Feed a 22-page brand manual to AI—can it really plan next week’s content for you?

Many companies have brand manuals, yet next week’s posts still start from a blank page. Designers flip through guidelines, marketers look for inspiration, and the boss wants something that can be published. The manual sits in cloud storage and doesn’t grow hands by itself.

AmirMušić wrote a sponsored article on X demonstrating how he connected this broken link to Lovart 2.0’s Arti. The video shows Oovie’s brand visuals, a concept for an egg brand. Pink backgrounds, breakfast scenes, fried-egg graphics, plus “Food to charge your day” and “Oovie Brand guidelines / 2026.” Establish the brand feel first, then talk about content.

He went through dozens of Reddit discussions about brands and content creation, and three problems kept coming up: applying brand guidelines 28.6%, finding content ideas 33.3%, and content production overload 38.1%. The numbers shouldn’t be treated as precise statistics, but the order feels real: guidelines, ideas, capacity—one bottleneck after another.

Oovie is a concept egg brand. AmirMušić first made a 22-page brand manual, then put the assets into Lovart’s Brand Kit, selected product photos, and turned on “Auto-update memory” and “Self-evolving skills” in Arti. His trick: send the brand manual directly to Arti and let it help build the Brand Kit. Once that’s done, Arti can use the brand assets and preferences in later projects.

The first test was lightweight: give it a creative concept, “everyone has their own breakfast.” Arti proposed “Different mornings,” mapped different egg dishes to morning habits, and expanded it into a week of recipe content. In 2 minutes it produced 5 ideas: soft-boiled eggs, French omelette, scrambled eggs on toast, shakshuka, and poached eggs. Each had a title and a mood, and all stayed within the same campaign direction. The author writes that a few years ago, preparing this kind of concept, visualization, and approval would take 3 to 5 days. Now drafts come out in minutes. There’s a discount to apply here: a draft can be reviewed, but that doesn’t mean it’s ready to publish. He also asked Arti to propose two directions, “Slow Sunday” and “Wake up bold,” with recipes, concepts, and details, to assemble a monthly content plan.

More practical are Scheduled Tasks. AmirMušić set up tasks that let Arti search for food and adjacent-category references, explain how the visual approach applies to Oovie, and then generate drafts with the Brand Kit. Two scheduled runs produced two directions: The yolk edit, treating the yolk as a fashion object; and Runny. Jammy. Set, using yolk texture and cooking stages as the theme. They aren’t guaranteed to make the timeline, but at least they provide a starting point you can judge.

He provided two templates. Daily Moodboard answers “what can the brand look like,” turning real design references into visual directions. Brand Visual Radar answers “what can we say to the audience,” producing content ideas, headlines, copy, and visual demos. The workflow is: open Delegated tasks, create a task, paste the prompt, and replace [CATEGORY] with your own category. Oovie used eggs and egg-based breakfast products. Attach the Brand Kit, choose Daily or Weekly, set the time, target project, and credit limit, and use Run now for a test run.

One comment in the replies gets closer to the point than the 22-page manual. A commenter said, memory matters more than the 22 pages. AmirMušić replied, true thing. Another detail is worth noting too: he disliked the deformed letter t in “edit” and the unrealistic eggs, so he had Arti store two quality guidelines—check letterforms and keep food realistic. Later, when starting a new project with no old conversation or attachments, Arti still retrieved those two guidelines from Memory. That’s more useful than generating one nice image.

Some people also asked whether this can work outside Lovart. AmirMušić said you can test it in other agents using the Skill mentioned at the beginning of the article, but he thinks Lovart’s agent is ahead in graphic and brand design. One commenter said agents don’t fail to design; they just make averages; with real references and constraints, the output doesn’t look like a generic generation page. Another commenter said if you give type scale, spacing, and font rules, half the slop disappears. Both comments point to the same thing: constraints need to be specific, and they need to be remembered.

What’s valuable about this workflow is that brand assets, feedback memory, retrieval, and scheduled tasks are connected, turning the brand manual into reviewable draft production. Don’t treat it as one-click publishing. Choosing dishes, reviewing images, and revising copy still require people. For AI practitioners with development backgrounds, what’s worth examining is how constraints are stored, retrieved, and reused; for ordinary users, run one task first—don’t build a fully automated factory right away.

Original article: https://x.com/i/article/2115819643201789952

Featured video: https://x.com/AmirMushich/status/2116237303740006597

Oovie brand page: https://amirmushich.link/oovie-brand