← The Genesis Care System← The evidence vault

RESEARCH.md

From the Genesis estate · ../../../docs/research/packages/crm_next_generation/RESEARCH.md

GENESIS RESEARCH SESSION: Next-Generation CRM & Relationship Intelligence — The Definitive Study

⚠️ SCALE NUMBERS BELOW MAY BE STALE. Live estate (S1418 probe, 2026-07-18): Neo4j ~28,458,058 nodes / ~86,871,006 rels · Qdrant 80 collections / ~19,863,151 points · agency 148 cells · API 10,960 paths · 82,693 commits. Re-probe live — never quote a baked total as eternal. Card: reports/s1418-fleet/docs/PROBE_SNAPSHOT.json.

Session ID: RS_2026-04-14_0245_crm_next_generation Date: April 14, 2026 Source AI: Claude Pro Max (Genesis Command Center) Carter Directives Verified: All 12 active and locked Genesis Health at Session Start: Neo4j 6.08M nodes | Redis 643K keys | Qdrant online | All 3 models serving | 110 daemons | 40 containers

EXECUTIVE SUMMARY

No CRM on the planet — out of 75+ platforms analyzed across five market segments — uses a graph database to model relationships. Every CRM stores relationships as rows in relational tables. This single architectural blind spot represents Genesis's most compelling product opportunity.

The CRM market ($101B in 2024, projected $263B by 2032) has erupted with $700M+ in AI-native startup funding since 2024, producing three new unicorns and a generation of founders who understand what Salesforce got wrong. But none have solved the fundamental data model problem. Genesis, with its 6.08M-node Neo4j knowledge graph, Qdrant vector search, and dual-LLM Actor-Critic architecture, is uniquely positioned to build the world's first graph-native CRM: the cardiovascular system that pumps relationship intelligence to every cell of the organism.

Recommendation: Build a graph-native CRM layer on existing Genesis infrastructure. Do NOT fork an open-source CRM. Do NOT buy a commercial CRM. Genesis already has 80% of what's needed. The missing 20% is schema design, email auto-capture, a relationship strength calculator, and a React frontend.

GROUND TRUTH — WHAT GENESIS ALREADY HAS

Before any research, we established live ground truth:

PHASE 1: THE FIVE WARRING TRIBES — COMPLETE LANDSCAPE

The CRM landscape has fractured into five distinct architectural philosophies. Each represents a different bet on what matters most.

TRIBE 1: THE INCUMBENTS

Control 60%+ of the market. Spending billions bolting AI onto 20-year-old architectures.

Salesforce — Agentforce - 9,500 paying customers, $540M ARR, 330% YoY growth - Atlas Reasoning Engine is LLM-agnostic - 84% autonomous case resolution in support use cases - Data Cloud essentially mandatory at $12K–50K/year extra - 77% B2B deployment failure rates due to data quality issues - Three different pricing models shipped in 18 months (per-conversation $2, per-agent $175K/yr, Foundations Edition $25/user) - Internal sales engineer reports: AI causing 10% seat reduction — threatening Salesforce's own per-seat revenue model - VERDICT: Impressive execution on a fundamentally constrained foundation. Architecturally trapped.

HubSpot — Breeze AI - Breeze resolves 50%+ of support tickets - Content Hub generating demand - Free tier still functional, $3,600/mo for Enterprise - AI features genuinely useful but incremental, not transformative - VERDICT: Best entry-level CRM with real AI. Not reimagining relationships.

Microsoft Dynamics 365 — Copilot - Deepest workflow integration via Outlook, Teams, SharePoint, LinkedIn - Dataverse as underlying platform - $65–135/user/month - VERDICT: Wins on ecosystem lock-in, not on relationship intelligence.

Zoho CRM — Zia AI - Built its OWN proprietary LLM (7B parameters) - AI included in ALL plans at $14–52/user/month (3–10× cheaper than competitors) - 16 billion API calls monthly, 50% YoY growth - Zia Agent Studio with 700+ predefined actions - MCP server for 15+ Zoho apps - 55-app ecosystem - VERDICT: The value play nobody should ignore. Enterprise perception is the only weakness.

