Hiring an AI employee is only the first decision. The operating model names who approves each action, what happens when the AI employee gets something wrong, and what your team sees each week.
Who does what once the role is live
The work splits clearly between the provider and your team.
- Struan: runs the jobs named in the job description at the agreed times, corrects mistakes as they are found, tunes the role as the work changes, and signs a weekly report.
- Your team: names the jobs, decides which work needs a yes, gives that yes, and keeps the relationships and the judgement calls the role does not make on its own.
How approvals actually work
The job description states which work needs a named person's approval, and who that person is. Gated work sits in the chat until that person says yes; nothing gated goes out on its own. Gates come down one at a time, only when your team decides to remove one, not automatically as the role runs for longer.
Illustrative example, not a customer result: a support role might draft every reply, but hold anything about a refund or a complaint for a named team member to approve before it sends. Routine replies might run without a hold, depending on what your team agrees in the job description.
What the weekly report actually contains
Each week, a named person at Struan signs a report your team reads. It is built from the same job description you agreed at the start, so it shows the work done against what was actually asked for. You can see a real job description, an approval record and a signed weekly report, reproduced in full, on the sample report page.
What happens when the AI employee gets something wrong
Mistakes are corrected as part of the service. When something is wrong, Struan fixes the work and tunes the role so the same mistake is less likely to recur. This is a managed correction process. It is a description of what Struan does, not a claim about a fixed error rate.
What the role can actually touch
The role only opens the inboxes, files and workspaces named in the job description. It does not have blanket access to every tool your business runs. If you want it to cover a new source, that is a change to the job description, agreed in writing, rather than an automatic expansion.
Changing or stopping the role
One instruction from your team stops the role. Struan follows a written stop procedure, and every action the role took before the stop stays in the action log. Data is held in a region you agree first, and how it is processed is set out in a data agreement before work begins. Full detail is on the security page.
Deciding what a role is worth, and what it costs
Cost and scope are set together at the start. The written quote separates building the role from the ongoing managed-service fee, including running, tuning and weekly sign off. Agree any change in scope with your named point of contact before the additional work starts. For a wider view of how Struan compares with hiring, software and outsourcing, including where Struan is not the better option, see the Compare page.
Book a sense-check call to talk through a specific role and how the approvals would work for your team. The first call is free.
If you are still deciding whether to hire an AI employee at all, What is an AI employee, and how does hiring one work? covers that first scoping decision.
You can also see real deployments written up in full on the proof page.