The same pipeline, with AI in the production seat

The need

A visual identity for Sleeper Guides that could run at Nike volume.

The constraint

A fraction of Nike's crew and budget, and a weekly sports calendar that never waits.

What I did

Proposed AI in the production seat. A fantasy platform and generated imagery are both a kind of make-believe, so it fit the brand. Set one treatment: the brand navy gradient across every hero, the title on the gradient, the same type on every article.

Why

I would not have proposed this at Nike, where the subjects were real athletes. Here it fit, and it made the volume affordable.

The need

A repeatable path from written article to finished hero, every week.

The constraint

Generated images drift off-brand and make errors; nothing can ship unchecked.

What I did

Tracked in Airtable. Human writers and editors; a shot list drafted with an AI assistant and finalized by me; a prompt agent; generation; a review agent; my final select.

Why

Put the judgment where it matters. The agents do the volume, and a person approves every image that ships.

The need

Enough capacity to hit a weekly cadence.

The constraint

A small budget and no studio.

What I did

Managed freelance producers, editors and illustrators on weekly turns. Built the prompt, review, stats and design agents that took the production seats.

Why

A crew is whoever, or whatever, does the work to the standard. Direction stays human.

The need

Hero images for every Guide, and a reason for fans to come back weekly.

The constraint

Image and video models were moving fast and were not always ready for the job.

What I did

Generated across Midjourney, Runway, Firefly and Kling. Built Fantasy Reporter, a weekly recap with an AI host, below.

Why

Use the tool that does the shot best, and solve the hard part instead of avoiding it.

The need

Every image consistent with the ones before it.

The constraint

Generated work has no shoot to go back to; the library is the only memory.

What I did

Built asset libraries and production guidelines, and pointed the prompt agent at the approved library so new images stay on-brand.

Why

The library is what keeps a generated program looking like one brand.

The need

Hold a weekly cadence.

The constraint

Weekly turns, live events, a contract with an end date.

What I did

Ran the Guides and the show every week, produced on-site at live NFL events, and handed the Fantasy Reporter process to an in-house editor.

Why

The handoff is the test: the process kept running without me.