AI Twin provides a private, model‑agnostic memory layer that lets users upload files, messages, and other personal data, then retrieve the right information at the right moment across the AI tools they already use. The service stores all data in the UK and EU with row‑level isolation, an append‑only audit log, and never accesses accounts without explicit consent. It supports universal input via typing, speaking, or forwarding, and runs a small on‑device embedding model for fast, local query ranking.
Funding
Funding not disclosed
Founders
Product
Problem
Individuals and professionals struggle to keep personal and business information—documents, messages, contacts, events—organized and readily accessible across the multiple AI tools they use, while also needing strong data privacy, auditability, and compliance with UK/EU regulations.
Solution
AI Twin offers a private, model‑agnostic memory layer that lets users upload files, emails, voice notes, and other personal data and have it typed into structured entities (People, Documents, Facts, Events). Each entity is stored with row‑level security, an append‑only audit log, and residency in the UK or EU, ensuring GDPR compliance and auditability. Users can retrieve information through a hybrid search that combines exact typed lookups with vector‑based fuzzy matching, all from a single query interface that works with the AI assistants they already use. Workspaces provide separate personal and business scopes with granular permission tiers, while features such as Magic Inbox, Smart Tasks, and Smart Forms automate the capture and organization of new data. The service is delivered via a subscription model, with pricing starting at £18 per month for personal use and optional practice workspaces for teams.
Target Audience
Primary customers are knowledge‑intensive professionals and small teams—such as lawyers, accountants, caregivers, and founders—who need reliable, compliant recall of personal and business information across AI tools.
Features
- Universal Input bar that accepts text, voice, attachments, URLs, and slash commands, routing content to the appropriate typed memory category
- Typed Memory model with immutable categories (People, Documents, Facts, Events) that are queryable, editable, versioned, and audit‑logged
- Hybrid Retrieval combining precise typed lookup and vector search in a single answer
- Row‑level isolation and four sensitivity tiers (LOW, MEDIUM, HIGH, CRITICAL) enforced at the database level
- Append‑only audit log visible to users for full provenance of every capture and action
- Workspaces for personal and business contexts with granular viewer/read‑write permissions
- Magic Inbox that transforms forwarded emails into structured memory entries (receipts, events, contacts)
- Smart Tasks and Smart Forms that automate routine intents (daily brief, data capture) with user confirmation