Automation
Smart Reel Intelligence System
A personal automation project that turned Instagram Reels into data-driven decisions.
Automation
Smart Reel Intelligence System
A personal automation project that turned Instagram Reels into data-driven decisions.
Client:
Internal project
(for a creator with 55K+ audience)
Role:
Product Design, Automation,
AI Workflow Design
Timeline:
~2 weeks
(built and tested solo)
Overview
Making creative strategy scalable through data and automation.
Raina is a content creator with 29M views in the past 90 days— and also my wife (the real boss). She posts videos weekly but had no scalable way to track what actually worked. I built an automated system using n8n, AssemblyAI, OpenAI, and Google Sheets to transcribe, tag, and analyze her Reels - helping her to decide what to post next, based on real data.
The Challenge
Raina creates a high volume of Reels — but figuring out what worked (and why) was time-consuming and based mostly on intuition.
Goal: Build a fully automated system that turns Reels into clear, structured insights — tracking hooks, formats, and performance, all without a team.
The Process
I followed a lean, technical-first approach — automating real problems while staying focused on clarity, performance, and decision-making.
Observe → Analyzed how Raina was handling her Reels manually
Map → Defined the core flow
Build → Set up a full automation system using n8n, OpenAI, AssemblyAI, and Google Sheets
Connect → Integrated all tools to run as one unified pipeline
Learn → Used the data to uncover patterns
Observe
Before designing anything, I spent time watching how Raina handled her Reels workflow. I wanted to see where her time went, what decisions she needed to make, and what caused the most friction.
I captured:
where she tracked metrics and notes
how she compared one video to another
how she chose what to post next
what created uncertainty
Map
Turning scattered tasks into a structured problem
I mapped her workflow end-to-end to understand the structure of the problem. The goal here was not to design a solution but to clarify how the current process behaved.
I documented:
every manual step and the time it required
all tools involved and how disconnected they were
repeated tasks across every new Reel
points where data was lost or inconsistently recorded
what information she needed but didn’t have
Patterns emerged quickly:
the workflow depended on multiple fragmented actions
important signals (hook, structure, CTA, timing) were not captured
no single source of truth existed
the process produced effort, not clarity
Mapping made it obvious which parts of the workflow could be automated, standardized or improved. This set the foundation for creating a system instead of another tool.
Build
Designing a low-code system that could operate on its own
With a clear understanding of the workflow, I designed a lightweight automation system using low-code tools. The goal was to remove all repetitive tasks and turn content into structured data without adding complexity.
I built a pipeline that performs:
video download
audio extraction
transcription through AI
metric collection
data normalization
automatic storage into a single database
This transformed a manual, multi-step workflow into a frictionless process triggered by a single input: the Reel link.
The build phase focused on reliability and simplicity. The system needed to work quietly, consistently and without supervision.
Full screenshot of the n8n workflow showing the end-to-end automation pipeline used to collect, process and analyze Reel content.
Screenshot of the n8n workflow showing the setup sequence where the system checks the sheet for new Reel links.
Screenshot of the n8n workflow handling video download, audio extraction and transcription using AssemblyAI.
Screenshot of the n8n workflow storing the transcript and preparing structured data for analysis.
Screenshot of the n8n workflow showing the AI analysis stage and the final output step, where insights are generated and written into the database.
Connect
Bringing data, insights and workflow into one seamless system
Once the core pipeline was working, I connected all modules to behave like one unified product.
The system now:
collects all content data
interprets transcripts and identifies patterns
stores everything in a clean structure
outputs insights that support real creative decisions
No UI, no dashboards at this stage.
Just a system that runs behind the scenes and turns behavior into information that matters.
This step ensured that all components were aligned, stable and ready to feed meaningful insights.
Dashboard
After the automation pipeline was stable, I designed a simple interface so Raina could understand the data without touching spreadsheets or technical tools. I used Lovable to build a lightweight analytics dashboard connected directly to the output of the n8n workflow.
The goal was clarity. Instead of raw rows, the dashboard presents:
Top metrics: reach, views, engagement and growth trends
Patterns: high-performing hooks, formats and themes
Behavioral signals: retention curves, watch-time drop-offs
Format distribution: which types of videos consistently perform
Actionable suggestions: clear next steps based on detected patterns
This interface turned the system into something she could use every day — not a technical backend, but a simple tool that explains why a Reel works and what to try next.
Screenshot of the custom Lovable dashboard displaying Reel performance, retention, format distribution and AI-generated insights.
Impact
The system removed guesswork from content planning. Instead of relying on intuition, Raina now has a clear view of what drives reach and engagement across her audience.
What changed:
No more manual work: downloading, extracting audio, transcribing and tagging each Reel
Discovery is faster: every Reel produces structured insights instantly
Patterns are visible: repeated hooks, formats and topics emerge without effort
Decisions improved: posting times, CTA choices and themes are now based on evidence
Outcomes
A consistent view of which hooks deliver reach and saves
A repeatable way to evaluate new Reels with zero manual overhead
More confident content planning backed by real data
A foundation she can scale as her audience grows
Impact in numbers
6–8 hours saved weekly
70s average processing per Reel
This system didn’t just automate tasks — it gave her a feedback loop she never had before.
Next Case:
Leannest | From zero to a full GLP-1 telehealth product in 45 days
Next Case:
Leannest | From zero to a full GLP-1 telehealth product in 45 days
This case study is confidential
Use the password from my resume.
Want to discuss a potential fit?
© 2025 All right reserved
Made with ❤️ by Renan.
Want to discuss a potential fit?
© 2025 All right reserved
Made with ❤️ by Renan.
Want to discuss a potential fit?
© 2025 All right reserved
Made with ❤️ by Renan.