Skip to main content
All CollectionsCollections
How to write good prompts for collections (personalizing summaries with one click)
How to write good prompts for collections (personalizing summaries with one click)
Hai Ta avatar
Written by Hai Ta
Updated over a week ago

When you're giving AI instructions that apply to all call summaries within a collection, imagine you're talking to a person. This way, you'll get the best results.

Here’s what it means in practice.

Example prompt. You can expand the text field to be able to write long prompts better.

1) Specify what type of information you want to see

It's important to use whole sentences and provide context around your request. This allows the AI to understand exactly what information you need and who it pertains to, resulting in more accurate and relevant summaries.

Example collection: Demo Calls

This collection would contain meetings where sales representatives showcase a product or service to potential clients (prospects) to demonstrate its value and features.

Don’t do this

prospect needs

Do this

Create a proposal that addresses the needs and pain points the prospect mentioned. Under each, list solutions we can provide.

Example collection: Designer Candidate Interviews

This collection would contain interviews where hiring managers assess potential designers to employ, focusing on their design philosophy, methodologies, and past experience.

Don’t do this

design philosophy

Do this

Summarize the interviewee's design philosophy, particularly how they approach user-centric design. Did they mention specific methodologies like "design thinking" or "agile design"? Compare their approach to our company's focus on iterative, user-tested designs. Also, note any examples they gave of past projects where they successfully implemented these principles, especially in complex scenarios like mobile app development or responsive web design.

2) Educate Wudpecker about specific keywords, terms and names

For example:

  • "Our company is spelled Wudpecker, not Woodpecker"

  • "If mentioned in the discussion, always include these in the summary: churn rate, objectives and key results (OKR), customer lifetime value (CLV), monthly recurring revenue (MRR)"

3) Specify the desired format

There are many ways to control the output layout of AI insights. Here are some examples.

  • "mention this first and after that, summarize the main discussion points"

  • “each section should have at least 3 bullet points”

  • “Make 4 headings, and name them “A”, “B”, “C” and “D”. Under each, include bullet points of details about that heading”

  • “add a table at the end with action items, responsible people, and deadlines”

  • “Write from my point of view, like 'I mentioned' instead of 'Alex mentioned'”

  • “Write in this format: [paste example text here]”

  • “make it informal in US English”

4) Mention what you don't want to include in the summary

If you've previously noticed unwanted content in your summaries, it's possible to also specifically ask AI to exclude that information. This helps keep the summaries clean and more comfortable to read.

For example, you can say:

  • “Don’t include small talk”

  • “Avoid repeating information”

  • “Exclude personal opinions or subjective remarks”

  • "Don't make tables"

Example prompt

Provide Feedback

Got any feedback regarding collections? We’d be happy to hear, so we can make the feature even better. Please leave a message for us by clicking on Help on Wudpecker’s browser application.

Did this answer your question?