Human+
A culture intelligence platform with AI-native diagnosis. Quantum Leap built the architecture, the agents, and the ICV 3.1 model. The platform is in production.
The HR tech market still measures culture with annual surveys and a single number. Engagement scores arrive months after the behaviors that produced them. Recommendations rely on category averages. The signal of what is actually happening inside an organization, in real time, with statistical confidence, has been missing.
A LATAM-based program structured around culture and HR transformation needed something different. They needed a platform that measures culture as a living state, distinguishes performative behavior from real behavior, detects systemic tensions that no single score can reveal, and turns the diagnosis into action via AI agents calibrated by role.
The category did not exist. We named it: Culture Intelligence Platform with native AI. We built the first one.
What was at stake: the credibility of a culture program in front of executive leadership that had stopped trusting engagement surveys, and the ability to compare culture rigorously across business units without losing the qualitative depth that makes culture work matter.
The engagement worked across Horizon 02 and Horizon 03.
Horizon 02 designed the ICV 3.1 model, the conceptual core that distinguishes Human+ from a survey platform. The previous version (a 4-component ICV) was producing a single score that looked reasonable but hid the real dynamics. The 3.1 model decomposes culture into 5 universal constructos (Coerência, Ritmo, Permeação, Conexão, Franqueza) with 21 submetrics, normalizes signals statistically, and exposes confidence and authenticity as parallel layers.
The hybrid RAG motor was designed alongside the model. It combines lightweight models for simple signal extraction with heavier models for insight generation. Sanitization sits at the context layer, protecting sensitive data per tenant before any prompt is composed.
Horizon 03 designed the modular agent architecture. 16 modules of AI planned, 4 base modules entering production with the Sprint 1 cycle. Two agents are already running: Coach AI for individualized coaching and Onboard AI for tenant setup with automatic capture of corporate values from public sources.
Multi-tenant architecture with RLS at every table, JWT with claims, contextual isolation per tenant. The platform learns culture by tenant, not in aggregate.
The architecture has three concentric layers.
Layer 01 / The ICV 3.1 model. The proprietary conceptual core. Computed at three levels (personal, team, organizational). Each constructo carries score (0 to 10), confidence (0 to 1), authenticity (0 to 1), and a final composite used for prioritized display. Built on top of behavioral signals, typed Feedback de Incoerência, missões culturais with textual evidence, and agent interactions.
Layer 02 / Tension diagnosis and fragility. The ICV does not stop at score. It detects 6 systemic tension types from cross-constructo patterns:
01 SILÊNCIO DOURADO high apparent coerência with low franqueza 02 CORAGEM COM DISFUNÇÃO high franqueza without ritmo or conexão 03 TRIBOS ISOLADAS high local conexão, low organizational permeação 04 CULTURA DE ELITE strong culture concentrated in few actors 05 DISSONÂNCIA INSTITUCIONAL speech and practice misaligned 06 VOZ SEM ESCUTA feedback exists but does not turn into action
Plus a daily fragility snapshot across three dimensions: actor concentration (45%), temporal fragility (30%), instrument dependency (25%). Each tenant gets a robust, moderate, or fragile badge that says whether the measured culture is sustainable or hanging by few threads.
Layer 03 / Agent layer. Multi-tenant agents calibrated by role and by tenant bias. Two are in production today:
01 COACH AI individualized coaching on personal profile and journey 02 ONBOARD AI guides HR team during tenant setup, auto-captures values
Four base agents follow in Sprint 1. The remaining 12 modules ship across Sprints 2 to 6.
Core technology. PostgreSQL 15 via Supabase with native multi-tenant RLS. PgVector for embeddings. OpenAI for embedding generation, GPT-5 for insight composition. JWT auth with claims. Modeled on top of 27+ tables in production, anchored by Empresa, Pessoas, Cargos, Histórico de Cargos, and Logs.
The platform is in production with the first piloto in active use. Sprint 0 delivered the schema foundation that ICV 3.1 required. Sprint 1 is stabilizing the RAG motor, the admin panel, and the first two agents (Coach AI + Onboard AI). MVP target is June 2026.
Beyond the technical delivery, the engagement created a category.
The HR market does not have a direct competitor. Culture Intelligence Platform with native AI, multi-tenant, calibrable by client cultural bias is a new positioning. Human+ enters as the first reference in it.
Decision quality changed for the pilot client. The ICV's tension diagnosis surfaces patterns the previous tooling could not see. Silêncio Dourado and Tribos Isoladas as named, measurable conditions changed how the HR team frames its interventions.
The IP became durable. The ICV 3.1 model and the multi-tenant agent architecture are proprietary, reusable across future engagements, and central to Quantum Leap's positioning as a Lab. The same architectural decisions (Supabase RLS, PgVector, role-calibrated agents) reappear across Lab.IA and BRAIN.