How we work · After launch
The first ninety days with an AI employee.
Launch is where the working relationship starts. A named owner picks the employee up on day one, a written check at week six asks whether the work actually changed, and ninety-day telemetry says whether your team is using it.
The three fixed points
Three points you can hold us to, and a written report each month in between.
In short
Working with an AI employee means a named person on your team owns it from launch, approves the scripts it works inside, and picks up whatever it escalates. 7 Minds Systems stays accountable for the employee itself. A written adoption check at week six and ninety-day usage telemetry show whether it earned its place.
What arrives
Every AI employee arrives with a name against it.
An AI employee goes live wired to the calendar, CRM, and inbox you already use, and we keep running it afterwards. What decides whether it earns its keep is the arrangement around it. Someone directs it. Its scripts are approved before it speaks to anyone. Every exchange it handles is written down.
That arrangement is settled during scoping and written into the build, so nobody is working out the reporting line in month two. Your named owner sets what the employee may decide alone and what it must escalate, and every escalation it raises lands with them.
We stay accountable for the employee itself, including how it runs and how it is corrected. Your owner directs the work; we keep the worker in good order. Where we run a system on your behalf, you remain the data controller and we act as your processor under a data processing agreement.
The twelve weeks that precede day one are set out on the how we work page.
The first ninety days
Three fixed points, and a record that runs between them.
Adoption is the thing that fails quietly, so it is checked on a schedule and written down each time. Every point below lands in front of you on paper, with the job that sits on your side of it.
Between the points
A written report each month covering what the employee handled and what it passed to a person. Your owner reads a sample against the approved script and sends the corrections back in writing, and the scripts are corrected from what that shows.
- Day one
It goes live with its owner named
The employee starts work connected to the calendar, CRM, and inbox you already use, inside scripts and escalation rules you approved before it handled a single live enquiry. The named owner chosen during scoping picks up whatever it escalates from the first morning.
- In your hands
- The approved scripts and escalation rules, and one name against the workflow.
- What you do
- Your named owner takes the first escalation, and starts correcting the script from what they read.
- Week six
The written adoption check
A written checkpoint that asks whether the workflow you hired the employee for has actually changed. It reads six weeks of the audit trail against that workflow, and it lands with your named owner.
- In your hands
- The written adoption check, and a clause you can invoke.
- What you do
- Arrive with what the trail shows. Where it escalated something it could have handled, where it handled something a person should have seen, and the jobs your team still does by hand.
- Day ninety
The usage telemetry review
Ninety-day usage telemetry counts how much of the workflow actually ran through the employee, which is what the review turns on. A thin record puts the role itself in question, and the scripts or the owner are corrected from what it says.
- In your hands
- Ninety-day usage telemetry, read against what you hired it to do.
- What you do
- Say whether the role was scoped to a job somebody on your team genuinely owns, and name the owner again if it was not.
What is settled in writing
The rules an AI employee works inside.
Everything below is agreed before launch and written into the build, so the answer to a hard question in month four is already on paper. This is what an AI usage policy needs to cover for a worker that talks to your clients.
Who directs it
A named person on your team
Chosen during scoping and written into the build. That person sets what the employee may decide alone and what it must escalate, and every escalation it raises lands with them. We stay accountable for the employee itself, so the work is directed on your side and the worker is kept in good order on ours.
What it may say
Scripts you approved first
The employee works only inside scripts you have approved, so what it can say is bounded from the start. Anything sensitive goes to a named person on your team for sign-off before it reaches your client, and in regulated fields every regulated step is approved by a human before it goes live.
What is recorded
An audit trail on every enquiry
Each build ships with an audit trail, so any enquiry can be opened and read back in full. It holds what came in, what the employee replied, which approved script it was working inside, and the point at which it handed the matter to a person.
Who holds the data
You remain the controller
Where we run a system on your behalf, you are the data controller and we act as your processor under a data processing agreement. A written data-handling boundary ships with every proposal, and data is held in the UK or EU on request.
When the employee gets something wrong, it surfaces in the audit trail, we correct the script, and you re-approve it before it runs again.
Scope the role, and the owner who will hold it.
Book thirty minutes and we will map the workflow, the person who owns it, and the checkpoints it gets measured against.
Thirty minutes, no pitch. You leave with the workflow named, the owner named, and the date it could be live.
Questions
Living with an AI employee, answered.
Plan on the first ninety days. The employee is working from day one, connected to your tools and bounded by scripts you approved, so there is no ramp before it answers anything. The settling in happens on your side, as your named owner learns what to let it decide alone and what to pull back. The week-six adoption check is the first honest read on whether the workflow has changed.
A named person on your team, chosen during scoping and written into the build before launch. Day to day they pick up whatever the employee escalates, and they hold the line between what it may decide alone and what must come to them. Each employee is scoped to one named workflow, so adoption stays a question with a single name against it.
Every enquiry it handled, readable in full. You can see what came in, what the employee replied, which approved script it was working inside, and the point at which it handed the matter to a person. The monthly report summarises the same record, and the underlying trail stays there for your own inspection.
It is a written checkpoint that asks whether the workflow you hired the employee for has actually changed. If your team will not use what we built, the walk-away clause applies and we stop. Where usage is thin it usually means the employee was scoped to a job nobody owned, so the correction is a clearer owner and a tighter workflow.
For a worker that talks to your clients, four things settle it, and all four are agreed before launch and written into the build. Who directs it is a named person on your team. What it may say is bounded by scripts you approved before it handled a live enquiry. What is recorded is an audit trail on every enquiry. Who holds the data is you, as controller, with us acting as your processor under a data processing agreement.
It runs on two sides. Your named owner directs the work, sets what the employee may decide alone, and receives every escalation it raises. We stay accountable for the employee itself, including how it runs and how it is corrected. A written report each month goes to your owner covering what it handled and what it passed to a person, and the scripts are corrected from what that shows.
See the roles you can hire on the AI employees page, or how a whole department of them cooperates in the Cooperating AI Workforce.
Book a strategy call
Thirty minutes to name the owner and the workflow.
Thirty minutes to scope this role properly: what it would answer, what it would escalate, and what it would cost to run.
Know the scope and the number before you start.
The build is scoped and priced in writing before it begins, with the exclusions named, and the AI workforce is built and trained on your services and voice before it answers anyone. You commit to a defined piece of work, not to a promise.
Book a thirty-minute call
Pick a time straight from the diary. If you would rather write first, email us and a person replies, usually the same working day.