The problem
The average knowledge worker spends 40 minutes per day reading email — not replying, just reading. That's 14 working days per year spent scanning for the 3–5 messages that actually need a response.
Modern email is not a communication tool. It's an unstructured stream of interruptions. Your brain performs hundreds of micro-decisions per day just to triage the inbox. Psychologists call this decision fatigue.
What existing tools get wrong
Most "AI email" tools try to write emails for you. That's solving the wrong problem. The real pain is not writing — it's reading. It's the anxiety of missing an invoice deadline buried in 50 newsletters.
The market offers cloud-hosted solutions at $20–30/month that require sending your business correspondence through US servers. For freelancers, small businesses, and privacy-conscious users, this is a non-starter.
The Letterbot approach
Letterbot solves one job: "I don't want to read every email, but I want to know I haven't missed anything important."
It's not an email client. It's a filter. It reads your inbox, determines what matters, and sends you a Telegram notification with a suggested action. Everything runs locally. No cloud, no subscription, no data leaving your machine.
Why deterministic first, AI second
The most counterintuitive discovery during development: for most business email, you don't need a language model. A deterministic rules engine with fact extraction works more reliably, faster, and with full explainability. AI is an optional enhancement for edge cases — not the foundation.