We want to synchronize the data of the meeting towards a CRM, and we chose HubSpot. Even though HubSpot has a native integration, we wanted to enhance that integration. So, in the first part, we’re enriching the contact and company information.
I created a Make agent here with some configuration to analyze the contact. The user provides the English data for the provided email address. The system enriches that address and uses different tools. One tool searches for the person in Apollo to get details based on an email, while another searches the web or looks up a company, for example, for the Romanian CUI, the official registration number. With that, you can go to another platform like Terminate to get more official information about the company.
The way the AI agent works is simple: you give it instructions, and the AI will decide which tool to use. It then comes back with the information as needed. You can enrich this with additional details like company revenue, employee history, or any other information you can find online or in platforms like Nepal.
For simplicity, I’ve added a few things here. I’ll run this now to show you how it works. First, I get the meeting participants, take one participant, and pass that over to the AI. The agent will then trigger other scenarios. If we go into the history of this, you can see it was triggered for Mario. For Mario, we have all the details from Apollo, such as his role as Head of Projects, LinkedIn profile, and more, which you might use for outreach automation.
The AI agent might also use the company searcher. We use web search with OpenAI to get quick answers. In the history, OpenAI returned details about the company and other relevant responses. We then move to Make it Future and continue.
Sometimes, the Make agent doesn’t respect only returning the answering JSON. I know that it’s under development, and improvements are coming. To make sure I always get a correct JSON, I use a completion module from OpenAI to check and correct the JSON, ensuring it’s valid and can be parsed for my next steps.
We’re running short on time—about 4 minutes left—before we move to Q&A. So once I have the data, I synchronize it towards the contact. I connect the fields like first name, last name, email, phone number, LinkedIn profile, and anything else I want to connect to the company. I also check for the company, and if it exists, I update it; if not, I create it. Then I create a deal, and the last step here is a follow-up check.
I haven’t set this up yet, but I will now. For example, I want to analyze the meeting and decide if I need to follow up with the customer. To do that, I need the meeting information, and I’m missing the MeetGeek transcript. I’ll map it as a text.
Once I pass the transcript to the AI, it will analyze it and tell me whether a follow-up is needed. If so, it returns the information in JSON format, indicating whether a follow-up is required. I could expand this to include other steps, like updating the deal stage based on the discussion.
For example, I could analyze the meeting and decide which stage is most appropriate for the deal. I can then check the result, create tasks, and assign them.
That’s the core idea. The limitations are based on your imagination. You can create follow-up tasks, update deal stages, and much more. I’ll share the JSON templates with everyone. Feel free to modify them, and reach out to me and my team for support.