Other Incumbents Evaluated: - Pipedrive: Visual pipeline focus, AI limited to lead scoring and email writing. $14–99/user. - Freshsales: Freddy AI for lead scoring and forecast. $9–59/user. Competent but unremarkable. - SugarCRM: Open-ish heritage, PE-backed, on-premise option. $49/user + add-ons. Stagnating. - Monday CRM: Work OS approach, flexible. $12–28/user. CRM module feels bolted-on. - Nimble: Social CRM, enrichment from social profiles. $25/user. Small but focused.

TRIBE 2: AI-NATIVE CHALLENGERS

Building CRM from scratch with AI at the core. The most relevant tribe for Genesis.

Rox AI — $1.2B Valuation (Sequoia + General Catalyst) - Deploys HUNDREDS of specialized AI agents per customer account - Knowledge graph + data lakehouse architecture - Customers: Ramp, OpenAI, MongoDB - $8M ARR projected at under 2 years old - Fastest CRM startup to reach unicorn status - RELEVANCE: Proves multi-agent CRM works at enterprise scale. Validates Genesis's agent architecture.

Reevo — $80M (Khosla + Kleiner Perkins) - Spans marketing, sales, AND customer success in one "stackless" platform - Most ambitious single-product bet in AI CRM - RELEVANCE: Shows the market rewards unified platforms over point solutions.

Day AI — $24M (Sequoia) - Founded by HubSpot's former CPO (10 years) - "The Cursor of CRM" — ingests all workday exhaust - Claims to know everything about your business in 15 minutes - 120+ customers, Pat Grady on board - Built on Anthropic's Claude - RELEVANCE: Proves unstructured-first approach works. Reference for OMEGA integration.

Clarify — $22.5M - Pioneered "autonomous CRM" — platform is FREE, pay only when AI agents complete tasks - "Ambient Intelligence" runs 24/7 with zero input - Time-series AI models for relationship mapping - RELEVANCE: Most customer-aligned pricing model. Time-series approach relevant to temporal decay.

Lightfield — By Tome's Founders - Keith Peiris + Henri Liriani killed Tome (25M users, $300M valuation) to build this - Starts from UNSTRUCTURED conversation data and DERIVES structure - Schema-less from day one — exact opposite of traditional CRM - 2,500 companies in first 3 months, 100+ YC startups, hundreds migrating from HubSpot - One-hour CRM migration agent launched March 2026 - RELEVANCE: Architecturally the closest to what Genesis should build. Unstructured → structured maps to OMEGA → Neo4j.

Sumble — $38.5M (Kaggle Founder) - Anthony Goldbloom (created Kaggle, sold to Google 2017) - Operates 2.6M-company KNOWLEDGE GRAPH - Enriched continuously from hiring data, company websites, social feeds, regulatory filings - Customers: Snowflake, Figma, Wiz, Vercel - RELEVANCE: THE most architecturally relevant platform to Genesis. Proves knowledge graph approach works for CRM. Key difference: Sumble maps companies; Genesis would map relationships between PEOPLE — a harder, more valuable problem.

Monaco — $35M (Founders Fund) - Founded by Sam Blond (former Founders Fund VC) - AI agents supervised by experienced human salespeople EMPLOYED BY MONACO - Hybrid human + AI model - Angels: Stripe's Collison brothers, YC's Garry Tan - Peter Thiel: "No product sells itself — though Monaco comes close." - RELEVANCE: Proves AI alone can't close deals. Human-in-the-loop matters.

Other AI-Native Startups Found: - Aurasell ($30M) — unified enriched data model - item (YC Fall 2025) — train AI agents by writing process documents - Pally (YC S25) — unified inbox + personal CRM - Persana (YC) — multi-agent GTM - Fixture (YC) — AI-native CRM - Veles (YC) — AI CPQ - Sweep ($22.5M) — agentic Salesforce layer - DevRev — knowledge graph CRM for product companies - Pylon ($51.3M) — post-sales omnichannel

TRIBE 3: OPEN SOURCE INSURGENTS

Architectural transparency and self-hosting. Critical for understanding what's possible.

