A public brand guide becomes a reusable SharePoint AI skill that judges any file against tone, color, typography, and photography rules — across whatever file types you throw at it.
The demo starts with Uppababy's online brand guide — a deep, publicly shared compliance document covering color palettes, typography, photography, and tone. That guide is the corporate knowledge the AI will enforce. Without it, the skill is a generic taste check; with it, the skill is a real compliance reviewer.
The narrator deliberately picked Uppababy because their guide is unusually thorough — color hex codes, photographic style rules, allowed and forbidden language. That depth makes the AI's judgements more like "here's the rule it violated" and less like "it doesn't feel right."
The Brand Review Skill is a SharePoint AI Skill — a reusable building block that takes a file as input and returns a structured review as output. It encodes the brand guide once and applies it everywhere. Once built, anyone in the org can call it; it lives in the skill catalog, not in someone's head.
This is the difference between a one-off prompt and a productized capability. A one-off prompt lives in chat history. A skill lives in the library, has a name, can be called from buttons, and can be governed centrally.
The skill handles whatever file format you give it — Word documents, PowerPoint decks, images, PDFs — and returns the same shape of review for each. The format detection and parsing happens inside the skill; the caller doesn't have to write a Word-reviewer and an image-reviewer separately.
This matters because brand compliance is rarely about one file type. A campaign produces decks, social images, copy docs, and emails simultaneously. A reviewer that only handles PDFs would force a workflow split; one that handles everything keeps the workflow whole.
For each file the skill produces two outputs: a numeric score and a written explanation. The score lets you sort and filter; the explanation tells you what specifically to fix. Same prompt, same brand guide, same scoring — the AI is consistent because the skill is consistent.
In the demo, one file passes and the others have issues. The notes describe what's wrong and what to fix — for example, color usage outside the palette, off-brand tone, photography that violates the photographic style rules. That's the level of specificity the public Uppababy guide makes possible.