CUCA - AI Relationship Intelligence
Turn Forgotten Contacts into Predictable Business Opportunities.
Screenshots
Problem & Context
- Dunbar's Number Limit (150 relationships): Biological memory limits cause buffer overflow beyond 150 contacts, leading to lost contexts.
- Traditional CRMs track transactional sales pipelines; they lack active trust insights and dynamic relationship intelligence.
- Network Drift & Decay: Unbalanced interaction frequency, broken referral loops, and silent contact freeze over time.
- Static Data Limitation: Contact tools only display static information, leading to reactive instead of proactive relationship management.
Implementation Scope
- Research social network theories (Dunbar, Weak Ties, Social Capital) integrated with Reciprocity, Context Depth, and Response State.
- Design minimal, interactive UI/UX layouts optimized for Today's Trust Actions without dashboard complexity.
- Develop the Sandbox Pilot Decision Engine to simulate communication policies without risking real-world reputation.
- Build the unified operational layers: Data & Memory Layer (relational document storage), Deterministic Rule Engine, and AI Orchestration.
Deliverables
- Mobile App Version 1.0 (Today's Trust Actions, Attention Engine, Relationship Graph) launching June 2026.
- Seed-stage architecture and completed MVP under the technology backing of Laboon Digital Creation Technology.
- Sandbox Pilot Decision Engine interface generating action recommendations based on context depth, reciprocity, and user goals.
- CUCA product landing page promoting AI Relationship Intelligence positioning.
Results
- Finalized seed-stage architecture and completed MVP, ready for closed pilots with business communities (BNI, JCI, CEO Community).
- Optimized Vertex AI & Gemini integration for real-time entity extraction from 30-second post-meeting voice notes.
- Secured Google Cloud engineering support and cloud credits to offset LLM server and graph database costs.
- Established a solid roadmap: Launching 1.0 in 2026, expanding to agentic relationship workflows in 2027.
Features
Relationship Graph
Remembers meeting context, introduction sources, and actual trust levels instead of just static phone numbers.
Attention Engine
Automatically categorizes and prioritizes key connections like 'Benefactors' or 'Gatekeepers' needing your attention.
AI Action Planner
Proposes smallest, natural follow-up actions like saying thank you or updating outcomes with human-like, non-synthetic drafts.
Trust Governance
Automatically detects and warns about spam risks, pushiness, unnatural tone, or hidden commercial agendas before action.
Voice-to-Memory
AI automatically extracts events, promises, and pain points from a 30-second audio note to update relationship memory.
Privacy-First (4 NOs Commitment)
Commitment to 4 NOs: no auto-sending, no contact spamming, no sharing relationship graphs, no synthetic intimacy.
Challenges & Solutions
Challenges
Avoiding the traditional CRM design trap: keeping the UI clean and action-oriented rather than dashboard-heavy.
Preventing the AI from becoming a mass spam tool that risks damaging the user's personal reputation.
AI data extraction accuracy: ensuring the AI never invents context or upgrades trust levels without factual evidence.
User privacy friction during contact import: establishing the 4 NOs privacy architecture to build initial user trust.
Solutions
Designed the Today's Trust Actions screen as minimalist swipeable cards acting as a personal assistant, avoiding tables.
Integrated an 'Authenticity Risk Score' to automatically detect and prevent generic or overly polished messages.
Built a hybrid system: core labels are computed by deterministic rules, while AI handles entity extraction and drafts.
Implemented 3 ascending privacy modes (Manual-only, Limited, Full) and absolute transparency with user approval required for everything.
Minimalist interface styled in accordance with existing kientaoso.com guidelines, optimized for responsive layouts and performance.
Related Projects
Insider - AI Analytics
AI Analytics for sales data from Kiot, Sapo and Haravan. Query data in Vietnamese natural language with AI chatbot.
Mind DNA
5-in-1 personality analysis app: Numerology, MBTI, Enneagram, Tu Vi, Four Pillars and Physiognomy. Decode your personality through both Eastern and Western systems — from numerology and psychological tests to Eastern astrology.
Vietnamese Family
Vietnamese genealogy and family management application, helping to preserve and connect generations in the family.