Waits for your business hours before it messages anyone
AI that works on its own, on a schedule or an event
Brief the assistant the way you would brief a colleague, then tell it when to start: every Monday, once on a date, or when a deal moves. It can work through a whole saved list, asks your team when it needs a decision, and writes down everything it did.
- New triggers start switched off
- Preview before it sends
- Silence is never a yes
One request, start to finish
Every Thursday at 10:00, the assistant follows up new bridal enquiries
- 1
Start it your way On a schedule, once, on an event, or only when you press Run now.
- 2
See who it reaches The preview names the contacts it would message and counts everyone it would skip, with the reason.
- 3
It asks when it doesn't know A question goes to your team in Needs you, with a deadline.
- 4
Read the run as a story What it said, what it changed and who answered its question.
01: TRIGGERS
Tell it what to do, and when to start
Most follow-up dies because nobody has time on Monday morning. A trigger is a job you hand the assistant once. You write the brief in plain language: every Monday, find deals that haven't moved in two weeks and draft a follow-up for each. Then you pick what starts it.
On a schedule
Every day, week or month at a set time.
Once
At a date and time you choose, such as the morning a sale starts.
When something happens
Three events can start one today: a new contact is added, a deal moves to a new stage, or a booking is completed. They fire however the change was made, in the app, by import, through the API or by dragging a card.
Run now
A trigger that only runs when you press the button.
From another system
Each trigger can have its own web address, so a form backend, Zapier or anything that can send a request can start it. The address is shown once, starts only that trigger, and can be replaced, which stops the old one straight away.
A run doesn't need a customer. Review last week's ad spend and tell me what to pause is a valid brief, and nothing is sent to anyone unless the brief says so.
Each trigger shows when it next runs and what happened last time. New triggers start switched off, and switching one off takes effect at once.
02: ONE BRIEF, A WHOLE LIST
Point it at a saved list and see who it skips
A trigger can act on every contact in a saved list, one run per person, all from the same brief. You write the job once, and each contact on the list gets handled on their own.
Before you switch it on, the preview names the first contacts it would act on and counts everyone it would leave out, with the reason for each. The preview starts nothing and costs nothing.
Always skipped
People who opted out, people on the do-not-contact list, contacts on a private number, people already mid-conversation with an assistant, and numbers on your AI exclusion list. Each reason has its own count.
Up to 500 a run
Each firing reaches at most 500 contacts, and the batch is trimmed to fit your remaining AI budget instead of being refused.
Paced, and polite about the hour
Messages are spread out so hundreds don't leave at once. A trigger that fires outside business hours waits until you open. If an outage would make it more than an hour late, that run is skipped.
Never twice
A retried firing cannot act on the same person twice, and one failed schedule doesn't break the next one.
Clean when you delete
Deleting a trigger, or the assistant it wakes, withdraws whatever it had queued.
Bring one follow-up you never get to
Tell us the job. We'll write the trigger with you on the call.
We reply in under an hour.
03: NEEDS YOU
When it needs a decision, it stops and asks
An assistant on a job can stop, put a question to your team, and carry on with the answer, keeping the work it has already done. It can ask for an approval, a choice between options it has narrowed down, or a free-text answer. Waiting costs nothing: no AI calls and no spend.
A deadline on every question
Between one hour and a week, with a fallback you can read. If nobody answers, a proposed action is never carried out, a question about a live chat passes that chat to a person, and a back-office run stops and says why.
The right person answers
Approvals need an admin. A choice or a factual question can be answered by anyone on the front desk. Questions raised on a private number are seen only by its owner and admins.
First answer wins
If two people answer, the second is told who decided and what they chose.
A queue built for answering
Needs you lists questions by deadline, not arrival, with For me, Everyone and Answered views. A live countdown turns red near the deadline, and keyboard shortcuts let you approve or decline without the mouse.
The customer isn't ignored
If the customer writes while the run is waiting, the assistant reads that message when it carries on.
You hear about it
A new question rings the bell in the app, and reaches your phone on WhatsApp if nobody opens it within fifteen minutes.
One question at a time per run, five at most over a run's life. Runs are stopped after three days as a safeguard, and time spent waiting on a person doesn't count, so a Friday question survives the weekend.
04: READ EVERY RUN AS A STORY
Every run, written down in plain sentences
Activity lists every conversation an assistant handled, newest first, and a run waiting on your team shows as Waiting on you. Open one and the Story view reads like a note from a colleague. The first line is the assistant's own summary, then why it started and how long it took, then each thing it did: replied to Mehak, with what it said; updated her event date; asked you a question, with the question.
Each change names the record it touched. A failed action is shown as failed, with its message. Retries read as one line instead of three failures. Voice notes it transcribed and photos it described sit in the timeline too. A Technical view is there for whoever wants it, and the page remembers your choice.
A stuck or failed run can be resumed or retried by hand, and a retry keeps what the run already spent.
05: SPLITTING UP BIG JOBS
Big jobs, split into parts it can check
For a large back-office job, an assistant can break the work into parts and hand each to a helper with a small set of tools, then read the results and carry on. It decides how to split the job at the time, so there is no flowchart to draw.
Helpers never message customers
They have no conversation, and every sending tool is removed.
Fewer tools, never more
A helper only gets a subset of the assistant's tools, and never one your account switched off.
Kept off live chats
Splitting is never used on a live customer chat.
One spend limit for the whole job
The assistant and its helpers share it.
A failed part doesn't sink the job
The assistant is told how many parts finished, and stops waiting for a helper after thirty minutes.
Off until you switch it on for an assistant. You set how many helpers run at once, which tools they get and a spend cap, and your settings can only tighten the built-in limits.
The honest part
What it won't do
Start on any event you like
Today three events can start a trigger: a new contact, a deal moving stage, and a booking completed.
Message more than 500 people in one run
That is the cap per firing, and a smaller AI budget trims it further.
Send late at the wrong hour
A run that would go out more than an hour late is skipped.
Read silence as a yes
An unanswered approval is never carried out.
Message people who said STOP
Opted-out contacts, the do-not-contact list and excluded numbers are always skipped.
Run forever
Runs stop after three days, not counting time spent waiting on your team.
Use helpers on a live chat
Splitting work is for back-office jobs only.
Works with
This page
AI that works on its own
- AI that works on its own
- Saved listsone run per contact on the list
- Contactsa new contact can wake it
- Pipelinesa deal changing stage can wake it
- Bookingsa completed booking can wake it
- AI assistantthe same instructions, tools and rules
- Team and notificationsquestions land in Needs you and the bell
Questions about AI on its own
Before you switch a trigger on
Will it message my whole list the moment I save it?
No. New triggers start switched off. Run the preview first: it shows who would get a message and who would be skipped, and it sends nothing.
What if it fires at night?
It waits for your business hours. If an outage would make it run more than an hour late, that run is skipped.
Who can answer its questions?
Approvals need an admin. A choice or a factual question can be answered by anyone on the front desk. The first answer wins, and the question disappears from everyone else's bell.
Does it use the same rules as my assistant in chat?
Yes. A trigger wakes one of your assistants, with its instructions, its tools and the actions you marked for approval.
What if a run gets stuck?
You can resume or retry it by hand from Activity. A retry keeps what the run already spent.
Can another system start a trigger?
Yes. Each trigger can have its own web address that a form backend, Zapier or similar tool can call. It starts only that trigger, and replacing the address stops the old one at once.
Book a demo
Hand the Monday chase to the assistant
Bring the follow-up your team never gets to. We'll set up the trigger with you, run the preview on your list and show you the story it writes.
We reply in under an hour.