Check ad and social images against brand rules before they run
Before an ad or social image is scheduled, a decision model checks it against your written brand rules and holds any image that seems to break one, so the marketing lead only reviews the flagged ones.
Uses image input, which TypeSafe AI labels a preview feature for Jev. Test it on your own sample images before relying on it.
An example pattern written by Agents Lookup. We have not yet found a public build of this exact job.
The use case
When several people and tools make images for ads and social posts, small brand mistakes slip through: an old logo, off-brand colours, missing disclaimers, or too much text on the image. Jev looks at each image with your brand rules written as questions and answers each with a probability. Images that clear every rule at a 0.85 confidence are marked ready for review. Images that break a rule, or where Jev is unsure, are held with the rule named. Jev does not write copy, approve spend, or publish anything. The marketing lead still looks at every image before it runs and decides on any exceptions to the rules. Start with the three or four rules that cause the most rework and add more once those prove reliable.
Example workflow
- Designers export final images to a Google Drive folder or a Canva folder synced to Drive.
- Make sends each image to Jev with the image preview header.
- Jev answers Noul questions for each rule, for example the current logo is shown, only brand colours are used, required disclaimer text is present, and text covers less than about a fifth of the image.
- Code marks the image ready or held and writes the results to a Google Sheet with the rule that failed.
- Held images are posted to the marketing lead in Slack with the failed rule and the image.
- The marketing lead approves each image before it is scheduled in the ad or social tool.
Tools
- Jev (TypeSafe AI)
- Canva
- Google Drive
- Make
- Google Sheets
- Slack
Risks and guardrails
- Image input is a preview feature in Jev 1.14 and TypeSafe says it is not production-ready yet — guardrail: test it on a few hundred of your own photos against a person's labels before relying on it, and keep a person checking every answer until the results hold up.
- Colour checks from a photo are approximate — guardrail: treat colour answers as hints and let the marketing lead decide on borderline cases.
- Rules written loosely produce loose answers — guardrail: write each rule as one clear statement and test it on images you know pass and fail.
- A passed image is not an approved ad — guardrail: a person still reviews every image and every ad platform's own policies before it runs.
Related recipe
Repurpose One Video or Webinar into Posts, Newsletter, Clips, and a Blog Outline: An agent turns one long video transcript into LinkedIn posts, a newsletter section, timestamped clip suggestions, and a blog outline in your brand voice, with a claims check and human approval before publishing.
Want it set up for you? Request a quote.
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Canva Skills (official page): github.com/canva-sdks/canva-claude-skills (checked 2026-11-09)
- Google Drive (official page): developers.google.com/workspace/drive (checked 2026-11-19)
- Make MCP server (official page): developers.make.com/mcp-server (checked 2026-11-09)
- Slack MCP server (official page): docs.slack.dev/ai/slack-mcp-server (checked 2026-11-09)