Insight

Write your question in natural language — no need to remember exact keywords, precise dates, or file names. Insight understands synonyms, intent, and conversation context. Ask about deadlines, agreed amounts, decisions made, promises offered, terms negotiated. The AI finds the answer navigating months or years of conversations, even when the information is buried in transcribed voice notes, attached PDFs, or screenshots. Works in English, Portuguese, Spanish — no need to switch to a technical vocabulary.

Every answer comes with clickable links to the exact excerpts of messages, audio, or documents that back it up. It's not loose hallucination — you can audit the origin of every statement. When the AI is uncertain, it explicitly admits and shows what it found closest, without making things up. Each citation opens directly at the right message in the original conversation, so you confirm context before deciding. This is the fundamental difference between RAG (Retrieval-Augmented Generation) and traditional chatbots: responses grounded in real sources, not probabilistic guessing.

PDFs, Excel/Google Sheets spreadsheets, contracts, PowerPoint or Keynote slides — any file uploaded to Drive becomes semantic search material. The system includes OCR extraction for images, document photos, and scanned PDFs, so WhatsApp screenshots and digitized contracts are searchable too. You can query by period ('what did I agree to in March'), by sender ('what did Ricardo promise'), by specific conversation, or cross multiple dimensions to recover promises, decisions, or complaints lost in the history.

Insight applies RAG architecture: your messages and documents are processed into small chunks and converted into embeddings — numerical representations of meaning stored in Qdrant. When you ask something, the query also becomes an embedding and the system retrieves the most semantically similar chunks, not raw keyword matches. Relevant excerpts are passed as context to a language model (GPT-4.1-mini or Claude Sonnet), which synthesizes the answer based only on those fragments. The AI doesn't 'know' anything on its own — it can only answer what's in your sources. More security, more auditability, less hallucination.

Sales reps recall promises made to clients months ago before a closing meeting. Managers retrieve meeting decisions that turned into informal WhatsApp chatter without minutes. Lawyers track mentions of specific clauses across thousands of client messages. Support teams resurface a customer's complaint history without opening 50 chats. Medical practices verify prior guidance given to patients. Recruiters find candidate commitments scattered across months. Founders recover details of investor conversations buried in noise. Anyone who has 'that important message I saw last week' and can't find it anymore — finds it.

Insight is the platform's main entry point — where most users open first. It cross-references data from Sessions (connected WhatsApp + Telegram), Drive (uploaded files), and Files (conversation media). You can save frequent questions as Projects to reuse without re-configuring context. Works together with Reporter (uses Insight to generate briefings) and Radar (uses Insight to summarize unread messages). Every query is logged in history, so you can review past questions, refine, and re-run with new parameters.

Q: Does the AI have access to my data? — Only during the query. Your conversations are AES-256-GCM encrypted at rest, and the AI receives only the chunks needed to answer each question, without training anything on your data. Q: Can I use offline? — No, RAG requires a cloud LLM (GPT-4.1 or Claude). Q: How much does each query cost? — Included in your plan with no per-query limit. Q: Does it work in groups with 200+ people? — Yes, and it's especially useful for recovering decisions in groups where topics shift quickly. Q: Does the AI make things up when it doesn't know? — No. When retrieved chunks don't contain the answer, the system explicitly says it didn't find it and shows the closest material.