Automated, and intelligent prospecting with GPT-4.
Lee este artículo en español
en LinkedIn:
Prospección automatizada e inteligente con ChatGPT-4
In an increasingly competitive world, improving customer attraction and digitizing business processes have become a pressing need for any business that wants to survive in today’s market. Fortunately, technology offers powerful tools to help achieve these goals efficiently and effectively.
Today I would like to introduce you to a data enrichment system that allows you to carry out intelligent prospecting in an automated way using GPT-4, and Make, my favorite process automation platform.
Take advantage of the section where I show you the GPT-4 prompts I created to hyper-personalize the messages.
The objective of this system is to achieve meetings with companies with which we have some kind of relationship with some of their employees through various sources of information and the application of advanced personalization techniques.
- We’ll use LinkedIn’s network of contacts to find the companies of people who have interacted with in the last 90 days.
- We extract data from the company website
- We will search for company information on LinkedIn.
- We will search for contacts with a specific position on the company’s LinkedIn.
- We will analyze the profile of the people, thus knowing their personality and carrying out an authentic personalization in the messages.
- We will create different hyper-personalized messages with the OpenAI GPT4 model and close more meetings.
In short, the idea is to analyze our network of contacts to detect companies that fit our ideal client.
For example, we detected the company Tesla, and Juan Carlos, a contact in this network, works for that company.
The next step is to obtain more information about the company, the website, email, telephone, and a list of people with the position of CEO or director of technology.
We will obtain a personality analysis for each person, thus understanding their behavior and knowing the form of communication we must use to personalize the different messages to be delivered.
Finally, we will create two hyper-personalized messages to send by email to the CEO or CTO and 1 LinkedIn message that we will send to the contact.
If you are looking for a solution to improve your sales processes, and attract more customers efficiently, take advantage of the section where the GPT-4 prompts created to hyper-personalize messages are shown.
Don’t wait any longer to automate your processes and improve your results!
Why automate customer prospecting:
Automating customer prospecting saves time and resources and improves the sales process’s effectiveness. By using technology such as data enrichment systems, and message personalization models, we can reach our ideal customers more efficiently and with more effective communication.
Furthermore, it is essential to note that more and more companies are adopting a more data-driven approach to selling and that personalization of communication has become necessary to stand out in an increasingly competitive market. Therefore, to stay ahead of sales trends, we must seriously consider the automation of customer prospecting.
In short, technology is revolutionizing the sales world, and we must be willing to adapt to take advantage of its benefits. Don’t get left behind, and start exploring the options that customer prospecting automation offers.
Technological stack to use
Well, going back to the topic of this article, the tools we are going to use are:
- Make our process automation platform.
- Apollo.io is the sales intelligence platform that allows you to find ideal prospects and enrich their data to convert them into customers.
- Humantic.ai is a platform that will help us assess and understand each contact’s personality and behavior using the latest generation of AI.
- OpenAI, yes, we will use GPT-4; I am joining the wave of ChatGPT releases.
- Mailtrack is an email tracking tool.
- Google Task, a Google application designed to create to-do lists.
- Google Sheet: A simple spreadsheet to record all the data we obtain.
Each tool fulfills a specific role within the process, from automating tasks with Make, prospecting with Apollo.io, personality assessment with Humantic.ai, and creating custom messages with OpenAI’s GPT-4.
However, I would like to point out that the cost of these tools can vary significantly depending on several factors, such as the size of the business, the amount of data being handled, the number of users, and the type of functionalities that are needed. Therefore, performing a detailed cost analysis before deciding to implement any system is essential.
Welcome to the world of automation!
Next, I will present the different automated processes for intelligent prospecting; all the automation has been created with Make (formerly Integromat).
Step 1: Search the websites of the companies to contact.
This will be one of the manual steps to carry out, I will present you how I do it, and I will also show you a second strategy to extract information from websites, but I am sure that you will devise other ways to obtain the data we require.
This step aims to obtain a list of companies websites to contact.
