# Adaptive Systems

> Deterministic AI Infrastructure. One platform, three products, and custom builds to the same standard. Front Desk: AI receptionists (voice + booking), AI support bots (RAG with citations) and lead qualification (LLM sub-scores + deterministic policy gate) for service businesses. Docket: AI document intake (invoices, receipts, purchase orders; PDFs, scans and phone photos; European and US suppliers) where every value is located on the page and checked by code, scans are read twice (vision model + independent OCR), with a risk gate for bank-detail swaps (IBAN or US ACH), duplicates and PO mismatches. Dossier: applicant document verification (bank statements, utility bills, payslips, tax documents, IDs) for lenders, letting agents and firms with AML duties: the PDF itself is examined for edits, balances and passport MRZ are checked by code, and every name, address and date is compared with the declaration; it checks documents, never scores people. In all three, models extract and draft; deterministic code decides; every decision is audit-logged.

Start here: [the one-pager](https://adaptivesystems.dev/one-pager) — all three products, how they decide, measured numbers, the stack and every link on one page.

## Pages
- [Home](https://adaptivesystems.dev/): Three products, Front Desk (conversations), Docket (invoices) and Dossier (applicant documents), plus custom builds: live numbers, a worked lead, a checked invoice, a caught forgery, measured results.
- [One-pager](https://adaptivesystems.dev/one-pager): All three products on one page: what they do, how they decide, measured results, security, the stack, how we work and every link to try them.
- [Docket](https://adaptivesystems.dev/docket): Document intake: invoices read by a model, every value checked by code, a risk gate for fraud and duplicates, live numbers and measured accuracy.
- [Dossier](https://adaptivesystems.dev/dossier): Applicant document verification: statements, bills, payslips and IDs read by a model, the file examined for edits, balances and MRZ checked, names and addresses compared; a published detection eval.
- [Try Dossier: forge it yourself](https://adaptivesystems.dev/dossier/try): Pick a forgery, we build a synthetic applicant's pack with it, and Dossier checks it in front of you; erased after 30 minutes.
- [Front Desk](https://adaptivesystems.dev/platform): Conversations: AI receptionist, support bot and lead qualification on one engine; architecture and flows.
- [Try Front Desk: score a lead](https://adaptivesystems.dev/platform/try): Write an enquiry and watch a real model score it, code check it, gates route it and the reply drafted, live.
- [Try Docket](https://adaptivesystems.dev/docket/try): Drop your own invoice: every value read, located on the page and checked, then erased after 30 minutes.
- [Services](https://adaptivesystems.dev/services): The five Front Desk modules on one engine.
- [AI Receptionist](https://adaptivesystems.dev/services/ai-receptionist): A voice agent that answers questions, checks your calendar and books or cancels appointments, using only your own facts.
- [AI Support Bot](https://adaptivesystems.dev/services/ai-support-bot): Answers customer questions from your documents and your hard facts, shows its sources, and says so when the answer isn't there.
- [Lead Qualification](https://adaptivesystems.dev/services/lead-qualification): Every inbound enquiry is scored with evidence, checked by deterministic rules and routed: full offer, starter, needs info, review or reject.
- [Self-hosted Automation](https://adaptivesystems.dev/services/self-hosted-automation): The whole stack (n8n, APIs, vector database, monitoring and backups) installed on a server you own.
- [AI Rescue & Audit](https://adaptivesystems.dev/services/ai-rescue): For a RAG bot that hallucinates, an automation that fails silently, or an agent nobody trusts: find out why, fix it, and add tests so it stays fixed.
- [MCP: Docket and Dossier in your AI assistant](https://adaptivesystems.dev/mcp): Connect Claude to the Docket and Dossier sandbox: the tools, the limits and what happens to your documents.
- [Live system](https://adaptivesystems.dev/live): Live activity, stats and health checks from all three products.
- [About](https://adaptivesystems.dev/about): Who we are and how we work.
- [FAQ](https://adaptivesystems.dev/faq): Hallucinations, prompt injection, data location, setup time.
- [Contact](https://adaptivesystems.dev/contact): Scoping call, app access, status.
- [Privacy](https://adaptivesystems.dev/privacy): Analytics consent, what the chat and forms store, erasure requests.

## Data
- [Front Desk live stats (JSON)](https://api.adaptivesystems.dev/public/stats): sanitized counters, service health, models of the production stack and the published eval (lead decisions + held-out support answers)
- [Docket live stats (JSON)](https://adaptivesystems.dev/live/docket/stats.json): documents processed (real and synthetic test samples counted separately), values verified, risk flags caught, measured accuracy of the production model
- [Dossier live stats (JSON)](https://adaptivesystems.dev/live/dossier/stats.json): applicants checked, outcomes, forgeries caught, the published detection eval
- [Full text](https://adaptivesystems.dev/llms-full.txt): all site content in one file