AI Workflow
End-to-End Design
to Scale User Growth
An end-to-end AI workflow to scale dating app visibility.
FAVORS has no marketing team. Instead of doing everything manually, I built an automated system where an LLM generates carousels directly in Figma via MCP and Buffer publishes them across five channels, while I keep control wherever it actually matters.
Production time per carousel dropped from two hours to five minutes, a 24x speed boost that scales user engagement without increasing operational overhead.
Role: Product Owner, System Design, Content Strategy
Tools: Claude, Figma (MCP), Flora, Buffer, RevenueCat
more posts in 11 days
Problem
FAVORS has no marketing team. Content only happens when someone sits down and makes it.
Every post needed someone to make it. About two hours per carousel.
TikTok rewards daily output. Two founders can’t keep up while building the product.
Lowercase Zurich German for a Gen Z audience. Generic tools miss it.
Hypothesis
If content production becomes a system instead of a task, two founders can post daily on five channels and test enough formats to find the ones that convert. Expected outcomes:
<5 min
production time per carouseloutput on every channel
in new users per day
The system
Figma is the hub. Claude, connected through MCP, plans each story and builds it directly in the Figma file as a carousel with a fixed arc: hook, content, call to action. Buffer then plans and publishes on three creator accounts and two brand accounts.
Why Figma and not a template tool: every carousel stays editable by hand, and one file is the source for design and automation.
Format library
Each format has its own job: tips, mistakes, founder stories, product walkthroughs and short conversion posts. Every format follows the same arc, so they can be compared. A format stays when it holds attention and drives installs.
Human in the loop
I approve every story, check every visual and release every post. The model is fast at producing options and bad at knowing when something is off-brand, unfair or simply not funny.
Measurement
Posts and views from Buffer, new users per day from RevenueCat. Daily peaks matched against what went live that day. Method: 11 days before the launch on 6 October 2026 compared with the first 11 days after.
Results
The strongest day since the app’s launch, 359 new users, came on day two of the system. That’s correlation, not proof. But the floor moved too: daily new users settled around 75, compared with around 30 in early September.
#7 in Social Networking
We made it to #7 among the top free apps on the Swiss App Store in October 2026, right between Discord and WeChat. Seeing FAVORS climb the rankings alongside apps of this scale was a pretty special moment. The rankings update live, so every day brings a new position.
Decision
People follow people. Over 30 days, the creator account earned about four times more views per post than the brand account (5,500 vs. 1,400). New formats are written in first person, the brand account carries product and conversion posts.
Key learnings
Rules beat prompts
Written skills keep every carousel on brand. 35 carousels so far from one rule set.
People follow people
The creator account earns about four times more views per post than the brand account.
Reach scales, depth doesn’t
Engagement fell from 0.83 % to 0.48 %. More output needs formats that start conversations.