Welcome to Week 9 of our NegoAI series.

In this week's video, we are going to tackle practical challenges in utilizing our behavioral assistant, specifically focusing on enhancing the depth of information we can provide it for more comprehensive analysis.

Problem 1 - Shallow Linkedin Scraping

We've made significant progress with our upgraded behavioral assistant, but we've encountered two key hurdles.

The first is the limitation of shallow LinkedIn scraping within our current system, which only captures headings and lacks crucial details like the "About" section and in-depth experience descriptions.

This restricts the quality of information we can feed the assistant, even though the assistant itself is very comprehensive.

Problem 2 - Character Limit Constraint

The second challenge is the character limit imposed by platforms like Microsoft Copilot and ChatGPT, where many of our clients utilize the assistant.

Our current behavioral assistant, at roughly 85,000 characters, far exceeds the 8,000 and 10,000 character limits of these platforms, necessitating a need to streamline it.

Today’s Focus: Solving Problem 1

Today, we address the first problem: how to furnish our assistant with more comprehensive LinkedIn data to facilitate a more robust profile analysis.

Our solution involves using the "Website to Markdown" tool.

This tool allows us to convert the main content of a web page, like a LinkedIn profile, into a Markdown file.

By attaching this Markdown file to the behavioral assistant within Typing Mind, we enable the assistant to convert the Markdown into a detailed JSON file.

This process captures a wealth of information beyond the basic headings, including skills, recommendations, courses, groups, newsletters, and publications.

This enriched dataset is significantly more comprehensive than the default output from our system's built-in scraper.

With this expanded profile data, we see a significant increase in the assistant's confidence score, rising to 95%.

This leads to a much fuller behavioral analysis, providing insights such as the Thomas-Kilmann conflict-mode profile, tips for interacting with a collaborative style, transparency notes, inference basics, and key caveats for interpretation.

This Week’s Exercise

Take 20 minutes to analyze the LinkedIn profile of someone you know professionally.

1. Instead of just skimming the headlines, try to go deeper. Read their "About" section carefully, look at the details within their experience entries, and note any skills, recommendations, or groups they are part of.

2. Think about how this more comprehensive information might inform your interactions or negotiations with them.

3. Consider what insights you could gain by using a tool like "Website to Markdown" to capture and analyze this richer data.

Preview of Next week

Next week, we'll tackle the remaining challenge: compressing our 85,000-character system instructions to fit within the character limits of platforms like Copilot and ChatGPT.