Deployment Files
Evidence from agents doing real work.
Demonstrations are designed. Deployments are discovered. The Deployment Files record what actually happened when agents encountered real systems, real permissions and real customers.
Organisations are described only with their consent, and may be anonymised where commercial sensitivity requires it.
Published files
Available now.
Deployment Playbook
How to run a controlled AI-agent pilot →A useful pilot does not prove that an agent can produce an impressive output. It determines whether a defined workflow should be deployed, remediated or rejected.
James Lawrence · 18 minutes
What each file examines
The questions a pilot rarely answers on its own.
- 01What work was the agent given, and what authority came with it?
- 02Which parts of the workflow needed redesigning before the agent could operate?
- 03Where did the system fail, and how was failure detected?
- 04What supervision, escalation and recourse were required?
- 05What did the deployment cost—in licence, integration and human time?
- 06What measurable commercial result followed?