Deployment 01 · Enterprise operations
Building an AI workforce inside Microsoft Teams
An organisation wanted AI to become part of its daily operation—not a small experiment or another isolated application, but a working capability inside the company.
“Prepare the weekly compliance report and send it for approval.”
Works with an organisational identity, approved tools, and controlled access to internal systems.
The action is not completed until a person with authority confirms it.
The organisation
An established organisation whose name is withheld at the client’s request. Its internal work moved through Microsoft Teams, email, and a set of systems built up over several years.
The problem as presented
The organisation wanted AI to become part of its daily operation—not a small experiment or another isolated application, but a working capability inside the company.
What we found
We spent approximately six weeks studying the organisation, its industry, its systems, and the technology available in the market.
We reached a simple conclusion: no single model, product, or framework could make a company AI-native. AI had to be introduced wherever work was already happening.
Most internal work already moved through Microsoft Teams. Building another interface would create unnecessary friction—another login, another interface, and another habit to learn.
The decision and trade-offs
A dedicated application would have given us more control over the interface, but it would have moved employees away from the place they already worked, and adoption would have depended on training.
Placing the AI workforce inside Microsoft Teams meant accepting the constraints of an existing platform. It also meant employees needed no new habit to gain the benefit.
We chose the existing environment.
The system
Each AI worker received a defined responsibility, an organisational identity, approved tools, controlled access, clear operating limits, and human escalation rules.
The workers connected to email and internal systems, so that a request made in a conversation could result in work being done in the systems that hold the information.
How people used it
Employees continued working in an environment they already knew. They delegated, asked, and followed up the way they always had.
The intelligence became part of the existing operation rather than a separate destination.
Control and permissions
An AI worker performs only the work it has been authorised to perform. Access is controlled, actions are limited to approved tools, and defined decisions are escalated to a person.
People remain in control of important decisions.
Result
The measured result has not yet been approved for publication by the organisation involved. We will publish it here once it has been.
What we learned
No single model, product, or framework makes a company AI-native. The environment where work already happens is a design constraint, not an implementation detail.
This deployment shaped how we now define responsibilities, identity, access, and escalation for every governed AI worker we build.
Making complex maritime work conversational
Do not begin with a software requirement.Begin with the work that is not working.
Tell us where people lose time, information becomes difficult to find, decisions slow down, employees repeat the same work, or an existing system holds the organisation back.
We will study the operation before deciding what should be built.