Twenty CRM — 39,700+ GitHub Stars (Leader) - TypeScript/React/NestJS monorepo (Yarn workspaces + Nx) - Both GraphQL AND REST APIs - TypeORM for data access, PostgreSQL primary store - Custom object system (Model Studio) — any entity type with arbitrary fields/relationships - Emerging MCP server (jezweb/twenty-mcp, 29 tools) - 300+ contributors, daily commits - Marmelab 2026 benchmark: 9/10 (highest score) - License: AGPL-3.0 (derivative works must remain open source) - RELEVANCE: Best UI/UX reference for Genesis CRM frontend. GraphQL schema maps to Neo4j GraphQL Library. Do NOT fork — use as architectural reference.

Erxes — GraphQL Federation Architecture - Apollo Router + tRPC microservices - Module federation micro-frontends - MongoDB + Redis + BullMQ stack - CRM is one "plugin" alongside marketing, operations, sales - RELEVANCE: Federation architecture could theoretically federate a Neo4j GraphQL endpoint as additional service. Cleanest integration path of any open-source CRM.

Corteza — Apache-2.0 License - Most permissive license of any open-source CRM - Go backend (superior performance for graph-heavy operations) - Federation via ActivityPub - "Aire AI" generates data models from natural language - Commons Conservancy Foundation ownership — can never be commercially captured - RELEVANCE: If Genesis ever needed an open-source CRM foundation, this is the license-safest option. Go backend is performant.

Krayin — MIT License + AI Module - Laravel-based, MIT licensed - Official AI module powered by LLMs via OpenRouter - Clean architecture, good for SMEs - RELEVANCE: Shows how to add AI to a traditional CRM via modules. MIT license is fully permissive.

Other Open Source CRMs Found: - Monica (22K stars) — personal relationship manager, data model maps perfectly to graph structure - SuiteCRM — Sugar-derived, PHP, benchmark 5/10, aging - EspoCRM — Entity Manager system, PHP, benchmark 7/10 - Huly — TypeScript/MongoDB, project management + CRM - Odoo CRM — ERP-integrated, LGPL, massive ecosystem - Atomic CRM — MIT license, 15K LOC, built-in MCP server, benchmark 8/10 - iDURAR — MERN stack open source CRM - YetiForce — security-focused - Dolibarr — micro-business - CiviCRM — nonprofit - vtiger — legacy - ERPNext CRM — Frappe-based - NocoBase — no-code platform with CRM module

KEY FINDING: No open-source CRM uses a graph database. Not one. Not even experimentally. Multiple developers have documented attempting graph-based CRMs and finding them "architecturally correct but practically unshippable" due to complexity.

TRIBE 4: RELATIONSHIP INTELLIGENCE SPECIALISTS

Focus on mapping relationship networks rather than tracking transactions.

Affinity — $120M Raised, $600M Valuation - Serves 50% of top VC firms, 96% retention rate - Auto-syncs email and calendar data across entire firm - Enriches from 40+ external sources - Detects inferred connections through shared work history - Scores relationship strength based on recency and frequency - API v2 (2024) is 100× more efficient than v1 - MCP connector announced 2025 - $2K+/user/year pricing - RELEVANCE: Category-defining platform. Proves market for relationship intelligence at premium pricing. Genesis can surpass by using actual graph database and five-dimension relationship strength.

Introhive — Intelligence Suite (October 2025) - Proactive relationship alerts - Alumni network monitoring - Relationship pathways based on shared experience - 90% data accuracy through passive email/calendar capture - RELEVANCE: Proves automated relationship intelligence works for professional services.

Louisa AI — Goldman Sachs Spinout - Maps "who-knows-who" across 150,000 users at Fortune 500 companies - Claims 33× warm intro reply rates - Oxford/Barclays backed - RELEVANCE: Enterprise-scale network mapping validates the graph approach.

