AI Use Status

Purpose

These statuses give employees and Dev a shared way to describe an AI idea.

They are not a heavy approval process. They are a simple vocabulary for answering: "What is the status of this AI thing?"

Before handoff, the employee or sponsoring team is responsible for describing the status of their own work. After the work meets the threshold for handoff, the status is shared between the original owner and Dev.

Experiment

An AI idea is an Experiment when it is exploratory and uses approved environments and appropriate data.

Typical signs:

• uses public, synthetic, or non-sensitive data;
• does not access production systems;
• can be deleted or abandoned without business impact;
• is used to learn, compare options, or test a concept.

Most early AI work should start here.

Guided Prototype

An AI idea is a Guided Prototype when Dev is helping with the approach, model choice, tools, supporting infrastructure, or product specification.

Typical signs:

• Dev guidance is useful or requested;
• the work may need company inference resources;
• the work may need AI-compatible search, fetch tools, or vector databases;
• the idea is still experimental, but the path is becoming clearer.

A Guided Prototype is still not production use.

Production Candidate

An AI idea is a Production Candidate when it may become part of a real company product, workflow, or operational process.

Typical signs:

• the prototype is useful enough to continue;
• production systems may be needed;
• sensitive data may be needed;
• real customers, employees, security, money, or operations may be affected;
• specialist ownership is needed.

Production Candidate status means the handoff process should start.

Approved Production Use

An AI capability is Approved Production Use when the responsible specialist teams have accepted ownership for their domains and approved the use for production.

Dev evaluates the technical, security, infrastructure, SRE, and DevOps parts of this decision. Other specialist teams may be involved when their domains are affected.

This policy does not define the production requirements. The responsible teams evaluate those requirements during handoff.

Rejected / Retired

An AI idea is Rejected / Retired when it should not continue in its current form.

Typical reasons:

• the idea did not work well enough;
• the risk is too high for the value;
• the need changed;
• a better solution exists;
• the prototype is no longer maintained;
• the production path is not approved.

Retiring an experiment is a normal outcome. Learning that an idea should not continue is still useful.