Using GPT-3 to Personalize Outbound Emails at Scale
Welcome to this comprehensive video tutorial on utilizing OpenAI in your clay tables. In this video, we will cover the fundamental aspects of integrating OpenAI and explore various practical use cases. If you're interested in learning how it works, you can watch the beginning of the video, and then dive into the use cases later on.
Use Case 1: Personalizing Emails Based on LinkedIn Profile
To start, we'll create a personalized email introduction using someone's LinkedIn summary. Here's how you can do it:
1. Go to the "Enriched Data" section and select "AI and GPT-3" (OpenAI).
2. Alternatively, you can search for "OpenAI" or click on "Buy Provider" and choose OpenAI.
3. Among the OpenAI options, select "Complete Prompt," as we will mainly focus on generating complete prompts rather than editing text.
4. Make sure you have your OpenAI API key inserted.
5. Now, let's proceed with creating the prompt. This step is crucial for getting accurate results.
6. When prompting, be extremely specific and clear, as if explaining the task to a fifth grader. Leave no details out.
7. In this case, we want to generate a one-sentence summary of someone's job focus based on their LinkedIn profile.
8. Start the prompt with "Tell me what this person's job focus is based on this input."
9. Map the relevant information from Google search, including the person's latest experience and their summary.
10. If the summary is not available, you can use their overall summary instead.
11. Provide a specific instruction like "Write a one-sentence summary of their job focus in under 10 to 15 words. Start each sentence with 'I notice on your LinkedIn profile, seems like you are responsible for.'"
12. Save the changes and run the prompt.
Example outputs:
• "I notice on your LinkedIn profile, it seems like you are responsible for investing and mentoring early-stage startups, driving growth within companies, and teaching compassion in elementary schools."
• "I notice on your LinkedIn profile, it seems like you are responsible for developing strategies and scenarios with organizations to renew business and improve performance through technologies and initiatives."
Note: It's interesting to observe that the AI generates different outputs even with the same inputs, which will be relevant when we explore generating a full email later on.
Use Case 2: Personalizing Emails Based on Company Description
Now let's personalize emails based on the company's mission using their description:
1. Create a prompt asking for the company's mission based on the provided input.
2. Fetch the enriched company data, specifically the company description.
3. Start the prompt with "Tell me what this company's mission is based on this input."
4. Specify the desired output as a line of 10 to 15 words that starts with "It looks like your mission at [company name] is to..."
5. Save the changes and run the prompt.
Example outputs:
• "It looks like your mission at LinkedIn is to help connect professionals around the world to advance their success."
• "It looks like your mission at Apple is to revolutionize technology, foster creativity, and empower people around the world."
Feel free to explore additional prompts for different companies and witness the accurate results generated by the AI.
Use Case 3: Extracting Google Snippets for Company Reviews
In this use case, we'll retrieve company reviews from Google snippets. Here's how you can do it:
1. Create a prompt and specify that you only want one snippet.
2. Save the changes and run the prompt.
3. You'll notice that the AI can also ingest data from snippets, providing valuable information.
4. If you need more data from the webpage, you can use the scrape website integration.
5. Create a new column and map it accordingly.
6. Normalize the data by prompting the AI to provide the employee rating for each company.
7. Make sure to specify that the answer should be in numerical values.
8. Save the changes and run the prompt.
Example outputs:
• LinkedIn: 4.5
• Apple: 4.2
• EY: 3.9
• Tesla: 3.6
Note: The AI might not always follow the exact instructions, so you can fine-tune the prompts accordingly to obtain the desired output.
Use Case 4: Prompting AI to Write a Full Email
Now, let's explore how to generate a fully AI-generated email:
1. Create a prompt asking the AI to write a casual email with 50 to 70 words.
2. Include the company description input and follow a specific template.
3. Start the template with a casual greeting like "Hello, [first name]."
4. Incorporate the company description input to show your familiarity with their goals.
5. Ask a relevant question related to their business.
6. Make the email personal and engaging.
7. Feel free to modify and improve the template based on your preferences and goals.
8. Save the changes and run the prompt.
Example outputs:
• "Hello, Noel. I noticed that your company, HSSBC, is focused on opening up a world of opportunity for customers and the planet. I'm curious how you're managing your bookkeeping to make it happen."
• "I noticed that your company, Siemens AAG, is a global technology powerhouse that has stood for engineering excellence and innovation. I was wondering if it would be helpful if someone were to manage to further drive your goals of solving world hunger."
Note: The AI-generated email examples provided may need further refinement and customization based on your specific requirements. Feel free to experiment and fine-tune the prompts to achieve optimal results.
By following these steps, you can effectively personalize lines based on LinkedIn summaries, target LinkedIn company descriptions, ingest and normalize data using AI, and even generate fully AI-generated emails. If you have any questions, don't hesitate to reach out to us at friends@clay.run or join our Slack community at clay.com/slack. We look forward to assisting you further. Thank you for watching!
.svg)
%201.webp)
.webp)
