual-ERP equal-citizens story (per context-sidecar cue [56]): “The same governed interaction model runs across NetSuite and Business Central.” Third-source confirmation (after read-talking-points and ai-governance-layer-video).
• Confirms MCP policy-scoping (per context-sidecar cue [60]): third-source architectural confirmation.
• Confirms SOC 2 mapping (per scene4-governance cue [6]): third-source (after 11-role-brief-cto and read-elevator-pitch). The scene4 narration uses “SOC two” (narration-friendly), not “SOC 2 Trust Services Criteria” — same claim, less specific framing.
• The installer-central narration absence is genuinely puzzling. The asset manifest shows it as a storyboard module with hasCaption: true (meaning it should have a caption file), but there’s no narration entry here. If this is intentional, the file is probably labeled as a storyboard clip rather than a narrated module demo. If unintentional, this is a minor gap in the Watch-track content. Flagged for mention on module-library but not treated as a contradiction.
• The “Suite Central two point zero” phrasing in scene2-intro cue [0] is a narration-friendly phonetic spelling of “SuiteCentral 2.0.” Not a product naming change.

Notes

• This file is the Web Speech synthesis data — narration is not pre-recorded audio. The browser reads the text aloud using a synthesized voice. That means the text is the authoritative source (there’s no pre-recorded version that could diverge from the text).
• public/js/narrator.js is the runtime that consumes this data (per 01-executive-summary asynchronous guide — noted earlier as docs/demos/EXECUTIVE-PACKAGE-V2-ASYNC-GUIDE.md). The two files are paired.
• The file has a Web Speech rate map (NARRATION_RATES) at the top that sets playback speed per video — most modules are 1.3x for speed, scenes are 1.08-1.35. This matters because the narration cue start times are wall-clock seconds at 1.0x speed; the browser speeds them up in playback.
• Individual module pages for the 14 modules without their own wiki pages are deferred. The narration content is ~12 seconds each, which is enough for a one-paragraph page but not enough to justify the filing overhead of 14 new pages. The module-library concept page catalogs them all with their narration descriptions. Future ingests (e.g., Connector Implementation Status or individual module feature guides) can justify promoting specific modules to their own pages.