A missed call is a job that goes to the next business on the list. A polite answer is not much better if nothing gets saved: a confident "you're booked" while the calendar stays empty.
We are building Notaray to close the gap between what was said and what was saved. Notaray is an AI front door for service businesses. It will pick up the requests your team can't, check your real calendar and confirm the booking once the calendar shows it. Anything uncertain goes straight to a named person with a deadline.
It is not live yet. The booking engine is in testing now, phone answering is on the way, and early access is opening for a small group of design partners. This post says what is built, what is not, and how to get in early.
Answering is easy. Landing the booking is the job.
Speaking and booking are two separate steps. A voice model produces the sentence the caller hears. A different call goes to the calendar to create the event, and that call can fail, time out, or find the slot already taken. If the agent speaks before the calendar has answered, it is confirming an intention, not a booking.
So Notaray is designed around one rule: it confirms a booking after the calendar shows it, not before. That is a design goal, not a performance guarantee. Nothing here has been measured, because no business is using Notaray yet.
Four steps, no guessing.
Every request is designed to run through the same four steps.
- Understand the request. Capture the requested time, the service and the details needed to act, under the business's own rules.
- Check the calendar. Look at the real destination calendar before promising anything. A suggested slot is not a reservation.
- Write one booking. One bounded attempt with a stable key, so a repeated request finds the original result instead of booking twice.
- Read it back. Ask the calendar what it now holds and compare it with the request. That comparison alone decides the outcome.
There are two outcomes and nothing in between. Confirmed means the calendar returned an identifier and the read-back matched the request. Pending means the result could not be verified: the request goes to a named person with a deadline and the original context, and the caller hears that it was requested. An unknown result never sounds like a success.
Every finished attempt is designed to leave a receipt: what was asked, what was checked, what the calendar returned and what the outcome was. It is the same pattern we use for every agent we ship: rules where rules work, a person where the result is uncertain, and a record of both.
What is ready, and what isn't.
- In testing now: the booking engine. Requests, calendar checks, read-back, receipts and follow-up, on Agnotiq's own operations first.
- Up next: browser voice in front of the same booking rules, subject to testing.
- Then: a limited phone pilot for overflow, missed and after-hours calls, subject to testing and to a review of call-recording consent where the pilot runs.
Notaray doesn't take calls yet. No integration is announced yet, and there is no price: the plan is to charge for verified bookings, not minutes, and prices are not set. The Notaray page on this site keeps that list current.
Who it is for.
Small service businesses in the United States and Canada that run on an appointment calendar and lose requests when the team is busy or closed: home services, clinics and wellness, salons and spas, and professional services.
If you are comparing AI phone agents today, the Notaray team published a test you can run on any of them, ours included once it exists: eight test calls that show whether an agent books reliably. Judge each one by the calendar, not by the voice.
Get in early and shape what ships.
Design partners tell the team how requests reach them and which calendar holds the truth. Their answers decide which calendar gets validated first and which cases the booking flow has to handle. It is not a purchase, a trial or a commitment, and a pilot place isn't guaranteed.
Tell the Notaray team where requests arrive and where bookings need to land. Spots are limited.
Request early access