On this page
AT A GLANCE
Blumo AI was my first attempt at turning a generative AI capability into a product for creating marketing content.
My contribution
I conceived and built the product as an early exploration of how AI could support a creative workflow.
Key facts
Key engineering decisions
Start with a marketing problem
The exploration started with a concrete question: how to speed up a first version of marketing content.
Think beyond the model
The challenge was not only to generate text, but to turn that capability into an experience someone could use.
Open a longer path
The project became a starting point for exploring more complex AI products and processes.
Project evolution
- 2022
- Content generation
- From model to product
- Starting point
The beginning
Blumo AI started in 2022, when generative models were beginning to become accessible products for people outside the technical world.
The initial idea was fairly simple: use AI to speed up the first stages of creating marketing content, starting with an idea or some context and generating a first version to work from.
Coming from advertising and marketing, it was a natural way for me to start exploring what could actually be built with this new technology.
What it did
Blumo let users start with information about a brand, product, or idea and generate different content directions with AI.
The goal wasn't to replace the creative process. It was to shorten the distance between a brief and a usable first draft: generate alternatives, compare them, edit them, and keep iterating from there.
Guided prompts brought together the context for each request: brand or product information, the goal of the piece, a CTA, links, hashtags that had performed well, and a promotion when relevant. The interface also made it possible to choose a model and adjust the result to a character count.
What became interesting to me wasn't only what the model could generate, but how to turn that capability into something people could use without needing to understand what was happening behind it.
Reconstruction of the original experience
From brief to first draft
Bring together the context
Instead of opening a new chat every time, a professional worked in a dedicated marketing space with guided prompts and a visual interface for preparing each request.
Reuse campaigns and promotions
A promotion could be saved and applied across multiple products at once — for example, a summer campaign — so it could be incorporated naturally into generated copy.
Generate and adjust
The system generated alternatives with CTAs, links, and hashtags; they could then be compared, adjusted to a character count, and iterated on again.
Select and fine-tune
Automation sped up the first version, while the professional selected and fine-tuned the direction that worked best.
From model to product
Blumo AI was my first experience thinking about AI as part of a product rather than as an isolated tool.
I had to think about what information to ask for, how the interaction should work, what the system should return, and how to make the output useful within a real workflow. The visual interface gave non-technical people access to APIs and models without asking them to understand how they worked behind the scenes.
Over time, the tool was also incorporated by advertising providers into their native ad dashboards.
Looking back, it was simple compared with the systems I build today, but it introduced a question that has stayed with me ever since: what needs to be built around a model for it to actually be useful?
Drafts and variations in the historical interface
About these screens
These are static reconstructions, not recovered production screenshots. They use the surviving product design as a visual reference.
Reconstructed subscription screen
What came next
What started with content generation eventually opened up a much broader area of interest.
Over time, I became less focused on what a model could generate on its own and more interested in how models could become part of real products and processes: automation, retrieval, tools, agents, evaluation, and complete AI systems.
Blumo AI was much simpler than the projects that followed, but it was also the starting point.