IMPORTANT: The only input data we need for the system is the website’s address.
My recommendation, use the power of the network of contacts.
Prominent sales executives say that it is easier to sell to people with whom we already have a certain relationship, those who already know us, so we will use the contact network to find the companies of the people who have interacted in the last 120 days.
The filter that I am going to apply is
- First-degree contacts.
- That they have been connected with us for at least 120 days.
- They are in Spain.
- They work in the marketing and advertising industry.
- That they have made at least 20 reactions and five comments on our publications.
- We have at least five contacts in common.
- That the company has at least 2,000 followers.
It is a good filter; it brings exciting data when applied, although the results can be further refined.
Once the filter is done, we are going to copy the list of websites to a Google spreadsheet; in this case, the format is as follows:
In the first column, we enter the list of the website; the rest of the columns will be auto-completed with the different automation that I explain below.
Don’t worry about the columns now; just keep in mind that there are columns that contain data, and there are columns that I use as controls, such as columns D, E, and F.
Now, I will show you another procedure to generate a list of websites that we can use to start a cooler prospecting.
Extracting data from websites with Simple Scraper.
We will use Simple Scraper, a Chrome extension that allows you to scrape websites from the browser without needing code.
The idea of this extension is to use it in an online directory of companies, extract the data, export it to a CSV file, and then enter it into the spreadsheet I previously explained.
You can see below how the use procedure is.
Well, at this point, we must have a list of websites in the Google spreadsheet; I will share my list with you.
Step 2: Extract data from the website, and register it in the spreadsheet.
This will be the first automation to be created with Make, and its objective is to extract the company’s telephone number, email, and LinkedIn profile from the website.
This automation becomes web scraping; if you need to become more familiar with the concept, web scraping is a Data Science process used to extract data from websites.
Input data:
- website address
Output data:
- phone #
- LinkedIn profile web address
As you can see in the following image, columns G, H, and I have been auto-completed, which belong to the data of the telephone number, email, and LinkedIn profile.
Step 3: Search and extract company information on LinkedIn.
To perform this automation, the Apollo.io platform can be used. A scenario will be created that inputs the company’s website address, searches the Apollo.io database for the necessary data, and records it in a Google Sheets spreadsheet.
Furthermore, Apollo.io allows you to generate your potential customers’ email lists and LinkedIn profiles.
The automation starts searching the Google spreadsheet for all the rows where column E (Apollo Company) has no data; we have not run the search in Apollo.io before.
For each row, we use the P (Domain) column to search for the LinkedIn profile in Apollo.io; we update the results in the spreadsheet.
Input data:
- website address
Output data:
- Company size (employees)
- Foundation year
- Industry
- Organization ID
In the following image, the columns K, L, M, and N have been auto-completed, which belong to the data of the Size of the company (employees), Year of foundation, Industry, Organization ID.
Step 4: Search for employees with a specific title on LinkedIn.
This automation aims to search LinkedIn for the profiles of the company’s employees with particular positions; in my case, I am looking for CEOs and technology directors.
We continue to use Apollo.io to perform data searches on LinkedIn. It is important to note that the search result on LinkedIn can be a list of people. Therefore, the data is grouped and inserted massively in the Google spreadsheet to optimize the consumption of operations in Make.
Input data:
- website address
Output data (list):
- Post
- Name
- Last name
- LinkedIn URL
- email status
- Country
- Phone
- Antiquity
- Organization ID
In the following image, I share the structure of another sheet that has the Google spreadsheet, in this sheet all the data of the people who work in the searched company will be recorded.
Step 5: Analyze the personality of the contacts.
Arriving at this point, we have the information about each contact; mainly, what interests us is the LinkedIn profile URL, which we will use for the following automation where we will use the Humantic.ai platform.
Humantic’s AI has proven to be a valuable tool for sales professionals, providing insights into the personality of prospects and customers that can significantly improve interactions.
Personality insights gained through Humantic AI have proven effective in driving better customer responses, more effective conversations, and ultimately helping sales professionals achieve their goals in half the time.
