Intent Intelligence

Intent intelligence

Shared AI core for Google lead leakage

The AI architecture is intentionally simple: Roofing homeowner intent signals are the primary intelligence source, customer context is the second input, light examples are supporting guidance, and the three product modes are task templates on one shared base.

AI foundation policy

No live model execution is enabled yet; this is a typed architecture boundary only.

Execution mode

placeholder · provider none · live calls false

Primary intelligence source

c_end_intent_intelligence

Start from supplied C-end buyer-intent signals before drafting any lead-leak recommendation.

Use customer context to connect homeowner intent to visible Google Business Profile, review, brand/profile, landing-page, competitor, or intake gaps.

Use light examples only as phrasing or pattern guidance, never as the primary source of truth.

Keep output owner-ready for Roofing and HVAC operators, with Roofing as the active launch industry.

Route only exception cases to human review instead of treating human review as the core intelligence layer.

Core input layers

This replaces the previous broad knowledge-base framing with a smaller intent-led stack.

Layer	Priority	Purpose	State
C-end intent intelligence	primary	Roofing-first homeowner intent signals that explain why a buyer searches, what they fear, and which public trust-signal gaps may leak Google leads.	seeded
Customer context assembly	secondary	Per-business intake, Google profile URL, website, service area, competitor, snapshot, report, and operator-note context for one customer.	context_assembly_ready
Light example/case guidance	supporting	Small human-approved examples that help phrase findings and remediation patterns only after intent signals and customer context establish the actual issue.	placeholder

Roofing C-end intent seed

Representative high-priority homeowner searches show how buyer emotion maps to trust-signal, review/reply, profile, and landing-page gaps.

Keyword	Category	Emotion	Trust signal	Landing page
best roofing company near me	trust_legitimacy	The buyer is trying to reduce uncertainty by comparing visible local trust signals.	review_recency_and_sentiment	reviews_testimonials_page
emergency roof repair near me	urgency	The homeowner is worried about active damage and wants a credible roofer immediately.	emergency_readiness	emergency_repair_page
hail damage claim roofing contractor	insurance_claim_concerns	The buyer wants storm expertise and may be under pressure from deadlines or adjuster uncertainty.	insurance_claim_support	storm_damage_insurance_page
hail damage roof replacement	high_value_lead_intent	The homeowner may need a major replacement and claim support after a weather event.	insurance_claim_support	storm_damage_insurance_page
insurance roof inspection	insurance_claim_concerns	The homeowner is uncertain whether damage is claim-worthy and wants credible guidance.	insurance_claim_support	storm_damage_insurance_page
licensed insured roofer near me	trust_legitimacy	The homeowner wants to avoid risk, liability, and low-quality workmanship.	license_insurance_and_certifications	comparison_or_why_us_page

Task templates

Free Snapshot, Deep Audit, and Monthly Assistant are templates on the same base, not separate AI systems.

Template	Mode	Automation	Purpose
Free Snapshot template	free_snapshot	automatic	Prepare a fast Roofing Google lead-leak snapshot by matching urgent and high-priority homeowner intent to obvious public trust-signal gaps.
Deep Audit template	deep_audit	human_review_when_needed	Expand the same C-end intent signals into a deeper owner-ready audit with competitor context, prioritization, and evidence-backed fixes.
Monthly Assistant template	monthly_assistant	manual	Reuse the same intent-to-gap mapping for future maintenance prompts, review/reply suggestions, and recurring trust-signal checks without adding Monthly Watch flow depth here.

No heavy knowledge CMS or multi-agent platform

The current scope is a seeded intent library, lightweight retrieval helpers, task-template boundaries, and human review only for exceptions or paid review support.