Understand the work
We map the task, the people involved and the systems that contain the information. Existing examples help establish what a useful outcome looks like and where errors matter.
Building an application around a specific task.
We start by understanding what people do today and what needs to change. The scope is agreed before development begins.
Design and test
We choose an approach for the task, define permissions and build a version that can be tested against representative inputs. For AI features, testing includes checking sources, unsuitable answers and the cases where a person needs to intervene.
Put it into use
Deployment planning covers access, monitoring and the people responsible for the process. We agree on support and changes separately, according to what the system needs.
What the brief needs to settle
A useful first version depends on a defined task and access to suitable test material. The brief can begin with an existing document, a sample request or a description of where work currently stops.
Which information can the application use?
We identify source systems, ownership and access requirements before connecting them. Sensitive records and permissions need a suitable plan for the project.
How will the team check the result?
For AI output, examples should include ordinary requests, ambiguous questions and cases with missing or conflicting information. We agree who reviews those results and how errors are reported.
What happens after release?
The agreement can cover deployment, monitoring and further changes. The people responsible for the application need a way to report a problem and understand its status.