As you can see, to use the Humantic.ai platform, we need to create two scenarios.
In the first, we look for all the rows of the sheet where we register the people and use the URL of the LinkedIn profile to request Humantic the personality analysis; this analysis is not immediate; it takes a couple of seconds, so we must create a second scenario to rescue the data.
As I told you before, we must create a second automation to retrieve the analyses requested from Humantic, which is what the following image does.
The automation begins by searching the Google spreadsheet for the analyses previously requested from Humantic; for each one, we request the result and save the information in the Google spreadsheet.
This is the result of evaluating my LinkedIn profile;
https://www.linkedin.com/in/franciscodebritofontes/
it defines me well; what do you think?
- Advice type: high steadiness
- Advice description: They focus on the results, but can still be quite procedural and analytical about how to get there
- Advice adjectives: Thorough Evaluator
- Email personalization:
- Closing Line: Logically summarize/ask
- Subject Length: 2-4 words
- Tone Of Words: Objective, informational
- Greeting: Yes (Say something formal/usual)
- Bullet Points: Recommended
- Closing Greeting: Formal
- Complimentary Close: Formal
- Length Of Mail: Medium
- Salutation: Yes (Something formal)
- Overall Messaging: Focused on removing doubts
- What to avoid:
- Don’t nudge them to do something by using the logic that others have done the same
- Don’t focus on relationship, focus purely on the merit of your product
- Avoid making strong statements, instead invite them to agree with you by asking them questions
- What to say:
- Use phrases like ‘the ROI of this’, ‘X% more’ etc.
- Help them weigh the risks by sharing objective proof points instead of anecdotes and examples.
- Keep a professional, business-like approach; especially if you tend to get informal quickly.
- Key traits:
- Risk Appetite: They have relatively low risk-appetite and are not very likely to go for something unproven and risky
- Ability To Say No: They might hesitate a little, but will go ahead and say no when necessary (or asked)
- Speed: They are unlikely to move very fast, especially when it comes to new products or services
- Decision Drivers: ROI matters the most to them, followed by process and finally proof of results
- Disc description:
- High Steadiness
- High Calculativeness
- High Dominance
- Ocean description:
- Somewhat Balanced
- Agreeable
- Somewhat Open
I share two different profile analyses.
With this information, as sales professionals, we will adapt the approach and strategy to meet better our potential client’s needs, which would result in a higher success rate in closing sales.
In short, Humantic’s AI is a valuable tool for sales professionals who want to improve their communication skills and achieve their goals faster and more effectively.
Step 6: Create hyper-personalized messages with OpenAI’s GPT4 model.
Once we have collected all the necessary information, it is time to create personalized messages that we will use to contact our potential clients. In this article, I’ll show you how to use the OpenAI GPT-4-0314 model to produce two email messages and one LinkedIn message. It is important to note that these messages will be highly personalized and adapted to the personality of each potential client.
It should be noted that this prompt uses up to 13 variables, which makes it highly customizable.
In the GPT-4 prompts, we use variables such as {NAME}, {COMPANY}, {INDUSTRY}, which will have as a value the data collected for each profile registered in the Google spreadsheet. With this, we managed to create highly personalized messages adapted to the needs of each potential client.
Hyper-customized GPT-4 prompts.
I have attached a sample.
Message 1 – Prompt to create the first email
Message 2 – Second message by email if you do not respond to the previous one
Message 3 – First message on LinkedIn
Step 7: Set reminders to send messages.
Now that we have different messages in the sequence to contact potential clients, it is crucial to schedule reminders to ensure the messages are sent at the right time.
Next, I’ll show you how to set these reminders, so you remember to send the messages.
You may wonder why we don’t send emails automatically, thus creating a completely automated system. The truth is that there are several reasons behind this decision.
Firstly, it is essential to read and review the messages generated by GPT-4 to ensure that they are appropriately tailored to each potential client.
