{"title":"New Home Agentic OS — New Home Star","description":"The New Home Agentic OS is a continuously learning AI platform built by New Home Star that operationalizes decades of home builder transaction intelligence. It ingests sales calls, CRM data, and builder knowledge to create a shared intelligence layer that coaches agents, surfaces buyer insights, and compounds competitive advantage over time.","siteUrl":"https://agentic-os-kk3f2kyb.manus.space","crawlUrl":"https://agentic-os-kk3f2kyb.manus.space/crawl","lastUpdated":"2026-08-15T17:04:17.591Z","architecture":{"layerCount":8,"layers":[{"id":"capture","name":"Layer 1 — Expertise Capture","description":"Ingests institutional knowledge and structured buyer intelligence from five sources: Zoom (the centerpiece — client strategy sessions, sales coaching, internal planning, builder consulting), Google Drive (assessments, playbooks, templates, strategic documents), Gmail (decisions, follow-up actions, client context, historical communication), New Home Connect (training content, videos, learning materials), and Qualtrics (NHS TCP-model survey responses capturing Familial Status, Life Stage, Economic Condition, hot buttons, motivators, objections, and buyer intent). Years of recorded institutional expertise already exist in Zoom."},{"id":"classify","name":"Layer 2 — Classification & Taxonomy","description":"Every piece of captured content is classified against a rich NHS data model: Builder Classification (Production, Luxury, Build-On-Your-Lot, Active Adult), Buyer Classification (TCP profile, hot buttons, motivators, objections, life stage, household composition), Customer Journey (Lead → Prospect → Qualified Prospect → On-Site Opportunity → Purchaser → Homeowner → Evangelist), and Builder Product Taxonomy (Builder → Community → Floor Plan → Homesite → Features → Inventory Home)."},{"id":"expertise-library","name":"Layer 3 — Knowledge Library","description":"The organized institutional memory of New Home Star, structured by seven knowledge domains: Sales, Marketing, Recruiting, Training, Operations, Builder Consulting, and Leadership. The emphasis is on how NHS knowledge is organized — not what software stores it. Only Approved knowledge pages are embedded into the vector layer."},{"id":"vector-memory","name":"Layer 4 — Vector Intelligence Layer","description":"Transforms human-readable knowledge into machine-queryable intelligence via a four-step flow: Human Knowledge → Classification → Embeddings → Semantic Retrieval. AI retrieves knowledge by meaning, not keywords. A buyer who says they want to wait until rates drop gets the most relevant NHS response to that exact situation."},{"id":"strategic-reasoning","name":"Layer 5 — Strategic Reasoning Layer","description":"Claude Console serves as the reasoning engine between intelligence retrieval and execution. Before any execution happens, this layer analyzes retrieved intelligence, evaluates options, and produces a validated recommendation. Responsibilities: Planning, Analysis, Recommendations, Validation, Decision Support. The system thinks before it acts."},{"id":"execution","name":"Layer 6 — Execution Layer","description":"Manus is the execution environment. It takes validated recommendations from the Strategic Reasoning Layer and produces tangible deliverables: Applications (builder portals, buyer-facing tools, internal dashboards), Agents (Sales Coach, Buyer Intelligence, Community Positioning, Onboarding), and Deliverables (landing pages, CRO audits, referral programs, assessments, dashboards, builder portals)."},{"id":"governance","name":"Layer 7 — Human Review & Governance","description":"Security (PII Protection, Access Controls, Builder Permissions, Data Segmentation), Governance (Human Approval, Knowledge Validation, Version Control, Audit Trails), and Quality Control (Output Review, Feedback Loops, Accuracy Monitoring). Every output is reviewed by a human before delivery. Every rejection is a learning signal."},{"id":"feedback","name":"Layer 8 — Continuous Learning","description":"Every builder, buyer, meeting, campaign, survey, and closed deal creates new intelligence. Approved outputs are re-embedded into the vector index and saved to the knowledge library. The system becomes more valuable with every interaction — not just more data, but more structured, classified, and retrievable intelligence that compounds over time."}]},"techStack":[{"category":"Orchestration","tools":["n8n (workflow automation)","Make.com (backup orchestration)"]},{"category":"AI / LLM","tools":["OpenAI GPT-4o (reasoning)","Anthropic Claude 3.5 Sonnet (reasoning)","Whisper (transcription)","text-embedding-3-large (embeddings)"]},{"category":"AI Agent Tools","tools":["Claude Console (prompt engineering workbench for designing and testing agent prompts before deploying to n8n)"]},{"category":"Vector Database","tools":["Pinecone (semantic search and RAG)"]},{"category":"CRM","tools":["HubSpot (primary CRM, API integration)"]},{"category":"Call Intelligence","tools":["Zoom (call recordings, API)","CallRail (call tracking, API)"]},{"category":"Builder Data","tools":["New Home Connect (NHC API — sales pacing, buyer segmentation)"]},{"category":"Document Storage","tools":["Google Drive (training docs, strategy content)"]},{"category":"Database","tools":["MySQL / TiDB (structured data, knowledge artifacts)"]},{"category":"Application","tools":["React 19 + TypeScript (frontend)","Express + tRPC (backend)","Drizzle ORM (database layer)"]}],"compoundingLoop":[{"step":"01 Builder","description":"Builder partners connect their Zoom, HubSpot, CallRail, and NHC accounts."},{"step":"02 Agent","description":"NHS sales agents conduct calls, coaching sessions, and buyer interactions."},{"step":"03 Buyer","description":"Buyer signals — intent, objections, community preferences — are captured automatically."},{"step":"04 Interaction","description":"Every interaction is transcribed, classified, and embedded into the expertise library."},{"step":"05 Deal","description":"Closed deals and outcomes feed back into the system, making every future recommendation smarter."}],"valuePropositions":["Better buyer intelligence from every interaction","Faster sales agent onboarding with institutional knowledge","Institutional memory that survives agent turnover","Consistent coaching across all builder partners","Compounding competitive advantage — the system gets smarter every day","No more reinventing the wheel — one shared system replaces everyone building their own"],"dataSources":["NHC Training IP (proprietary sales methodology)","Zoom call recordings (API)","CallRail call logs (API)","HubSpot CRM (API)","New Home Connect (NHC API)","Google Drive documents","Gmail decision and relationship context","Qualtrics TCP-model survey responses (Familial Status, Life Stage, Economic Condition, hot buttons, motivators, objections, buyer intent)"],"primaryTools":["Zoom","CallRail","HubSpot","New Home Connect"],"builtBy":"New Home Star (NHS) — largest independent new home sales company in the US","faq":[{"q":"What is the New Home Agentic OS?","a":"An AI-powered operating system built by New Home Star that captures, classifies, and compounds sales intelligence from every builder interaction."},{"q":"Who uses this system?","a":"NHS sales agents, builder partners, and NHS leadership."},{"q":"What data does it ingest?","a":"Zoom call recordings, HubSpot CRM data, CallRail call logs, New Home Connect training and builder data, Google Drive and Gmail knowledge, and Qualtrics TCP-model survey responses."},{"q":"What AI models power it?","a":"OpenAI GPT-4o, Anthropic Claude 3.5 Sonnet, OpenAI Whisper, and text-embedding-3-large."}]}