Writing Prompts for Meeting Summaries

Writing an AI prompt can be a daunting activity. Not everyone is familiar with how to do it well. Most people simply apply trial and error to dial in the right questions to ask. Here are a few tips and techniques you can use in your prompts for your own use case.

1. Give the AI some context

We've found that giving the AI a bit of context about who you are and what your industry is, helps the AI narrow in on the right way to summarize your transcript.

For example, we'll work with the assumption we are using LazyNotes to summarize a client interview for an interior designer:

You are a meeting summarization assistant for a interior designer.

2. Describe how you want your notes summarized.

Next, add a short instruction about how you'd like your summary. You could add specifics about how many words, max, but we let the AI decide.

When I give you a transcript please try to summarize the answer 
these questions:

3. Choose the right questions for your job.

Now to the hard part - what questions to ask and how to ask them.

We recommend thinking about what important facts you want to consistently track about each meeting. Are there fields in your CRM tools that you always fill out? Every industry is different, but for our interior designer we'll add these to our prompt:

Client Asks: What did the client ask for?
Styles: What styles or examples inspire the client?
Importance: What was important to the client?
Deadlines: Were any deadlines or timelines mentioned?

For more generic meetings you might want to use questions like:

  • Topics: What are the main topics discussed?
  • Action Items: What are the main action items? Include the person/party responsilbe for each one if possible.

Here are a few example questions for different industries that may inspire your own questions:

Venture Capital

Company Summary: What is the company name and 5 word description? 
Product: What does the product do? 
Traction: How much traction does the company have?  How many signed customers or users?
Deal terms: What is the founder asking for in terms of deal terms?
Team: Tell me about the team and what companies they worked at previously. 
Business model: What is the business model and pricing?
Follow ups: Where any follow up actions discussed? If so, list them out.

Interviews

Strengths: What are the candidates main strengths? 
Weaknesses: What are some potential weaknesses of the candidate?
Red flags: Were there any items discussed that would prevent the candidate from being employed here?

Construction/GC

Topics: What were the key topics of discussion? 
Action Items: What are the action items and next steps? 
Issues: What are the subcontractor or trade issues we need to address?
Client asks: What did the client ask us to do? 
Potential problems: Where there any areas that might be a cause for concern?

Lecture notes

Topics: What are the main topics of the lecture? 
Key Examples:Were there specific examples outlined that demonstrate the concept? 
Terminology: What are some new key terms to be sure to understand? 
Implications:What are the implications of the points made in this lecture? 
Emphasis: Was anything emphasized by the lecturer? 
Exam Material: Did the lecturerer highlight any questions or specific topics that would be covered on an upcoming exam? 
Assignments: Were any upcoming assignments discussed? Is there any homework due?

One final thought on questions. We've noticed that the more concrete the questions can be, the better job the AI will do in answering. For example, the AI still will not do an excellent job answering more subjective questions like

  • How qualified was the candidate for the job?
  • Was the client happy?
  • How exciting is this opportunity?
  • How compelling was the speaker?

But language model tech is changing every day so feel free to give these questions a try. Just don't expect miracles (yet).

4. Optional: Formatting

This one is optional but improves readability. Notice that we added a one-phrase topic in front of each question (Ex. "Action Items:"). This hints to the AI to create header topics for each answer. Doing so makes the output summary easier to scan. This is optional, but highly recommended.

5. Putting it all togther

So our interior designers prompt should now look like this:

You are a meeting summarization assistant for a interior designer.  
When I give you a transcript please try to summarize the answer 
these questions:
Client Asks: What did the client ask for?
Styles: What styles or examples inspire the client?
Importance: What was important to the client?
Deadlines: Were any deadlines or timelines mentioned?

Last step is just adding the prompt to LazyNotes. Tap the Person Icon and then "Edit Recording Settings" to add your summarization prompt. (We also provide starting templates if you have writers block.)

And finally, feel free to get more creative with your prompts. We're continiually amazed at the pace of AI innovation and the creativity of our users. If you feel like you've unlocked a new capability with your prompts, we'd love to hear about it. Drop us a note at info@lazynotes.app