Skip to main content
Skip to content
Menu

Evaluations Beginner 3 parts Updated

Map your AI capability gaps visually

You'll have a diagram or infographic showing where your chosen AI model succeeds, where it fails hard, and what that means for which tasks you can safely hand off versus which need human review.

What you'll make

AI Capability Gap Map

  1. Success Zone: tasks where AI works autonomously
  2. Caution Zone: tasks requiring human review
  3. Failure Zone: tasks to avoid or heavily supervise
  4. Pattern Key: visual legend explaining the mapping criteria

Upload an infographic or diagram you can reference when deciding how much to trust AI on different types of work

The 3 parts

Each part ends with a check you can see.

  1. Stress-test your AI workflows

    Find where your AI model breaks by testing it on edge cases of your real work tasks.

    Done when

    You have 10-15 test cases across your work tasks with clear pass/fail results and identified 3+ systematic failure patterns.

  2. Categorize your oversight needs

    Sort your work tasks into trust levels based on your stress-test results.

    Done when

    You have three ranked categories of tasks with specific criteria for each zone and at least 2 tasks per category.

  3. Create your capability gap visual

    Design an infographic or diagram that clearly shows your AI trust zones and decision criteria.

    Done when

    You have a visual map someone else could use to understand where you trust AI vs require human oversight, with clear visual distinctions and a usable legend.

    Capstone: you put together your AI Capability Gap Map.

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

  • Create a visual map showing where your chosen AI model succeeds vs fails on your actual work tasks
  • Identify 3-5 specific task types where AI needs human oversight vs can work autonomously
  • Have a tested methodology for spotting AI evaluation blind spots in your own workflow

Who it's for

  • Anyone using AI for multiple work tasks who wants to know where to trust it vs double-check
  • People who've noticed AI gives inconsistent results but haven't mapped the pattern
  • Teams deciding which workflows to automate vs keep human-in-the-loop

What it won't cover

  • Building custom evaluation benchmarks or datasets
  • Technical model comparison across different AI providers
  • Statistical significance testing or formal confidence intervals

Skills you'll practise

  • Task-based capability mapping
  • Evaluation bias detection
  • Visual information design
  • Systematic failure analysis