How it works
Auto-classification
Recipients tagged by relationship type: investor, customer, team, partner, family. You can override anytime.
Communication DNA
Per-relationship tone and frequency patterns. Pranan knows you sound different to a board member than to a hiring manager.
Context notes
Add private notes about a relationship and Pranan respects them in every draft.
Why this is the hardest layer to build
Voice matching is technically the easier problem. Statistical patterns from a single user's sent mail are bounded and repeatable. Relationship intelligence has to track per-contact context, which means the model state grows linearly with every recipient you correspond with.
For a user with 500 active contacts, that's 500 small models to maintain, each updated whenever you send to that contact. The infrastructure has to be efficient or the cost spirals fast.
Pranan's approach is to factorize: a base voice fingerprint plus per-contact deltas. The deltas are small and only update on relevant signals.
Try relationship intelligence free
300 drafts a month free. Voice fingerprint from your sent mail in under 5 minutes.