Moreover, not all customers are in the same time zone or respond the same way to emails sent at different times. By manually sending the emails, we can give each one a unique and personalized touch.
The following image shows the flow of the automation that I have generated. The Google spreadsheet looks for the rows where column J (Reviewed publications?) contains the value YES, and column K (Scheduled reminders?) has no value. For each row, the columns G (First Message), H (Second Message), and I (Third Message) are obtained, and a task is created in Google Task, thus scheduling a reminder in our calendar. This way, we ensure to send the messages at the right time and not forget any potential client.
You can see below how the reminders appear in the Google calendar.
Step 8: Registration of sending messages in the Google spreadsheet.
In this step of the automation, the date, and time of sending messages to potential customers will be recorded in the Google spreadsheet. For this, a scenario has been created that is activated when the message-sending tasks are marked as completed.
This registry is important since it allows us to keep track of and analyze the results of the messages sent and thus be able to evaluate the effectiveness of our communication strategy. In the Google spreadsheet, you will be able to see the details of each message sent, including the date, time, the message sent, and to which potential customer it was sent.
This information will be of great help to make adjustments and improvements in our communication strategy in the future, and thus be able to increase our chances of capturing the attention of potential clients and achieving our business objectives.
In the following image, you can see the information that is registered in the Google spreadsheet for each message that we have programmed to send.
Step 9: How to detect opens and clicks on email messages with Mailtrack.
One way to tell if potential customers engage with your emails is by using an email tracking tool like Mailtrack for Gmail.
Mailtrack tracks events in emails, such as a link being opened or clicked, and sends an email notification the moment an event is detected.
To record this data in Google Sheets, we can create an automation that waits for Mailtrack notifications and records the date and time of the event in Google Sheets. In this way, we will track the interaction of potential customers with our emails and evaluate the effectiveness of our email strategy.
The following image shows how the last open and last click events on the links containing the messages are recorded.
Step 10: Detect booked meetings.
Once we’ve gotten to step 10 of our lead prospecting, and follow-up system, we’re ready to start filling out our meeting calendar. However, we must keep sight of the fact that we need to keep all the information in our system up to date and in one place.
That is why the automation presented in this step is vital to closing the system cycle and having all the information necessary to make informed decisions. The idea is to register in the Google spreadsheet the date and time of the meeting that a contact schedules for us.
Why is it essential to have all this information recorded? The answer is simple: this allows us to analyze the results of our prospecting and follow-up strategy. We will be able to see if the prospected contacts are correctly receiving our message if they are opening and clicking on the emails and scheduling meetings with us.
The automation presented in this step starts when a newly scheduled appointment is detected in Calendly. Then, all the meeting information is obtained, from the date and time to the people who will attend and the link to the room created in Google Meet.
This information is used to create a contact in the Google contact book, which allows us to keep our book updated and have all our contacts’ information in one place. Finally, the Google spreadsheet is searched for any contact with the email scheduled for us, and the date and time are recorded in column Q (Meeting date).
In short, this step is vital to closing the loop on your lead prospecting and follow-up system.
Registering the date and time of scheduled meetings will allow us to analyze the results of our strategy and make informed decisions. Don’t underestimate it!
You can see in the following image how the Q (Meeting date) and R (Meeting URL) columns have been autocompleted.
We have reached the end; now it is your turn to act.
In conclusion, automated and intelligent prospecting with GPT-4 and Make is a tool that allows us to optimize our processes and increase our business opportunities. It is no longer necessary to spend long hours researching and searching for prospects manually; creating an automatic system allows us to do it quickly and efficiently.
If you have a small or medium-sized company, consider automating your processes with Make, since this will save time and money and increase the productivity of your work team.
On the other hand, if this article has been valuable and exciting, I invite you to share it on your social networks and comment on your experiences or doubts in the comments section.
In short, technology offers us endless possibilities to improve our companies and make them more competitive in the market. Do not miss the opportunity to take advantage of it!
Prospección automatizada e inteligente con ChatGPT-4