4Degrees — Network Graph Visualization - Investment-focused relationship mapping - Frequency-based scoring - RELEVANCE: Shows demand for visual relationship graphs in financial services.

Other Relationship Intelligence Platforms: - Dex ($12/month personal CRM) - Clay ($204M raised, $3.1B valuation — data enrichment as relationship intelligence) - Boomerang AI (relationship intelligence beyond LinkedIn) - Nektar.ai (buying group detection) - Rings AI (investment intelligence) - Nat.app (personal PRM) - Regards (relationship visualization) - Covve (mobile personal CRM) - Cloze (algorithmic follow-ups)

TRIBE 5: REVENUE INTELLIGENCE PLATFORMS

Capture conversation and activity data at scale.

Gong — $300M ARR - Revenue Graph is closest to true relationship intelligence layer - "Mission Andromeda" (February 2026): MCP interoperability, AI sales coaching - Not becoming a CRM — becoming the intelligence layer ABOVE CRM - RELEVANCE: Potential data source for Genesis rather than competitor. MCP integration path exists.

Clari + Salesloft — Merged December 2025 - $510M+ combined funding - Unified revenue orchestration managing $5T+ in observed revenue - RELEVANCE: Shows market trend toward consolidation and unified platforms.

Other Revenue Intelligence: - Outreach — sales execution - People.ai — activity capture - Chorus (acquired by ZoomInfo) — conversation intelligence

PHASE 2: DEEP EVALUATION — TOP 20

EVALUATION CRITERIA (The "Carter Test")

For each platform, we evaluated: 1. Core philosophy — relationships vs transactions? 2. AI capabilities — real vs marketing? 3. Data model — rigid vs flexible/graph? 4. Automation — what runs without human input? 5. Integration — API quality, Neo4j/Qdrant connectable? 6. Mobile — real or afterthought? 7. Self-hosted — can you run it on your own infrastructure? 8. Pricing — cost vs value 9. Uniqueness — the ONE thing nobody else does 10. The Carter Test — does it feel like data entry or relationship management?

TIER 1: ARCHITECTURALLY TRANSFORMATIVE FOR GENESIS

#1 Sumble — Most architecturally relevant. 2.6M-company knowledge graph. Validates graph approach works for CRM at scale. Maps companies; Genesis maps relationships between people.

#2 Twenty CRM — Best open-source reference. 9/10 benchmark. GraphQL + REST APIs map to Neo4j GraphQL Library. MCP server exists. 39,700+ stars, daily commits. AGPL license means reference only, not fork.

#3 Rox AI — Most successful multi-agent CRM. $1.2B valuation. Hundreds of agents per account. Knowledge graph + data lakehouse. Proves Genesis's agent architecture is the right bet.

#4 Lightfield — Most radical insight: start from unstructured data, derive structure. Maps directly to OMEGA pipeline + Qdrant. Schema-less from day one.

#5 Affinity — Category-defining relationship intelligence. 50% of top VC firms. Auto-capture from email/calendar. Measures recency × frequency. Genesis surpasses with five-dimension model.

TIER 2: STRONG FEATURE OR STRATEGIC RELEVANCE

#6 Attio — Most flexible commercial data model. $141M raised, Google Ventures. Custom objects without rigid schemas. MCP server + App SDK. Sub-50ms latency.

#7 Clarify — Most customer-aligned pricing (free platform, pay per task). Ambient Intelligence with zero input. Time-series relationship models.

#8 Day AI — HubSpot CPO's revenge. $24M from Sequoia. Knows your business in 15 minutes from workday exhaust.

#9 Creatio — Only new-wave CRM with on-premise deployment. $1.2B valuation. Composable low-code architecture. 19 industry verticals.

#10 Close — Native MCP server. Built-in VoIP/SMS/email. "Chloe" AI sales agent that calls, qualifies, books, follows up. Bootstrapped pragmatism.

#11 Erxes — GraphQL Federation with Apollo Router. Could federate Neo4j endpoint as additional service. Cleanest integration path.

