Skip to main content
You own the use case. This is the path we recommend to take it from a first idea to a supervised production service, and a method you can reuse for the next one.

The lifecycle

1

Framing & POC

Pick a scope that proves business value and is feasible. Validate the idea quickly before investing. Recommendation: agree on shared selection criteria so the next use cases are chosen the same way.
2

Audit & MVP specification

Define the target scope, run a security review, and specify the MVP. Bring in the systems and data the agent will need.
3

MVP

Build an industrializable version: connectors, authentication, evaluation, and monitoring in place. Iterate with real users (UAT).
4

Industrialization

Apply engineering discipline: Git versioning, CI/CD, observability, roles and permissions, security and performance tests.
5

Go-live

Promote to production and make the solution auditable (AI Act, GDPR). Onboard the teams who will use it.
6

Operation & continuous improvement

Keep it running and secure: monitoring, KPIs via Insights, and updates to keep pace with the state of the art.

Deploying it

Deployment runs on two independent cycles: the platform moves through its own Dev, Pre-production, and Production environments with CI/CD, while each use case has its own dev and prod workspaces on the production environment, versioned by Git. This is covered in detail in Deployment strategy, including the workspace-or-self-care choice.

Next steps

Deployment strategy

Platform environments, and the two-workspace Git model per use case

Who does what

The teams and rights behind this lifecycle

Opening strategies

How to let people build without chaos