From Concept to Creation
In this module, you will build a simple AI assistant that solves a real problem. This is not a toy exercise — you will define a problem, design prompts, test outputs, and iterate until your assistant produces reliable, useful results. By the end, you will have a working AI workflow you can demonstrate and explain.
Step 1: Define the Problem
Every good AI system starts with a clear problem statement. What task do you want your assistant to handle? Who is the user? What does a good output look like? Be specific. 'Help with studying' is too vague. 'Summarise textbook chapters into flashcard-style Q&A pairs for exam revision' is a buildable problem.
Step 2: Design Your Prompts
- System prompt: define the assistant's role, expertise, and behaviour rules
- User prompt template: structure how user input is formatted for the AI
- Output format: specify exactly how the assistant should respond
- Error handling: what should the assistant do when input is unclear or outside scope?
Step 3: Test and Iterate
Run your assistant with real inputs. Does it produce useful output? Is the format consistent? Does it handle edge cases? Test at least 5 different inputs and note where the output breaks down. Then refine your prompts based on what you learn. This test-and-iterate cycle is how professional AI systems are built.
Key Takeaway
Building an AI assistant follows a clear process: define the problem, design prompts (system + user + output format), test with real inputs, and iterate. The quality comes from the iteration, not the first attempt.