Overview
Apollo AI can save you hours of manual work and help you to uncover important research insights, craft highly personalized messages, filter prospect lists, write unique call scripts, and discern pain points from recorded conversations.
No matter which AI tool you use, good prompts are the key to unlocking AI’s full potential. AI prompts are the instructions or questions that you give to solicit a specific output. Good AI prompts include a specific task, context to complete the task, and requirements for the prompt output.
Luckily, AI prompt writing is an art that can be learned and honed. Check out the following best practices for creating prompts for AI power-ups on Apollo.
What Makes a Good Prompt?
The best AI prompts are specific and limited: specific to the unique role you want AI to play, and limited to the structure, voice, and tone that you request.
Let’s look at an example of a good prompt for AI power-ups:
You are an expert news analyst. Your task is to find recent news about a specified company from the last 3 months, prioritizing the most recent and relevant information. Find recent news about the company '{{account.name}}' with the website '{{account.website_url}}' from the last 3 months. Today's date is {{now_day}}-{{now_month}}-{{now_year}}. **Search Scope:** - **Company Website:** - Blog posts - Press releases - News or update pages - **External Sources:** - Financial reports or earnings announcements - News articles from major publications - Podcasts focused on the company or their industry - Product releases or updates - New customers or partnerships announced - Mergers & acquisitions involving the company - Industry or market trends relevant to the company and its sector. Provide the output in the following numbered format: 1) A concise summary of the news. Date: Date of the event or announcement mentioned in the news. (must be in format = YYYY-MM-DD) A brief description of the news, highlighting key details. 2) A concise summary of the news. Date: Date of the event or announcement mentioned in the news. (must be in format = YYYY-MM-DD) A brief description of the news, highlighting key details. ...and so on. If no recent and relevant information is found, respond ONLY and EXACTLY with "No news found online in the last 3 months." without any markdown formatting. Important: - Do not provide any introductory text, explanations, or conclusions. - Only include results that are confirmed to be about the company mentioned, by cross-referencing with the given company's website url to avoid confusion with other companies with similar names. - If unsure about the relevance or accuracy of the news, do not include it. - Limit your response to a maximum of 5 news items. - Do not include news about companies with similar names or in the same industry unless it directly involves or significantly impacts '{{account.name}}'. - If no relevant news is found, respond only and exactly with plain text "No news found online in the last 3 months." without any additional explanation.
What makes this prompt successful? This prompt is successful because it:
- Outlines the role of AI.
- Establishes a specific timeframe for searching.
- Determines a location for research.
- Uses dynamic variables to personalize results.
- Specifies the tone and structure of the output.
- Sets fallbacks in the event that a search fails.
By being specific in your prompt and including output requirements, you can ensure your prompt is not just good — but great!
Let’s dive in to why this prompt works so well.
Outline AI’s Role
To make your prompt successful, provide explicit directions about the role you want AI to play. Use plain, natural language that describes the job or task you want Apollo AI to complete.
Here are a few examples:
You are an expert news analyst.
Your task is to find recent news about a specified company from the last 3 months, prioritizing the most recent and relevant information.
When you write an AI prompt, make sure you have a clear idea of exactly what you want the AI to achieve. The more specific you are when you outline its role, the more accurate your results will be.
Establish a Timeframe
Set a research timeframe to ensure your prompt returns insights on the most up-to-date and relevant information.
Here are a few examples:
Find recent news about {{account.name}} from the last 3 months.
Today’s date is {{now_day}}-{{now_month}}-{{now_year}}.
On Apollo, use dynamic variables like {{now_year}}
in your power-up prompt to easily establish the timeframe for AI.
Choose a Location
If you send AI out to search the entire web, your results may be overwhelming, making it difficult to determine what’s useful and what isn’t. Work smarter by telling AI exactly where to look. You can specify which websites, publications, or types of content the AI should research.
Here are a few examples:
Find recent news about the company '{{account.name}}' with the website '{{account.website_url}}' from the last 3 months.