#12 Corteza — Apache-2.0 license. Go backend. ActivityPub federation. "Aire AI" data model generation.

#13 Salesflare — Best automatic data capture. Email signature scraping, LinkedIn pulling, calendar logging, zero manual input. €350K seed, bootstrapped independence.

#14 Salesforce Agentforce — $540M ARR demands respect. 84% autonomous resolution. But 77% deployment failure rate and complexity explosion.

#15 Zoho — Own LLM, AI included at $14/user, 16B API calls/month. Value king.

#16 Gong — $300M ARR Revenue Graph. MCP interoperability. Intelligence layer above CRM.

#17 Monica — Open source personal CRM. Data model maps PERFECTLY to graph: contacts as nodes, relationships as typed edges, activities as temporal edges. Best schema reference.

#18 Introhive — Proactive relationship alerts. Alumni network monitoring. 90% data accuracy via passive capture.

#19 Monaco — $35M, AI + human hybrid. Proves human-in-the-loop matters for closing deals.

#20 item — YC Fall 2025. Train agents via process documents. Most aligned with Genesis's documentation-driven architecture.

WHAT EVERY RELATIONSHIP INTELLIGENCE PLATFORM GETS WRONG

The entire industry measures relationship health using the same primitive metric: recency × frequency of contact. This is like measuring cardiovascular health by counting heartbeats.

Five critical dimensions remain unmeasured by ANY platform on the market:

  1. SENTIMENT — Whether the last interaction was positive, neutral, or contentious. A customer emailing angrily every day has high frequency but a deteriorating relationship. Genesis's dual LLMs can perform sentiment analysis on every communication.

  2. RECIPROCITY — Whether effort flows in both directions. One-sided relationships where you always initiate signal a fundamentally different dynamic. Graph analysis computes this trivially: out-degree/in-degree on communication edges.

  3. DEPTH — Distinguishes superficial pleasantries from substantive discussions. A 30-second "happy holidays" email ≠ a 90-minute strategy deep-dive. Qdrant semantic search classifies interaction depth automatically.

  4. NETWORK POSITION — Whether a contact is peripheral or a critical bridge between clusters. Neo4j's built-in graph algorithms (PageRank, betweenness centrality, community detection) compute this directly.

  5. TEMPORAL DECAY — Different relationships weaken at different rates. A close friend survives months without contact; a new prospect goes cold in days. Configurable half-life per relationship type, same model as GRL decay.

The market gap is stark: No platform answers the questions that matter most: - What would happen to our deal pipeline if this person left? - Who are the hidden bridges between our client clusters? - What's the shortest warm introduction path to Target Company's CEO? - How resilient is our relationship network to the departure of any single contact?

These are graph traversal problems — trivial in Cypher, impossible in SQL.

THE GENESIS CRM VISION: CARDIOVASCULAR SYSTEM

Traditional CRM is a blood bank — stored, static, requiring manual infusion. Genesis CRM is a cardiovascular system — actively pumping relationship intelligence to every part of the organism.

The Heart: Neo4j Knowledge Graph Every person, company, interaction, deal, project, and document becomes a node. Relationships are first-class edges with properties: strength (0–1), sentiment (−1 to +1), reciprocity ratio, last interaction timestamp, interaction count, relationship type. The existing 6.08M nodes become the foundation. Consulting research on clients automatically flows into their relationship record because it already lives in the graph.

The Blood: Qdrant Semantic Intelligence Every email, call transcript, meeting note, and document processed through OMEGA gets embedded via Qwen3-Embedding-8B. Semantic search enables: "find every conversation where a client expressed frustration about delivery timelines" or "which prospects discussed topics similar to our conference presentation?"

The Lungs: Dual-LLM Processing Qwen3.5-397B (Actor) processes: email summarization, follow-up drafts, sentiment analysis, meeting prep briefs, relationship strength explanations. GLM-4.7-355B (Critic) validates: quality-gates auto-generated content, verifies sentiment assessments, detects hallucinations, flags anomalies. Actor-Critic ensures the CRM never sends an embarrassing auto-generated email.

