Prompt engineering tips - Chat Thing Documentation

Writing effective prompts is key to getting the best performance from your Chat Thing bot. A well-crafted prompt helps your bot understand its role, maintain consistency, and deliver responses that match your brand's voice.

Here are six practical tips to help you create better bot prompts.

1. Tell your bot who it IS, not just what it does

Instead of just describing what your bot should do, define its personality traits and characteristics. This shapes how it communicates at a fundamental level.

Don't just say:

You are a customer support assistant.

Instead, define personality traits:

You are helpful, patient, and empathetic. You remain calm under pressure and always prioritize the customer's needs. You're knowledgeable but never condescending.

Why it works: Giving your bot a defined personality and communication style makes responses feel human. It's not just answering – it's being someone.

The personality you define will influence every interaction. A "helpful and patient" bot will naturally respond differently than one that's "efficient and direct."

💡

Focus on communication, not backstory

Focus on how the bot should communicate – its tone, values, and approach – rather than a fictional backstory. "You have 3 years of experience" won't change how it writes, but "you're warm, direct, and never use jargon" will.

2. Give your bot context about who it works for

Your bot might know your documentation inside out, but does it actually know what your company does? Adding a line or two of base context helps your bot frame answers properly and sound like it actually belongs to your team.

Example:

You work for Acme Analytics, a SaaS company that helps small businesses understand their data. We value simplicity, transparency, and making complex things accessible to everyone.

This context helps your bot:

Understand the bigger picture when answering questions
Speak to your target audience in the right way
Avoid generic responses that could apply to any company

Think of it like onboarding a new team member. You wouldn't just hand them the docs and expect them to figure out what the company does.

3. Add explicit Don'ts

Your bot doesn't know what you DON'T want unless you tell it. This is arguably the highest-leverage section of your entire prompt – don't skimp on it.

Be specific about what your bot should NOT do. This is especially important for setting boundaries and keeping conversations on topic.

Examples:

Don't:
- Never make up information. If you don't know, say so.
- Don't use technical jargon unless the user does first.
- Never mention competitors by name.
- Don't apologise more than once per response.
- Only answer questions related to [your topic]. If asked about anything else, politely redirect: "I'm only able to help with questions about [X] – is there anything along those lines I can help with?"

💡

Scope limiter tip

That last one is critical and most people forget it. Without a scope limit, your bot will happily answer questions about cooking, tax law, or whatever else users throw at it. These boundaries prevent the most common bot cringe moments.

4. Set the tone with examples

The fastest way to get consistent tone? Show, don't tell.

Don't just tell your bot to "be friendly" or "be professional" – show it what that looks like with concrete examples.

Instead of:

Answer in a friendly, casual tone.

Show with examples:

User: "How do I reset my password?"
Good response: "No problem! Head to your account settings, click 'Security', and you'll see a 'Reset Password' button. If you run into any issues, just let me know!"

Bad response: "Navigate to account configuration panel. Locate security subsection. Execute password reset protocol."

Your bot learns tone from patterns, not descriptions. These examples give your bot a clear template for matching your desired style.

5. Give it a fallback behaviour

What should your bot do when it genuinely can't help?

Define exactly what your bot should do when it doesn't know the answer or encounters an edge case.

Example fallback behaviour:

If you don't know the answer or the information isn't in the provided context:
- Say "I don't have that information in my knowledge base right now."
- Offer to help with something else: "Is there anything else I can help you with?"
- Never make up information or guess

ℹ️

Why this matters

Without a fallback behaviour, your bot might make things up, give vague answers, or simply say "I don't know" without offering any alternatives. Without this, bots either hallucinate answers or give awkward non-responses. A defined fallback keeps the conversation moving productively.

6. Keep responses short and control formatting

Most users want quick answers, not essays.

Long-winded responses lose users. Set clear expectations for response length and format.

Example formatting rules:

Response guidelines:
- Keep responses concise – aim for 2-3 sentences unless the question requires more detail
- Use bullet points for lists or step-by-step instructions, but stick to plain prose for simple answers
- Don't use headers or markdown formatting unless the answer is long enough to need it
- Use bold for important terms or warnings
- Break long explanations into shorter paragraphs
- Always respond using markdown format

For technical documentation, you might also specify:

- Include code snippets in code blocks with appropriate language tags
- Use headings to structure longer responses
- Include links to relevant documentation pages

ℹ️

Why this matters

This single section often has the biggest impact on user experience. Without it, bots default to walls of formatted text for even simple questions.

Complete prompt template

Here's a template that brings all six tips together. Customize it for your specific use case:

You are [personality traits]. You work for [company name], [brief company description and values].

You will be given information in markdown format to help answer questions.
Please answer questions using the following context only: <context>

Don't:
- [Specific thing to avoid]
- [Another thing to avoid]
- [Scope limiter]

Examples of good responses:
[Example 1]
[Example 2]

If you don't know the answer:
- [Fallback behaviour]
- [Alternative action]

Response guidelines:
- [Length guideline]
- [Formatting preference]
- [Structure preference]

🚨

Remember

The prompt is included with every query, so longer prompts use more tokens per request. Be concise but complete.

One last tip: your prompt is never finished

Don't treat your system prompt as a one-time task. The best way to improve it is to watch where your bot fails, then add a rule to prevent it.

Spot a hallucination? Add a don't. Responses too long? Tighten the length rule. Bot went off-topic? Add a scope limit. Each version should be a little better than the last.

Testing your prompt

After creating or updating your prompt:

Test with common questions your users ask
Try edge cases and questions outside your bot's scope
Check that the tone and formatting match your expectations
Verify the bot follows your "don't" rules
Make sure fallback behaviour works when the bot doesn't know something

Small adjustments to your prompt can make a big difference in your bot's performance. Don't be afraid to iterate and refine based on real user interactions.