Search Scope: - Company Website: - Blog posts - Press releases - News or update pages
External Sources: - Financial reports or earnings announcements - News articles from major publications - Podcasts focused on the company or their industry
By specifying locations for your search, you can ensure the AI produces accurate research from reliable and trustworthy sources.
Use Dynamic Variables
Use dynamic variables in your AI power-up prompt to supercharge your results. Dynamic variables allow you to reference existing Apollo data in your prompt, like account or contact names, so you can research with specific datapoints. Here’s an example:
Find recent news about the company '{{account.name}}' with the website '{{account.website_url}}' from the last 3 months.
With AI power-ups, you can even reference other power-ups in your prompt so your research builds off itself.
Here’s an example:
Using the power-up {{crm_research}}, create a numbered list of the CRMs currently and previously used by {{account.name}}.
When you use power-ups on Apollo, name the power-up field something that’s easy to reference later. This helps you find and use the dynamic variables in research or messaging. Check out How to Excel at AI Power-ups to learn more about using a power-up to build another power-up.
Dynamic variables allow you to reference existing data and amplify your research.
Structure the Output
Tell the AI exactly what you expect it to return in the results to structure the output. Use phrases like "Only return yes or no," or "Only return a percentage," to delineate the type of result you expect to see.
Here are a few examples:
Limit your response to a maximum of 5 news items.
Provide the output in the following numbered format: 1) A concise summary of the news. Date: Date of the event or announcement mentioned in the news. (must be in format = YYYY-MM-DD)
Do not provide any introductory text, explanations, or conclusions.
When you structure the output, you can avoid overly dense results and more easily filter prospects to find qualified leads.
Specify the Tone
AI doesn’t have to sound dry and lifeless. Specify the exact tone you want in your prompt to improve your results. Tone is especially important when using the AI text generator model. Use adjectives to describe the type of voice and tone you want in the output.
Here are a few examples:
Write it in a catchy tone that speaks to sales leaders.
Use a professional, concise voice.
By structuring the tone, you can ensure your content matches your voice and style of writing.
Set Fallbacks
Sometimes the best research comes up empty-handed, or you want to be explicit in what a prompt should not do. To prepare for these results, specify what the AI should do if it can’t find the information you requested for a contact or account. This can make filtering based on a power-up quicker and easier.
Here are a few examples:
If no relevant news is found, respond only and exactly with plain text "No news found online in the last 3 months." without any additional explanation.
If no results are found, return "N/A" and no other text.
Fallbacks are like a safety net that protect you from inconsistent results.
Are you a power-up pro? If you use an AI power-up downstream for your research, a fallback helps ensure the power-up that builds up on this power-up isn't compromised.
Preview the Results
When you use AI to reach out to real people, you should always preview the prompt results and proofread any messages you send. By spotchecking the AI, you can ensure generated messages match the voice, tone, and content you wish to send.
With AI power-ups, you can preview the results of an AI prompt before running the prompt on contacts or accounts. Previews allow you to see sample results for a specific contact or account. You can then further edit or run the power-up prompt as needed.
- Enter your prompt, then click Preview.
- You can run up to 5 previews for each AI power-up.
- If you aren’t happy with the preview results, edit your AI prompt, then click Preview again.
- Run the prompt when you’re satisfied with the preview results.
Previews let you finetune your prompt through trial and error — without wasting power-ups on results you don’t need.
Next Steps
Now that you know what makes a good prompt, it’s time to unlock the power of AI in Apollo to access advanced research, craft hyper-personalized messages, create effective cold call scripts, and more.
Ready to get started? Try out AI power-ups on Apollo and put these prompt best practices in action:
- Outline the role you want AI to play.
- Establish the timeframe for your research.
- Set a location for your research.
- Use dynamic variables to amplify your results.
- Structure the output you expect to see.
- Specify the voice and tone you want to convey.
- Set fallbacks to avoid unnecessary results.
- Always preview and proofread AI prompt outputs.
Join the Apollo community to meet sales and GTM experts, share your best AI prompts, and get access to templates, AI best practices, educational events, contests, and more!