Best Practices for General Assistant Queries

The General Assistant produces better results when requests are clear, focused, and well structured. This article outlines practical best practices that help improve accuracy, relevance, and consistency across common tasks. These guidelines apply to content creation, research, image generation, and working with files.

Written By Kristė Vagnerytė

Last updated 23 days ago

Focus on One Task per Chat

Each chat includes the full conversation history. Mixing unrelated tasks in a single chat can lead to:

  • Confusing or incomplete answers

  • Reduced relevance

  • Slower responses

Best practice:
Start a new chat whenever you switch to a different task or topic.


Provide Context Upfront

Including basic context in your request helps the assistant understand your intent.

Whenever possible, specify:

  • The goal of the task

  • The audience

  • The type of output you expect

Clear context reduces the need for clarification or rework.

Example

Less effective:

"Write something about our Q2 results."

More effective:

"Write a 2-paragraph executive summary of our Q2 results for the board meeting. Focus on revenue growth and key wins. The audience is C-level executives."


Be Specific About the Output

If you have expectations around format or length, include them directly in your request.

You can specify:

  • Length (short summary, detailed document)

  • Structure (headings, bullet points, table)

  • Tone (professional, neutral, informative)

This helps the assistant tailor the response more precisely.

Example

Less effective:

"Summarize this sales report."

More effective:

"Summarize this sales report in 3–5 bullet points. Use a professional tone and focus on actionable insights for the sales team."


Use Clear and Direct Language

Simple, direct requests work best.

Avoid:

  • Long, unstructured prompts

  • Combining multiple instructions in one sentence

  • Vague requests such as “make it better” without explanation

Clear instructions lead to clearer results.

Example

Less effective:

"Make this better and also add some stuff about pricing and maybe reorganize it and make it sound more professional but keep it short."

More effective:

"Rewrite this proposal in a professional tone. Add a pricing section with three tiers. Keep the total length under 500 words."


Use Files Intentionally

When working with uploaded files:

  • Upload only files that are relevant to the task

  • Clearly state what you want to do with the file

  • Ask focused questions about specific sections or data

If you no longer need the file, start a new chat before moving to another task.

Example

Less effective:

"What do you think about this?" [uploads 3 unrelated files]

More effective:

"Analyze the attached file. Create a summary of total revenue by region and identify the top 3 performing products."

Or for document review:

"Review the contract.pdf file. Extract all payment terms and deadlines, and present them in a table format."

For multi-file tasks:

"Compare the budget figures in 2024_Budget.xlsx with the actuals in 2024_Actuals.xlsx. Highlight any variances greater than 10%."


Use Deep Research When Appropriate

Deep Research is best suited for complex or broad topics that require more detailed analysis.

Use it when:

  • You need deeper insight

  • You want to explore multiple perspectives

  • A high-level overview is not sufficient

For quick or narrow questions, standard research is usually faster and more effective.


Jump to Previous Answers

In longer chats, finding earlier assistant responses can be difficult.

The Jump to answers feature lets you quickly navigate to previous assistant answers without scrolling through the entire conversation. It automatically detects assistant responses and allows you to jump directly to them.

This is especially useful when reviewing earlier results, documents, or analyses in longer chats.

💡 Pro Tip:
Use Jump to answers to quickly return to earlier assistant responses when working in long conversations.

Refine Results Incrementally

If the first result is close but not exact:

  • Ask for small adjustments

  • Clarify a specific section

  • Request a different structure or focus

Incremental refinement often produces better results than rewriting the entire request.


Common Mistakes to Avoid

  • Leaving out the goal or audience

  • Mixing multiple tasks in one request

  • Expecting company-specific knowledge without providing it

  • Using very short prompts with little context

Avoiding these issues leads to more consistent and useful output.


What to Read Next