🔤 Words to Know
Mission 4 — Key Terms
The vocabulary you need to understand and apply this mission's concepts. When in doubt, ask your AI Buddy.
These terms appear across the briefs, challenges, and prompts for Mission 4. Read them before you start, or use this page as a reference while you work.
All Terms — Mission 4
Classifier
Term
An AI step that reads input and outputs a single label to indicate its type or category.
Example: A classifier reads customer feedback and outputs "COMPLAINT" so that feedback gets routed to the problem-solving path instead of the thank-you path.
---
Router
Term
Logic (if-then) that directs input to different paths based on a classifier label. The router contains no AI - it's pure traffic control.
Example: A router in an email system reads the classifier label "SUGGESTION" and sends the email to the suggestion-handling path, where it gets documented and forwarded to the product team.
---
Logic Layer
Term
The decision-making intelligence added to a workflow between input and response. It's what makes a workflow smart instead of just fast.
Example: A support workflow gains a logic layer that classifies incoming tickets as urgent vs routine, then routes urgent tickets to immediate response and routine tickets to scheduled batch processing.
---
Decision Tree
Term
A branching structure where each branch represents a different input type and its corresponding actions.
Example: A decision tree for expense reports has three branches: (1) routine expenses under $500, (2) unusual expenses, and (3) policy violations. Each branch routes to different approval processes.
---
Input Type
Term
A distinct category of input that should be handled differently from other categories.
Example: In an employee survey workflow, input types might be satisfaction (satisfied, neutral, dissatisfied), department (sales, operations, support), and tenure (new, mid-career, senior).
---
Classification
Term
The process of assigning a label to input based on its characteristics.
Example: A focus group note is classified as "feature request," "competitive concern," or "user frustration" based on what's being discussed.
---
Tiny Decision Canvas
Term
A one-page template for defining input types, their meanings, and corresponding actions before building the system.
Example: A customer service canvas might show: PRAISE (customer satisfied, action: thank and ask for review), COMPLAINT (problem reported, action: apologize and solve), SUGGESTION (idea offered, action: thank and document).
---
Branching Logic
Term
Conditional pathways in a workflow where different inputs take different routes.
Example: A hiring workflow branches based on job level: entry-level candidates go through fast-track screening, mid-level candidates go through detailed interview rounds, and senior candidates go through executive interviews.
---
Personalization at Scale
Term
Treating different inputs differently (tailored responses, different tones, different actions) while maintaining efficiency and consistency.
Example: A customer service system handling 10,000 emails per day classifies them and routes them: angry customers get immediate executive attention, satisfied customers get a thank-you email, suggestions get routed to product team. Each type is handled right at scale.
---
Conditional Path
Term
A route through a workflow that is triggered only when a specific condition (like a classifier label) is met.
Example: An expense approval workflow has conditional paths: expenses under $500 auto-approve, expenses $500-2000 need manager approval, expenses over $2000 need director approval.
---
System Prompt Constraint
Term
A specific instruction in an AI system prompt that limits or controls the AI's output format or behavior.
Example: A classifier might normally say "COMPLAINT (the customer seems frustrated about shipping)" but with the constraint "Output ONLY one word" it says simply "COMPLAINT".
---