Show HN: Brand Agent – plans and creates full on-brand marketing campaigns

I’ve been building an AI design tool called Brandiseer, and I’m now working on a new component inside it called the Brand Agent. I wanted to share what I’m building and get feedback from the HN community.

What it does

Most AI design tools generate nice one-off images, but they don’t understand a brand or the broader context of why something is being created.

The Brand Agent tries to handle the part before the design: planning the actual campaign.

You tell it something like:

“We’re launching a new product line next month.”

“I’m opening a second store.”

“I need a Christmas campaign.”

“We’re running a limited-time offer.”

…and it generates:

A structured campaign plan Key messages, target audience, asset list, timeline.

A checklist of required visuals Example: hero banners, announcements, ads, product visuals, social graphics, email assets, etc.

All visuals automatically generated Everything is created using your brand’s actual identity — colors, fonts, tone, patterns — which Brandiseer learns by analysing your website or uploaded assets.

Editable assets Each visual can be edited or regenerated while keeping the same visual language.

Why I’m building this

Most small businesses and solo founders don’t have a designer or a marketing person. They open Canva or an AI generator, stare at a blank canvas, and have no idea what assets they even need.

The real bottleneck isn’t design ability — it’s planning. “What should I create?” is a harder problem than “How do I make it look good?”

The Brand Agent tries to remove that friction.

How it works under the hood

I use GPT-4.1 and Gemini 2.5 Flash for reasoning + planning.

The brand profile is generated from website analysis (colors, fonts, tone, patterns, layout signals, etc.).

Visual generation uses a combination of on-brand prompts + editable assets.

Everything is stored in a brand memory graph so the system stays consistent across multiple generations.

Early prototype

I have a working demo that shows the agent planning and generating campaign assets. It’s still rough, but the workflow is already useful.

What I’d love feedback on

Use cases I’m missing

Where AI planning tends to break

Whether this fits into existing SMB workflows

How to make the agent more deterministic and reliable

Any concerns around consistency, UX, or over-automation

If anyone’s interested in testing it or giving feedback, I’d appreciate it.

Happy to answer questions. — Daniel