AI inside a global media operation

Context

A global advertiser running media across many markets and agency partners, with a strong appetite for AI and a limited tolerance for experiments that never land.

The problem

Enthusiasm for AI outran the operating model: no shared view of which tasks were worth automating, who owned the output, or how quality would be reviewed.

Approach

• Started from the operating model — process, data, ownership — before tooling.
• Prioritised use cases by the time they returned and the risk they carried.
• Set review and accountability practices so AI-assisted output had a named human owner.
• Kept a short list of experiments running rather than a long list of intentions.

Result

A clearer basis for deciding which AI work to fund and which to stop, grounded in how the media operation actually runs.