A customer calls off tomorrow's 10:00 job
An hour of Friday, and the work booked into it, about to go to waste.
An orchestrator and agents, per job
They answer your customers, chase your work and follow up what went quiet, round the clock, in your words. Nothing goes out without your say.
Goals
Who you're hiring
Every team is built this way. Hire one, or several.
A real situation, start to finish · fleet operations team
Thursday morning at a business that sends people out to jobs. Nobody at the office heard about any of it, and Friday's van still goes out full.
An hour of Friday, and the work booked into it, about to go to waste.
Nobody at the office touched it. You set the rule once: refills inside your policy go on their own, anything outside it comes to you.
What this is not
A chatbot answers a question and stops. A team owns the job until it is finished, does the work in your systems, and comes to you when it should.
Which job?
These are the teams we run most often. Each one works its own channels, every initiative switches on by itself, and if your job is not here we build it.
Start with one initiative, not a platform. Most people start with a single leak they can already name, and switch the next one on once it is paying.
You stay in charge
It runs on its own. These four things stay yours, and every one of them is a switch or a screen in the product.
Start with everything past you. When the "is my phone ready" replies are always right, let those go out on their own and keep checking the rest. Refunds never stop waiting for you.
Typical approval rate, illustrative: about 70% of replies approved unedited in week 1, about 90% by month 3. The spending limit shown is an example. Our name for the queue is approvals; the limit is a spending cap.
It learns your business
Every week it knows a little more about how you do things. Nothing it learns is hidden from you: read it, correct it once, or delete it, and the change applies everywhere in one go.
Customers who write "wife's phone" usually mean the account is in the wife's name. Look the ticket up by both names before answering.
Dear Customer, your repair has been completed. Please attend our premises to collect at your earliest convenience.
Hi Priya, your phone is ready. Pop in any time before 5.30 and we will have it waiting.
Learnings are confirmed after three or more separate jobs; stale facts are flagged and expire (how it works). Every correction is kept as a before-and-after, applied in one go, undoable; every version of every instruction is kept (how it works). Our name for this: memory and the training centre.
Try the teams before you trust them
They play through what actually happens here, and they have to pass. Run a team through your own situations before it goes live (we call these Stories). Every team turned up already rehearsed, and every fix comes with a test proving it.
27 kinds of check, pass or fail and cost shown per run (how it works). The cost line is an example. Nothing in a rehearsal is sent to a real customer.
The question everyone asks second
Nothing changes for your customers
Your customers keep emailing the same address. Your helpdesk keeps working, with your existing tickets pulled in when you connect it. Your CRM stays the place your records live. Alerts land in Slack. Texts and WhatsApp messages go from your own number. If we do not already connect to something you use, describe it and the team builds the connection, logins included.
Every team is built round the way you already work: your systems, your tone, your rules. Tell us what needs doing and we will show you what it would look like.