Applied AI - Putting Google NotebookLM, custom OpenAI ChatGPTs and Anthropic Claude to work
Hands-On with Generative AI for Rapid Research and Preparation
In the latest AI Office Hours session on March 20, I walked through a hands-on demonstration of generative AI tools that can accelerate research, synthesize complex data, and prepare for high-stakes meetings. If you missed it, you can watch the full recording here: View Recording here (25 min).
Scenario: Preparing for a Last-Minute Executive Meeting
To make the session as practical as possible, I created a real-world scenario. Imagine getting a last-minute invite to present AI insights to Amazon’s executive leadership. There’s no time for traditional research methods - I need to gather intelligence, structure key talking points, and craft a compelling narrative quickly.
AI Tools & Techniques I Used
ChatGPT for Data Extraction & Transformation
Task: Compile a pursuit matrix by extracting executive bios and structuring them into an Excel-ready format.
I started by using ChatGPT’s browsing capabilities to pull Amazon’s leadership bios from their website. Instead of manually copying and pasting, I had ChatGPT transform the list into a CSV file that I could open in Excel.
There were a few hiccups - ChatGPT temporarily crashed, as AI demos love to do - but ultimately, it delivered a nicely structured dataset. This example showed how ChatGPT can support users by taking over tedious data transformation tasks - in this case by organizing key players, their roles, and relevant background information into an Excel-ready format.
Claude for Financial Analysis & Risk Insights
Task: Extract key financial metrics and risk factors from Amazon’s 10K report (a dense SEC filing).
I uploaded the first 33 pages of Amazon’s latest 10K to Claude and asked it to summarize performance indicators and risks.
Initially, Claude hit a character limit, forcing me to adjust my approach. After a few refinements, it successfully broke down the report into profitability metrics, revenue trends, and risk factors - without requiring me to sift through all 97 pages.
Claude - when prompted appropriately - can even generate a structured summary and interactive visualizations, making it easier to absorb key takeaways. This is a prime example of using AI to turn dry, overwhelming data into clear, actionable insights.
NotebookLM for Document Synthesis & Conversational AI
Task: Merge multiple sources (Amazon’s 10K + a 2025 priorities article) into cohesive talking points.
Next, I uploaded the 10K report along with an external analysis of Amazon’s top five priorities for 2025 into NotebookLM.
This allowed me to have a conversation with my research - I asked NotebookLM to cross-reference leadership principles with Amazon’s financial strategy. It not only synthesized the information but also cited exact sources, eliminating hallucinations and ensuring credibility.
I also added a YouTube video on Amazon’s leadership principles, allowing the tool to weave cultural context into my meeting prep. This turned a standard financial report into a narrative that aligned with Amazon’s internal philosophy.
Custom GPTs for Personalized Meeting Prep
Task: Build a custom chatbot to prepare for my conversation with Amazon’s CFO.
I used ChatGPT’s custom GPT builder to create a "Brian Prep Bot", named after Amazon CFO Brian Olsavsky. I loaded it with:
Amazon’s 10K report
Articles outlining Amazon’s business priorities
Brian Olsavsky’s LinkedIn activity to incorporate his most recent thoughts
By doing this, I could ask the bot for meeting openers, key discussion points, and tailored insights based on real, up-to-date information.
Instead of generic AI-generated suggestions, the chatbot pulled directly from Amazon’s strategy and the CFO’s own words, making the prep highly personalized and relevant.
AI-Generated Podcasts for Passive Learning
Task: Convert my research into an interactive podcast to review while on the go.
NotebookLM generated a 16-minute podcast summarizing Amazon’s financials, leadership focus, and priorities.
This let me passively absorb my meeting prep while multitasking - a perfect way to internalize key points while commuting, exercising, or in between meetings.
Key Takeaways
AI can rapidly convert unstructured data into structured insights, making research significantly faster.
Large Language Models (LLMs) aren’t just chatbots - they are data transformation engines that can extract, summarize, and synthesize massive amounts of information.
System limitations and quirks require workarounds, but with persistence, AI tools unlock efficiencies that would take hours or days manually.
Custom AI workflows, such as building a personalized GPT or generating a research podcast, take AI-assisted productivity to another level.
This session demonstrated how generative AI can elevate research, streamline preparation, and create dynamic learning experiences in a matter of minutes. I hope you find this recap and recording useful!