
Build a design brief classifier in code
You'll have a working script that takes messy design requests (Slack messages, emails, Figma comments) and sorts them into project type, urgency, and design system relevance so you know which ones AI
What you'll make
Design Brief Classifier
- classify_request() — the main LLM classification function: what to include
- load_overrides() — your manual rule system for systematic failures: what to include
- process_batch() — batch processing with summary stats: what to include
- main() — orchestration and CSV output: what to include
Paste your complete Python script with example output from at least 10 real design requests you've processed
The 3 parts
Each part ends with a check you can see.
Sample Your Actual Design Requests
Pull 20-30 real design requests from your channels and hand-categorize them by what actually matters to your workflow
Done when
You have 25+ real design requests sorted into 4-6 categories that reflect how you actually prioritize work, with at least 3 examples per category
Test AI Classification and Find the Gaps
Use AI to classify your sample requests directly, then compare results to your hand-labels to find systematic failure patterns
Done when
You have AI classification results for all sample requests, plus a documented list of at least 8 failure cases organized by failure type (ambiguous request, missing context, multi-category, etc.)
Configure the Classifier Script
Take an AI-generated classification script, customize it with your categories and override rules, and test it on fresh design requests
Done when
You have a working Python script configured with your categories and at least 5 override rules, tested on 15+ new design requests with documented accuracy vs your manual triage
Capstone: you put together your Design Brief Classifier.
Where this helps
An elective: a single lesson that sits outside rtwo's job certificates and doesn't count toward one.
About this lesson
What you'll be able to do
- Run a working Python classifier on at least 20 real design requests from your inbox/Slack
- Identify the 3-5 classification patterns where AI gets it right vs where you need manual override rules
- Have a documented priority system that routes design work by type, urgency, and design system impact
- Know which kinds of design requests benefit from AI pre-processing and which need human judgment from the start
Who it's for
- Design managers drowning in mixed-priority requests across channels
- Solo designers who want to triage incoming work systematically
- People comfortable running Python scripts and editing configuration files, even if they don't write code from scratch
What it won't cover
- Setting up Slack/email API credentials (assume you can export message lists)
- Production deployment or team-wide tooling setup
- Advanced design system architecture or component libraries
- Writing Python functions from scratch — we'll modify AI-generated code, not build it ourselves
Skills you'll practise
- Classification rule design from messy real-world text
- AI output calibration for domain-specific categories
- Override system configuration for systematic AI failures
- Design request pattern recognition and priority mapping