LLMs for People matching

How good is ChatGPT at people matching? Let's try!

First, let's be clear on what we want to test. "People matching" could have several meanings: skills, prospects, or even dating mates!

In this experiment, I want to search among a network of people to identify who among them could be a relevant prospect to engage in selling my services.

Let's assume the LLM was pre-loaded with information about what I am selling, where, and how, along with a list of prospects and additional insights that could have been captured from public sources or a CRM.

Then, I will ask the LLM to:

1. make an ordered list from the most promising to the least promising contact names,

2. explain its reasoning in detail,

3. suggest the best way to reach out to them (what should be the message, how it should be conveyed, and how).

To assess the results, I have exported my list of contacts from LinkedIn, which presents a short version of who you are connected with, the position of the connection, the company, as well as other meta-data.

The model gave a ranking of those contacts, from who are the most likely to be a decision maker versus the others.

Can this list be used as a "plug and play" with automation in place that would reach out to those contacts? No, the data in there are sometimes not accurate, not making sense, and the approach feels very generic and spammy.

Was the list of some value for me? Definitely, the LLM was able to provide some relevant perspectives that I would qualify as "signals" I would not have been able to see by myself.

Conclusion: It's just a matter of time before filters will be even more precise, able to manipulate data sorted in natural language, and lead us to even more automation. 

What I have just explained is actually an example of agentic workflow.