The Platform
How PrivitIQ Works
A unified intelligence layer built for business data you can't afford to expose. Four steps from raw data to real answers — entirely within your control.
01
Ingest
All your data. One secure pipeline.
PrivitIQ ingests structured and unstructured business data from a wide range of sources — no matter the format or origin.
- 📄 Documents — PDFs, Word files, contracts, legal records
- 📊 Structured data — accounting spreadsheets, financial reports
- 🏥 Healthcare records — PHI-sensitive files with controlled handling
- 🌐 Web crawling — continuously ingest business-owned content
- 💬 Communications — operational records and business channels
Data Sources
📄 contracts.pdf
📊 financials.xlsx
🏥 records.pdf
🌐 website crawl
📝 policies.pdf
💼 agreements.pdf
↓
Secure Ingestion Pipeline
Unified Data Layer
🔒 Internal — Legal Team
🔒 Internal — Finance
🌐 Public — Customer FAQ
Access controlled per dataset
02
Unify
One governed data layer. Full control.
All ingested data is processed, classified, and indexed into a secure governed layer. You define the structure — by brand, department, or use case.
- 🗂️ Multi-dataset configuration — segment data any way you need
- 🔐 Access control — public (customer-facing) vs private (internal)
- 🏷️ Intelligent classification — documents tagged by type and sensitivity
- 🔍 Semantic indexing — vector embeddings for deep meaning retrieval
- 🔄 Continuous sync — web crawl and scheduled re-ingestion
03
Query
Ask in plain language. Get precise answers.
PrivitIQ's conversational AI engine understands complex, multi-part questions and routes them through the right intelligence pipeline — simple or advanced.
- 🧠 Intent-aware routing — conversational, factual, or complex queries
- 🔍 Multi-query retrieval — decomposes complex questions into focused searches
- 📐 RRF reranking — evidence ranked by relevance across all sources
- 🤖 Multi-model AI — specialized models for extraction, reasoning, embeddings
- 🔁 Fallback strategy — model redundancy for reliability and cost optimization
- 💾 Session memory — context builds across your conversation
"Compare Q1 vs Q2 expenses across all departments"
→ Intent: Complex query
→ Decomposed into 4 sub-questions
→ Parallel retrieval across docsets
→ RRF reranking · Top 20 chunks
Structured answer with table
Delivery Channels
💬 Web Chat
📘 Facebook Messenger
📸 Instagram
🔌 REST API
All channels. One data layer.
04
Respond
Every channel. One truth.
Answers stream to any channel your business uses — without duplicating data or losing consistency across platforms.
- 💬 Web chat — embedded on your workspace or public site
- 📱 Messaging platforms — Messenger, Instagram, and API-enabled channels
- 📊 Structured output — tables, summaries, and cited evidence
- ⚡ Streaming — real-time token-by-token delivery
- 🔒 Access-aware — public channels only see public-designated data