LLM Best Practices

Large Language Model (LLM) chatbots like the ones underlying in Ask This Guy are powerful tools for accessing and managing information. To help you get the most out of your chatbot experience, here are some practical tips and important guidelines to keep in mind. Following this framework will help you craft effective prompts, ensuring better results and smoother interactions with ATG

Framework for writing good instructions (prompts)

Be clear and specific

Always try to express your questions or requests as clearly as possible. Instead of saying, "Show me the report," try "Can you show me the sales report for Q1 2024?" This helps the chatbot understand exactly what you need.

Provide context when needed

If your question is about a specific project, team, or topic, mention it. For example, "How do I submit an expense report for the marketing team?" This extra context helps the chatbot give you a more accurate answer.

Engage in dialogue

If the chatbot's answer isn't quite what you expected, don't hesitate to ask follow-up questions or request clarification. You can say, "Can you explain that in simpler terms?" or "Could you give me an example?"

Rephrase or simplify your questions

If the chatbot doesn't understand your question, try rewording it or breaking it into smaller parts. For example, instead of asking a multi-part question, ask one part at a time.

Use examples in your prompts

Providing examples can help ATG better understand your intent. For instance, if you want a summary in a specific style, you can say, "Summarize this article in bullet points, like this: example." This makes it easier for the AI to match your expectations.

Always verify important information

ATG is designed to be helpful, but it can sometimes make mistakes or provide outdated information. For critical decisions or important tasks, double-check the chatbot's answers against official sources or documentation. LLMs are advanced, but they are not perfect. Their answers may sometimes be incorrect or incomplete. Always use your judgment and verify important information.

Cite your sources when sharing answers

If you share information from the chatbot with others, let them know where the information came from. This builds trust and helps others verify the information if needed.

Respect confidentiality

Only share information that you are authorized to access. Avoid entering sensitive or personal data unless you are sure the chatbot's privacy and security settings are appropriate.

Don't share sensitive data unnecessarily

Be mindful of the information you provide. If you're not sure whether something is confidential, check with your administrator or team leader before sharing.

Avoid using only keywords

Typing just keywords or very short phrases can confuse the chatbot. Always try to use complete sentences.

Don't overload with jargon

While LLMs understand a lot of technical language, using too much jargon or acronyms can sometimes lead to misunderstandings. Keep your language simple and clear, especially when you're just starting out.

Use polite language

While chatbots don't have feelings, using polite language (like "please" and "thank you") can make your interactions feel more natural and positive.