AI Tools Best Practices

What we verified before publishing

Fact-check status: passed

The following claims were verified through Perplexity Sonar before the questions were finalised. The host can reference any of them during play:

• Machine Learning (ML) is the practice of using algorithms that learn by extracting patterns from large bodies of data[3].
• Chain of Thought prompting involves dividing a single task into more manageable steps to help large language models produce accurate and consistent results[1].
• Giving AI tools context, such as audience, purpose, and desired output type, is a fundamental best practice for effective results[4].
• Users should always review and verify AI-generated outputs for accuracy, biases, or mistakes before sharing them[5].
• Zero-shot prompting provides no examples, while few-shot prompting includes examples to illustrate the desired pattern[1].
• AI tools should be used for concepting, drafting ideas, and visualizing concepts rather than for final production like real photos[7].
• A clear workflow definition includes identifying specific goals, tasks, and where generative AI can assist, such as in writing or data analysis[7].
• Enterprise AI tools, like Copilot for Microsoft or Gemini for Google Workspace, protect company data from being used to train public models[5].
• Asking AI for many iterations, such as 50 ideas instead of 10, leverages the tool's ability to generate without fatigue or resentment[4].
• Step-by-step instructions break complex tasks into smaller stages to guide AI models more effectively[7].
• Role-playing prompts ask the AI to act in a specific perspective to shape its response toward a desired outcome[7].
• Branching allows users to edit prompts and explore alternative outcomes after an initial answer is generated[4].