When writing good sales emails, you normally have to choose between quantity or quality. Quality normally comes in the form of thoughtful personalization because no one wants to read some salesy copy paste pitch. To personalize every email would take a lifetime or a large team of sales reps. With this automation you can personalize at scale, increasing your open and response rates dramatically.

Automating both the data scraping and email generation is the key to unlocking scaled, personalized emails for your business.

How we do it:

We built this automation with Agenthub which is a no-code AI automation builder. It's like playing with legos that let you automate really complex and time consuming business workflows. Here is a preview of what this outbound sales automation looks like.

At a high level, this automation is gathering the necessary information about the lead you’re emailing by reading their company website, generating an email with AI and sending (or drafting) that email.

(If you want to jump straight into sending emails, scroll to the bottom of this post and you’ll find some ready to use automations.)

1. Read Leads from Your Google Sheet/Excel Spreadsheet

Most of our users have a bank of leads in a Google Sheet or Excel spreadsheet. This automation is making the assumption that your leads are also stored in the same way but I’ve included an example automation below that lets you plug in emails/websites one at a time.

We read the leads row by row from your spreadsheet and extract the email address and company URL for processing. You’ll notice two separate Read Google Sheets nodes in the automations, this is so we can read both the column containing the email and website separately.

2. Crawling the Website

Using this company URL (their landing page) we want to explore their content a bit to find the best page to scrape. Generally, most websites have a ‘/about’ page linked somewhere which contains a comprehensive write up of their product/vision. This is often the best piece of context to give the AI so it can write a thoughtful email.

Our web crawling node is built for exactly this. Give it a link to any website and it’ll extract all links present on that page. For example if we were to feed it https://www.agenthub.dev/ it would return…

'https://agenthub.dev/privacy-policy', 'https://twitter.com/AgentHub_AI',
'https://discord.gg/xtbrafmzC7', 'https://docs.gumloop.com/getting-started',
'https://agenthub.dev/pricing', 'https://agenthub.dev/about',
'https://agenthub.dev/blog', 'https://agenthub.dev/tos',
'https://www.agenthub.dev/blog ...

These are all the links we have in our footer and header on our landing page.

3. Select the Most Informative Page

From this list of links we want to grab their about page and scrape it. However, not all websites have an about page. Some have a /info, some have an /about-us page, some have none at all meaning the landing page itself is the best source of info. The irregularity of this task makes it perfect for AI. It’s a very straightforward task for most models to pick the best link from a list and return it.

We use an ‘extract key info’ node on AgentHub for this task because it trims all the fluff from the AIs response and is made for extracting values from large bodies of text.

4. Scrape That Page

Now that we have the exact link we’d want to scrape we can pass that into the Web Scraping node on AgentHub to extract all the text.

The output of the node will be the text of the website you gave it as if you highlighted it all and copy pasted it into a document.

5. Ask AI to Generate the Email Body

We now know exactly what the company is about and the AI is ready to craft your perfect email. We pass in their /about page content as context and write a prompt that tells the AI what we want written as well as the tone we expect.

This is where you can get creative. You can craft your perfect email by tweaking the AI prompt. Here is the prompt we use but you’re able to edit is as much as you’d like!

For example, I really don’t like when people ask to “explore synergies” so I specifically ask the AI not to use fluff words like that (I really hate synergizing).

“The context contains the content of a company's 'about' page. You are an expert in generating cold outreach emails. Please generate an email from Max Brodeur-Urbas to this company that proves we know what they do and {insert more details about what you want to convey here}.

Keep it extremely short, only 3 sentences. Be somewhat casual, don't use words like synergies or overly formal language.“

6. Ask AI to Generate the Subject

This step is quite easy and enormously effective! A personalized email subject goes a long way in terms of open rates. We feed the newly generated email body and ask the AI to generate a succinct subject.

We use the Ask AI - Single Value node on AgentHub to achieve this. It uses functions under the hood to ensure that the AI only returns the value you care about, no extra fluff before or after. If you’re unfamiliar with functions, here’s a writeup I did about them

7. Add Meeting Links/Outro

Normally, people like to have the same sign-off and call to action near the bottom of their email. Whether it’s a link to their calendly or just their email signature, we don’t need AI for this since it doesn’t change. I write my signoff once and slap it on the end of the AI generated email. We use a combine text node on AgentHub to combine the ai generated text with your pre-written sign off.

This is the sign-off I include in the automation. This is my real-calendly so book some time if you have questions and/or want to chat.

Here is a link to my calendar if you wanted to chat this week!
https://cal.com/max-brodeur-urbas/30min

____
Max Brodeur-Urbas
Co-Founder at AgentHub.dev

8. Save Email as Draft OR 🚀

We have a perfectly formatted email ready to send! We need to send it somehow, luckily we have a Gmail and Outlook integration on AgentHub which lets you either send that email out as is to the recipient or save it as a draft.

Saving the email as a draft is great for peace of mind. It’s nice to have the ability to double check things before sending them off, especially when you’re trying things out for the first time. Generally our users start sending emails directly from AgentHub after a couple of trial runs/tweaks to their automation.

Example Output

Here is an example of an email sitting as a draft in my Inbox after running this automation. I'm able to proof read it and send it off with a single click!

Prebuilt automations:

Now that you know how it works, here are 4 variations of this automation so you can start generating outbound emails with AI right out of the box.

Here’s the complete automation with both the Google Sheets and Gmail integrations

Here’s that same automation without the Gmail Sender if you didn’t want to authenticate with Gmail

And finally a simple automation without the Google Sheet and Gmail integrations so you can test out the automation without any authentication

Customizing Automations

The beauty of AgentHub is that you can customize any of these automations to match your exact needs. All of these automations are editable so you can tweak them as much as you’d and save them under your account.

Further Customization Ideas:

• Have it scrape the founders linkedin profile as well for extra context.
• Have it update your google sheet rows with the status of each lead.
• Scrape the leads from some external data source like the YC company directory
• Output email drafts to your Google Sheet for later use.
• Have it search the website for a valid email address as well as crafting the email.