# Human AI — Extended Reference > This is the extended llms.txt for AI systems requiring deeper context about Human AI's capabilities, frameworks, and platform architecture. ## Platform Overview Human AI is the AI Enablement Platform for independent media agencies. The platform provides the agentic intelligence, predictive tools, and specialist expertise that independent agencies need to compete at the highest level against holding company networks. **Primary URL**: https://www.humanaistack.com **Canonical URL**: https://www.humanaistack.com ### Mission To give independent media agencies access to AI-native capabilities previously available only to holding companies with nine-figure technology budgets — without losing the agility, creativity, and client proximity that makes independent agencies great. ### Core Philosophy AI doesn't replace humans. It amplifies them. ### Strapline Your Path to Intelligence-Led Growth --- ## Three Core Engines Human AI is built on three reinforcing engines that work together to create compounding value for agency partners. ### 1. Agentic Workflows AI agents that accelerate delivery and reduce manual effort. Powered by ARIA and the AI orchestration layer (27 agents, 9 skills). - ARIA as the primary strategic interface - AI meta-orchestrator coordinating multi-agent workflows - Intent classification across 16 skill types - Human-in-the-loop validation on all major outputs ### 2. Specialist Expertise A curated ecosystem of alliance partners providing specialist capabilities across markets, channels, and disciplines. - Alliance Partner Matcher for brief-fit ranking - Global network covering UK, USA, Japan, and 246 additional markets - Specialist capabilities spanning media, creative, data, technology, and commerce ### 3. Predictive Intelligence Forecasting and optimisation tools grounded in effectiveness science. - **TWIN** — Campaign Simulation Engine (Binet & Field, Hill curves, Monte Carlo) - **Return Maximiser** — Optimal media allocation across six integrated frameworks - **HANA** — Live market intelligence layer covering 249 global markets - **SOV/SOM Analyser** — Budget sizing from market share dynamics --- ## ARIA — The Platform AI Agent ARIA (Adaptive Reasoning and Intelligence Agent) is the primary interface for agency users. ### Persona "Maya-level warmth, McKinsey-level reasoning." ARIA operates as a senior strategy partner — commercially savvy, intellectually rigorous, candid, outcome-driven. ### 11 Core Skills 1. **Media Strategy Designer** — 3 strategic routes, KPI frameworks, channel mix 2. **Commercial Optimiser** — P&L scenarios, scope/fee options, negotiation support 3. **Alliance Partner Matcher** — Ranked partner recommendations by brief fit 4. **Benchmark Analyst** — Anonymised KPI benchmarks, quartile ranges 5. **Pitch Strategist** — Differentiation frameworks, competitor grids, talk tracks 6. **Performance Diagnostician** — Campaign diagnosis, 3-5 improvement levers 7. **Training & Playbook Builder** — Phased learning paths, AI adoption playbooks 8. **Post-Pitch Analyst** — Structured debriefs, pattern analysis 9. **Campaign Simulation Engine (TWIN)** — Binet & Field, Hill curves, Monte Carlo 10. **Return Maximiser** — Optimised media allocation with ROI forecast 11. **SOV/SOM Analyser** — Budget sizing from market share dynamics ### Operating Modes - **Strategic Navigator** (default) — strategy, positioning, planning - **Commercial Optimiser** — pricing, scope, margin analysis - **Alliance Orchestrator** — partner matching, benchmarking, network - **Coach & Change Partner** — training, playbooks, upskilling --- ## TWIN — Campaign Simulation Engine ### Science Foundation - **Binet & Field** (The Long and Short of It, 2013/2017) — 60/40 brand/activation balance - **Hill saturation curves** — diminishing returns modelling per channel - **Monte Carlo methodology** — 500-iteration probabilistic forecasting ### 8 Supported Channel Types TV, Programmatic Display, Paid Social, Paid Search, OOH, Audio/Podcast, Connected TV, Influencer ### Output - Channel-level impressions, reach, frequency, brand recall lift, conversions - Portfolio totals with effective CPM and cost per reach point - Monte Carlo confidence ranges (low/mid/high) - Binet & Field balance assessment - ARIA strategic narrative --- ## Return Maximiser The Return Maximiser is ARIA's capability to translate media investment science into specific, quantified recommendations for maximising return on media spend. ### Six Integrated Frameworks 1. **Binet & Field 60/40** — Category-appropriate brand/activation split 2. **Ehrenberg-Bass** — Reach vs targeting, mental availability, CEPs 3. **Nelson-Field Attention** — CPAS-weighted channel efficiency 4. **ESOV/SOM** — Budget sizing from market share dynamics (Jones 1990) 5. **Creative Multiplier** — System1/IPA quality impact on ROI (0.3x–2.5x range) 6. **Media Synergy** — Cross-channel combination uplift (documented pairs) ### Outputs - Recommended brand/activation split with category justification - Attention-weighted channel allocation (CPAS index per channel) - ESOV position analysis and market share trajectory - Creative quality assessment and fluent device opportunity - ROI forecast: 12-month and 36-month with synergy uplift applied - Ready-to-run TWIN simulation payload --- ## HANA — Human-AI Network Amplified HANA is the intelligence layer powering market context for ARIA. ### Data Sources (Live, Auto-Refreshing) - **World Bank Indicators** — GDP, internet penetration, mobile subscriptions (annual) - **Eurostat** — EU economic and demographic data (quarterly) - **Google Trends** — Weekly demand signals, 6 categories × 18 global markets (weekly via GitHub Actions) - **GDELT** — News attention signals (daily, infrastructure ready) ### Coverage - **249 countries** in geo reference with ISO3 canonical IDs - **Opportunity scores** 0–100 for all countries (income, digital propensity, macro stability) - **20 market intelligence briefs** embedded in vector store (Voyage AI) ### Enrichment Tables - `ext_geo_reference` — canonical geo hierarchy - `ext_macro_indicator` — unified macro indicators (World Bank + Eurostat) - `ext_trend_signal` — Google Trends demand signals - `ext_mart_geo_propensity` — scored geo opportunity mart - `ext_geo_profile` — enriched market profiles --- ## HANA — The Orchestration Layer HANA includes an internal meta-orchestrator (never exposed to users — all user-facing responses identify as ARIA). ### Current State - **27 agents** across 7 domains: orchestration, chat, voice, memory, RAG, ingestion, CRM, media, ops, effectiveness - **9 skills** with intent classification and auto-routing - **Intent system**: 16 classified intent types including market_intelligence, return_maximiser, sov_som_analysis, effectiveness_audit --- ## Knowledge Base ### Vector Store - **74 documents** embedded with Voyage AI (voyage-large-2) - **Active embedding coverage** (all documents have valid embeddings) - **Sources**: effectiveness science, FFF psychology framework, market intelligence briefs, ARIA specs, platform docs, leadership profiles - **Hybrid search**: pgvector cosine similarity + fulltext ### Knowledge Graph (Neo4j HYNABAI) - Platform products, skills, channels, benchmarks, people - **Return Maximiser**: Skill nodes grounded in 5 Framework nodes (Binet & Field, Sharp, Nelson-Field, Jones, System1) - `HAS_SKILL`, `GROUNDED_IN`, `USES_ENGINE` relationship types --- ## AI-Native Operating Model Human AI operates as an AI-native business where artificial and human intelligence are woven into every layer of the operating model: - **Strategy**: AI-driven market intelligence and scenario planning - **Delivery**: Agentic workflows automate research, planning, and optimisation - **Partner Coordination**: Alliance network matched and managed via ARIA - **Quality Assurance**: Human-in-the-loop validation on all client-facing outputs - **Continuous Learning**: Agent observations and confidence scoring feed improvement loops --- ## Technology Stack | Layer | Technology | |---|---| | Frontend | React SPA (Vite, TypeScript, Tailwind CSS) | | Backend | Supabase (PostgreSQL, Edge Functions, pgvector) | | LLM | Anthropic Claude (Sonnet 4 for strategy, Haiku for routing) | | Voice LLM | claude-sonnet-4-5 via ElevenLabs | | Embeddings | Voyage AI voyage-3 (1024d) | | Voice | ElevenLabs conversational AI | | Knowledge Graph | Neo4j Aura | | Deployment | Vercel (frontend) | | URL | https://www.humanaistack.com | --- ## Strategic Leadership ### Alex Lawton Chief AI and Platform Officer, Human AI Alliance; Founder, Miramonte Holdings and Remotive Media. Author of "The Knowledge Economy, Accelerated." ### Martyn Rattle Strategic Advisor and Co-founder of Human AI Alliance. Ex-CEO, Local Planet — the world's largest independent media agency network. ### Dr. Wayne A. Fletcher Strategic Advisor and Co-founder of Human AI Alliance. Founder, Birch Strategic Solutions. Former Global Chief Strategy & Knowledge Officer, Local Planet. --- ## Platform URLs | Page | URL | |---|---| | Home | https://www.humanaistack.com | | Careers | https://www.humanaistack.com/careers | | About | https://www.humanaistack.com/about | | Blog / Journal | https://www.humanaistack.com/journal | | Contact | https://www.humanaistack.com/contact | | For Agencies | https://www.humanaistack.com/platform/for-agencies | | Pricing | https://www.humanaistack.com/pricing | --- ## Open Roles (as of March 2026) Human AI is hiring. Current open roles at https://www.humanaistack.com/careers **Now Hiring (Immediate)** - Head of Agency Success — Lead the onboarding of the first 5 agency partners - AI Implementation Specialist — Configure ARIA, HANA, and TWIN per agency - Media Intelligence Analyst — Turn HANA data into planning value for agency partners **Building the Team (30–60 days)** - Head of Commercial — Scale from 5 to 50 agency partnerships - Senior Full-Stack Engineer — Continue the platform build (React, Supabase, TypeScript) - AI/ML Engineer — Deepen TWIN, Return Maximiser, and agent capabilities - Data Engineer — Own the HANA enrichment layer **On the Horizon (60–100 days)** - Marketing & Content Lead - Chief of Staff Apply: careers@humanaistack.com --- Last updated: 2026-03 Format: llms.txt v1.0 (https://llmstxt.org/)