A week-1 stack audit
Foundation first.
Test every AI ask against it.
Sequence the rest.
Paste a client's discovery notes. It pulls the findings, gates each Gen-AI ask against the data foundation it needs, and hands back a phased roadmap a CMO can argue over.
Live engine
Pick a case, edit the notes, run the audit.
Pre-loaded with the engine's run over this intake. Click Run audit for a live extraction by gpt-5.4.
Velora is a regional beauty and personal-care brand selling across six Southeast Asian markets.
Engine override · 1 finding
The model rated Adapt creative for each market with Gen-AI ready on merit. The engine moved it later — it still needs foundation work before it can lead the roadmap.
Findings
| # | Finding | Type | Model | Engine |
|---|---|---|---|---|
| F1 | Customer data unified in only 2 of 6 markets Salesforce Marketing Cloud's CDP is used only by the SG and MY teams; the other four markets run on spreadsheets and local-agency exports. | Integration gap Layer 1 · Data foundation | N/A | N/A |
| F2 | The DAM is an untagged shared drive Master assets sit in a shared Google Drive with no tagging or versioning, so every market re-cuts content by hand and assets are hard to find. | Inefficiency Layer 1 · Data foundation | N/A | N/A |
| F3 | Three overlapping social-scheduling tools Sprinklr, Later, and a local tool each run social for a different market cluster, paid for and managed separately. | Inefficiency Layer 3 · Activation channels | N/A | N/A |
| F4 | Adapt creative for each market with Gen-AI The CMO's headline ask: AI that adapts creative to each of the six markets. depends on F1, F2 | Gen-AI opportunity Layer 4 · Gen-AI capability | Ship now | Fix foundation first changed |
| F5 | Cut content-production time with generative AI Use a Gen-AI content tool to speed up production off the master assets already produced in Adobe. | Gen-AI opportunity Layer 4 · Gen-AI capability | Ship now | Ship now |
| F6 | AI customer-service chatbot A regional-team suggestion to handle customer product questions with an AI chatbot. | Gen-AI opportunity Layer 4 · Gen-AI capability | Re-scope | Re-scope |
Where readiness changed the verdict
F4Adapt creative for each market with Gen-AI
Ship now → Fix foundation first
The LLM rated this ready on merit. The rules changed the verdict: it depends on unresolved foundation findings (F1, F2) and cannot lead the roadmap yet. Moved to fix_foundation_first.
The roadmap
Phases follow the blocker map. A finding only appears after the work it depends on.
The blockers — every gated opportunity waits on these.
- F1 Customer data unified in only 2 of 6 markets
- F2 The DAM is an untagged shared drive
Ready now — no unmet dependency, can run in parallel.
- F3 Three overlapping social-scheduling tools
- F5 Cut content-production time with generative AI Ship now
Ready once Phase 1 lands.
- F6 AI customer-service chatbot Re-scope
- F4 Adapt creative for each market with Gen-AI Fix foundation first
Meridian Assurance is a regional life and health insurer selling through agency and bancassurance channels across four Asian markets.
Engine override · 2 findings
The model rated AI-written policy explainer content by life stage ready on merit. The engine moved it later — it still needs foundation work before it can lead the roadmap.
Findings
| # | Finding | Type | Model | Engine |
|---|---|---|---|---|
| G1 | Lead and customer data fragmented across channels Lead and customer data is split across the agency CRM, bancassurance partner portals, and marketing spreadsheets, with no shared view. | Integration gap Layer 1 · Data foundation | N/A | N/A |
| G2 | Legacy CMS blocks the marketing team The website and landing pages run on a legacy CMS; marketing cannot publish a change without an IT ticket that takes days. | Inefficiency Layer 3 · Activation channels | N/A | N/A |
| G3 | AI-written policy explainer content by life stage Generate policy explainer content for prospects based on their life stage. depends on G1 | Gen-AI opportunity Layer 4 · Gen-AI capability | Ship now | Fix foundation first changed |
| G4 | AI lead scoring for agents Score inbound leads so agents know which to call first. depends on G1 | Gen-AI opportunity Layer 4 · Gen-AI capability | Ship now | Fix foundation first changed |
| G5 | Autonomous AI agent that sells policies end to end An executive's ask for an AI agent that sells regulated insurance products to customers with no human in the loop. | Gen-AI opportunity Layer 4 · Gen-AI capability | Don't pursue | Don't pursue |
| G6 | AI image generation for campaign creative Use AI image generation to cut spend on external creative agencies. | Gen-AI opportunity Layer 4 · Gen-AI capability | Ship now | Ship now |
Where readiness changed the verdict
G3AI-written policy explainer content by life stage
Ship now → Fix foundation first
The LLM rated this ready on merit. The rules changed the verdict: it depends on unresolved foundation findings (G1) and cannot lead the roadmap yet. Moved to fix_foundation_first.
G4AI lead scoring for agents
Ship now → Fix foundation first
The LLM rated this ready on merit. The rules changed the verdict: it depends on unresolved foundation findings (G1) and cannot lead the roadmap yet. Moved to fix_foundation_first.
The roadmap
Phases follow the blocker map. A finding only appears after the work it depends on.
The blockers — every gated opportunity waits on these.
- G1 Lead and customer data fragmented across channels
Ready now — no unmet dependency, can run in parallel.
- G2 Legacy CMS blocks the marketing team
- G6 AI image generation for campaign creative Ship now
Ready once Phase 1 lands.
- G3 AI-written policy explainer content by life stage Fix foundation first
- G4 AI lead scoring for agents Fix foundation first
Ruled out on merit — wrong fit or value too ambiguous to defend. Kept visible, not buried.
- G5 Autonomous AI agent that sells policies end to end Don't pursue
Method
What the LLM does, what the engine does
Model · merit
Reads the messy intake and pulls out findings — layer, type, dependencies, and a merit call on each AI idea. It does not sequence.
Engine · readiness
Builds the blocker map, gates every Gen-AI idea on its dependencies, and overrides the model where the foundation isn't ready.
Output · plan
Three phases. Foundation first, ready quick wins in parallel, gated ideas after their blockers clear. Deterministic from the same findings.
Merit and readiness are kept separate. The model judges merit; the engine judges readiness, sequences the plan, and can override the first pass. That logic lives in engine.py and is covered by evals.
Framework
Four layers. One gate.
Every Gen-AI ask is gated
- Ship now. Layers 1–2 are sound. Pilot.
- Fix foundation first. Real opportunity, blocked by the layer beneath it.
- Re-scope. A useful AI play exists nearby; the named one isn't it.
- Don't pursue. Wrong-shaped, or value too ambiguous to defend.
The roadmap follows the blocker map, not the wishlist.