The Nervous System: Predictive Intelligence Graph neural networks on Neo4j data predict: weakening relationships (pre-churn detection), dormant connections to reactivate (opportunity detection), optimal introductions (network intelligence), outreach timing and channel (behavioral modeling).

The Heartbeat: Automated Maintenance Daemon Daily daemon scans the relationship graph for: decay threshold crossings → follow-up suggestions, life events/job changes → congratulations, stuck deals → escalation alerts, churn-predictive patterns → preemptive engagement.

BUILD vs BUY vs EXTEND — THE DEFINITIVE RECOMMENDATION

BUYING FAILS

No commercial CRM exposes its data model at the graph level Genesis needs. Attio's custom objects are relational tables underneath. Affinity's relationship intelligence is proprietary. Salesforce/HubSpot are architecturally incompatible.

FORKING FAILS

Twenty's AGPL-3.0 would contaminate Genesis's codebase. Replacing PostgreSQL with Neo4j inside an existing CRM means rewriting the entire data access layer — at which point you've built a new CRM anyway.

BUILDING WINS

Genesis already has 80% of the infrastructure: - Neo4j (6.08M nodes) → relationship graph, running - Qdrant → semantic search, running - Dual LLMs → sentiment/summarization/generation, serving - OMEGA pipeline → document ingestion, built - Redis (643K keys) → caching/alerts, running - MCP server → tool integration, 40+ tools - RedPanda → event streaming, deployed - 935 Python files with Neo4j imports → deep graph infrastructure

WHAT TO BUILD (the missing 20%):

  1. Contact/Company/Deal node schema in Neo4j (~1 Cursor session, pure Cypher)
  2. Email/calendar auto-capture daemon (~2–3 sessions, Gmail API + Neo4j)
  3. Five-dimension relationship strength calculator (~2 sessions, Python + graph algorithms)
  4. GraphQL API via Neo4j GraphQL Library (~1 session, schema file + config)
  5. React frontend inspired by Twenty's patterns (~3–5 sessions)
  6. Enrichment pipeline with Clay-style waterfall enrichment (~2 sessions)
  7. The Heartbeat daemon for daily relationship scanning (~1 session)

PHASED ROLLOUT:

INTEGRATION ARCHITECTURE WITH GENESIS

Neo4j ← → GraphQL: Neo4j GraphQL Library v5+ auto-generates full API from .graphql schema with @relationship directives. Define Person, Company, Deal, Interaction, Communication types. Queries, mutations, filtering, sorting, pagination — all auto-generated. Existing 6.08M nodes queryable alongside CRM data through same API.

Qdrant ← → Semantic Search: Dedicated crm_communications collection (4096-dim cosine). Every communication embedded via Qwen3-Embedding-8B (port 8014). Natural language queries across all relationship history.

OMEGA ← → Document Ingestion: Client documents flow through existing 9-layer pipeline. Extracted entities automatically linked to CRM nodes in Neo4j. Consulting research becomes part of client relationship record with zero data entry.

Dual LLMs ← → Intelligence: Actor (8010) handles summarization, drafts, sentiment, prep briefs. Critic (8011) handles quality-gating, verification, hallucination detection, anomaly flagging.

MCP Server ← → External Access: Extend with CRM tools: search_contacts, get_relationship_strength, find_warm_path, get_deal_pipeline, suggest_follow_ups.

RedPanda ← → Real-time Events: Email arrivals, calendar updates, deal stage changes, threshold crossings publish to topics. Downstream consumers trigger Heartbeat alerts, enrichment updates, strength recalculations.

MARKET DATA AND COMPETITIVE INTELLIGENCE

Genesis structural cost advantage: Every AI-native CRM burns OpenAI/Anthropic API credits per agent. Genesis deploys agents on its own 8× H200 GPUs (1.15TB VRAM). This compounds with every agent deployed.

Research conducted by Carter Hill, Day 7 Public Benefit Corporation Genesis Command Center — April 14, 2026