Asmi calls you every morning. You talk - it handles the day. It calls services (dentist, salon, plumber, bank, insurance) or people (friends, colleagues) to coordinate, book or resolve things. Updates you on iMessage or WhatsApp when done. It can navigate IVRS, wait on hold and handle complex conversations well.
It started with a phone call from a close friend from my PhD days.
After a little chit-chat and a lot of catch-up, he asked me if I wanted to build something that actually helps people. Not in a vague, hand-wavy way but genuinely does the things you don't want to do yourself, and helps you do the things you actually care about, better.
The outbound call thesis is interesting - handling the messy phone-mediated workflows is genuinely the hard part of agent execution. One compliance layer worth building in now: Article 50(1) of the EU AI Act requires voice AI agents interacting with people in the EU to identify themselves as AI at the start of the call. For an agent calling a dentist's receptionist on behalf of a user, the opening line needs to be something like: "Hi, I'm an AI assistant calling on behalf of [user name] to [purpose]." That's also good caller experience - businesses responding to calls appreciate knowing what they're talking to. This obligation sits with whoever deploys the agent (Asmi), not just the underlying model provider. The August 2 deadline is 48 days out. If EU expansion is on the roadmap, building the disclosure into the call-start template now is much easier than retrofitting it later.
The clinic's website is broken. The second number rings forever. The third transferred me to a voicemail from 2019. I've been at this for 45 minutes. The appointment itself takes 20.
