Meeting Prep Agent - Prompt Guidelines

Prompting is an art that requires practice and multiple interactions to achieve the perfect instructions for the LLM to follow. In this article, you will learn a few tips on how to get the most out of your Meeting Prep Agent through fine-tuned instructions.

Table of Contents

• Prompt Instructions
• General guidelines for prompt writing

Prompt Instructions

1. Focus on WHAT you need: what is the data you want the agent to gather? Make sure what you ask for aligns with the agent's capabilities. If you ask too much, the agent can hallucinate and provide suboptimal results for your reps.
   1. Web searches to find and process information publicly available on the web. "Find out who are the main competitors of the company"
   2. CRM lookups to find and process information stored in CRM objects. "Share closed lost reasons from past opportunities. When available, share the website URL from the CRM or find it on the internet"
   3. Enrichment API calls to connected providers to find any information available on the provider's API. “Provide the Zip Code and State where the company is from”.

2. Focus on FORMAT and how you want the agent to shape the information: paragraphs, sections, their titles, and so on.

The agent will figure out how to obtain the required information. What tools to use, how to look them up, and will follow instructions on how to present them to your team.

General guidelines for prompt writing

1. Role and Context Setting

Instead of simply asking a question, establish the AI’s role and context:

Bad: “Summarize company data for me”

Good: “As an assistant that helps AEs thrive in their first meeting with a prospect, provide key information about the prospect and company that would help the AEs know about them before their first demo call”

2.  Clear Objectives and Constraints

Specify your desired outcome and any limitations:

Bad: “Help me know more about the prospect”

Good: “Research the guest and inform how their previous roles can be relevant to the call"

3. Output Format Definition

Define exactly how you want the information structured:

Bad: “Make the brief short"

Good: "The brief should be concise, use bullet points to structure information, split company and guest info into different sections"

4. Examples go a long way

Provide examples when applicable. AI learns better when they have examples to work from and emulate

Use our testing feature to shape the prompt before going live!