How It Works

The mechanism

Veylith connects to your inbox via standard protocols (IMAP/SMTP or OAuth for Google Workspace and Microsoft 365). Once connected, incoming messages pass through three stages before you see them.

1. Classification

Each message is read and classified against the categories you define: client queries, invoices, internal updates, newsletters, cold outreach, urgent items. The AI model runs on our UK servers. No message content is sent to a third-party service at any point.

2. Routing

Based on the classification, messages are moved, labelled, or flagged. An invoice from a client goes to your accounting folder. A query marked urgent gets a desktop notification. Newsletter traffic gets filed or deleted before it reaches your inbox view.

3. Summary

Each morning you receive a digest of what came in overnight: how many messages were processed, what was flagged as urgent, what was routed, what was removed. You review what needs attention. The rest is already handled.

What stays in your control

You set the categories. You set the routing rules. You decide what gets flagged and what gets removed. Veylith executes your instructions. The AI is the mechanism, not the decision-maker.

You can audit any decision: every classification and routing action is logged, readable, and reversible. If Veylith misclassifies a message, you correct it once and the model learns the correction for your inbox.

Infrastructure

All processing runs on dedicated UK-based VPS infrastructure. Local language models handle classification. No data crosses outside the UK. No third-party AI provider is involved at any stage. We own the hardware and the model weights.