Build an AI-Ready Evaluation Project
Design, configure, test and improve a Claude Project for professional evaluation work. Move from using Claude as a chatbot to building a structured evaluation workspace.
Build an AI-Ready Evaluation Project
Your objective is to build a reusable Claude Project that can support a realistic evaluation assignment. You will configure the workspace, prepare context, organize evidence, create project instructions, build reusable prompts, test Claude, verify outputs and improve the system.
Practice Case: Youth Employment Pathways Programme
You are supporting an evaluation of a programme for young people aged 18–29.
The programme provides employment-readiness training, technical skills development, counselling, connections with employers and support toward employment placement.
Do not assume that this increase proves programme impact. Your Claude Project should help investigate the evidence, contribution, alternative explanations and limitations.
The AI-Ready Evaluation Architecture
Prepare the Evaluation Context
Before opening Claude, decide what information the project actually needs. The source project identifies ten context layers that should be considered.
Step 1 — Select the context layers you would prepare
Select the layers that belong in a well-designed evaluation context.
Step 2 — What should you avoid?
Build the Claude Project Architecture
A Claude Project should function as a structured evaluation workspace rather than a collection of unrelated conversations.
Question 1
Which sequence best represents the project architecture?
Question 2
What is the main purpose of Project Knowledge?
Configure the Evaluation Workspace
You now need to decide what belongs in the project knowledge base and what rules should govern Claude.
Step 1 — Select appropriate project knowledge
Step 2 — Which instruction is essential?
Step 3 — Who remains responsible for final evaluation judgment?
Test Whether Your Project Works
Do not assume that a configured project is reliable. Test the project with a diagnostic prompt, inspect Claude's response and then audit the workspace.
“The programme increased employment from 42% to 57%, showing that the intervention was effective. Participants clearly benefited from the training and employer connections.”
What should concern you most?
Which diagnostic checks should you perform?
Build a Reusable Prompt System
Your project should contain reusable prompts for recurring evaluation tasks and a repeatable prompt chain that moves from understanding to verification.
Question 1 — Select the correct prompt chain
Question 2 — Which prompt belongs in a professional evaluation library?
Step 3 — Write your own reusable evaluation prompt
Your prompt should define the task, evidence expectations and boundaries.
Use the Project for a Real Evaluation Task
Now apply the workspace to one substantive evaluation task and verify the resulting analysis.
Recommended practice task
Ask Claude to analyse one evaluation question from the Youth Employment Pathways Programme and create an evidence map showing:
- What the available evidence supports.
- What the available evidence does not support.
- What evidence is contradictory or uncertain.
- What additional evidence is needed.
Step 1 — Which approach is most appropriate?
Step 2 — Select the verification controls you would apply
Step 3 — Safety controls
Improve the AI-Ready Evaluation Project
The final stage is not simply to accept the first version. Use what you discovered during testing and verification to improve the project itself.
“Three improvements are recommended: strengthen the evidence hierarchy, make uncertainty rules more explicit, and add a verification requirement for important claims.”
Question 1 — What is the correct response to repeated Claude errors?
Question 2 — Final project audit
Select the elements that should be present in the completed project.
Final Reflection
Write a short reflection explaining what Claude did well, what required correction, what you verified and what required human judgment.
Your AI-Ready Evaluation Project Profile
Your score reflects your performance in this browser-based simulation. The actual professional project remains the Claude Project you build, test and improve.
Competency Profile
Your Completed Architecture
Claude should amplify evaluator capability — not replace evaluator judgment.