How to Make the AI Classify Your Leads on LinkedIn

Author

Robert Hangu
January 06, 2025

Anyone who has ever sent out a large number of messages on LinkedIn knows this: Sending is easy, replying is hard!

Why?

Because once you’ve segmented your target audience, it becomes straightforward to personalize the message – thanks to AI. (If you don’t know how, you can watch a video I’ve created on this topic.) Plus there’s no past chat history to review. You start off fresh.

But here's the tricky part: managing the flood of responses you receive. The more cold messages you send out, the harder it gets. Each prospect responds in a different way, so each chat becomes unique. And the longer the conversation, the more time-consuming it gets.

It's a multiplicative problem: number of responses x average length of chat.

Let’s say you message these people:

Sarah – Marketing Manager at Tech Innovations

James – Product Designer at Creative Solutions

Emily – Sales Executive at Global Enterprises

Hi Sarah, I came across your profile and was impressed by your work in digital marketing at Tech Innovations. I’d love to connect and share insights on effective strategies in our industry! What are you currently working on that you’re excited about?

Hello James, I noticed your portfolio and your innovative approach to product design. I’m currently exploring collaboration opportunities and would love to hear about your experience at Creative Solutions. What are you currently working on that you’re excited about?

Hi Emily, I’ve been following Global Enterprises and am impressed by your sales achievements. I’d love to connect and learn more about your strategies in building client relationships. What are you currently working on that you’re excited about?

This is easy to compose and send. Notice the underlined question at the end? That’s where things start to get tricky:

Sarah’s reply

James’ reply

Emily’s reply

Hi Rob! Thank you for reaching out. I appreciate the kind words! Right now, I’m focused on launching a new digital campaign aimed at enhancing our customer engagement. It’s a bit challenging, but I’m excited about the potential impact. Let’s keep in touch!

Hey Robert! Thanks for the message! I’m currently working on a groundbreaking product design for a new app that I believe could revolutionize user experience. I’d love to chat about collaboration opportunities! If you’re in town we could grab a coffee next week.

Hello! I really appreciate your interest in my work. At the moment, I’m diving into a project that involves refining our sales strategy to better align with client needs.

Sarah is polite, but maintains a friendly distance, James is obviously open to collaboration, and Emily… well, she doesn’t even bother to properly finish her response.

You can’t know this without reading each message!

But AI can: That's where few-shot learning comes in!

It’s tightly connected to the EDI framework—Examples, Data, Instructions. You teach an AI model via a series of examples how to classify each message: “interested”, “cold”, etc. Then you’ll eliminate uninterested leads (like Emily), and you can let the AI draft appropriate responses for the interested ones.

First you get all your LinkedIn messages into Google Sheets via the “Bardeen” web extension. Once in table format, you’ll need the “GPT for Sheets” add-on. This is the solid foundation on which I run my automations and experiments – you can read about it in a previous email and watch a video.

This is a how the messages look like (click the image to zoom in):

The conversations are in column “Message Thread”. From there the AI extracted various other points of information, including a lead classification, and a follow-up message in my own tone and writing style.

Then I can filter by lead stage and have a ready-to-use follow-up message. Not bad!

For this I’ve used two prompts in succession: one that classifies the conversations and another one consists of multiple message templates, used selectively, based on lead stage. This is done with a =SWITCH() formula: 

=SWITCH(K2, "no engagement or superificial", "that's good to hear! what's the biggest challenge in [insert contact's core activity here]? cheers", "talk about themselves but no question", [...], ""curious about me", [...], [...])

It’s effectively a mini AI agent right inside Google Sheets!

Then I’m only left with adjusting the message drafts for any errors and sending them out.

I won’t bore you with the complete prompts here. But if you like to learn more about how to build these automations yourself, join the Make Work Obsolete community for $69/month. You’ll get access to the complete prompt handbook, a full AI course curriculum, and weekly Q&A calls hosted by AI tech experts. Try it 7 days for free, no strings attached!

I’ll see you inside!

Cheers,
Robert