Analyzing a 10-K with ChatGPT

Introduction

This post is second in a series on how AI can help with various aspects of researching an account, one of the core parts of the job of a sales rep. As a sales rep, it’s important to really understand your customers: their business, their goals, their fears. The more you understand them, the better you’ll be able to help them and in turn earn their trust because your approach is authentic to their specific needs — they’ll feel like you’re part of the team, not just another vendor.

• Part one: Basic account research with Perplexity.ai

• Part two: Analyzing a 10-K with ChatGPT

• Part three: Analyzing an earnings call transcript with ChatGPT

In our previous article, we walked you through using Perplexity AI to get some basics about a company really quickly. Now we want to help you learn from more complex, time consuming sources such as company financials.

Analyzing a company’s 10-K is a no-brainer for a savvy sales rep. A 10-K—which is published yearly by publicly traded companies—is packed with info about a company's financials and strategies, but it's also a beast to read. Lucky for us, GenAI tooling like ChatGPT is here to help! We are using ChatGPT instead of Perplexity which was used in the previous post on company research, because while Perplexity is really good at generating and analyzing Google-style search results using general public web info, GPT is excellent at analyzing a specific file or dataset. In this case, what we want to do is give GPT the 10-K and ask it to go to town.

Analysis walkthrough

In continuation from our last article, we’re a sales rep at Endgame learning about Zscaler, a new prospect in our book. Like I said, 10-Ks can be a beast—Zscaler’s 2023 10-K is 157 pages (!!)—which is where ChatGPT comes into play. Let’s give this a spin:

1. Download the latest 10-K from Zscaler’s investor relations website. Most 10-Ks are readily Google-able—and if not, ALL are available via the SEC (more on this later).

2. Kick off a new chat with in ChatGPT using GPT4—note that this does also require ChatGPT plus—and upload the 10-K.

3. Give some context, and ask away!

There’s enough content floating around crafting the ultimate prompt that I wont delve into too much detail there, but one easy way to prevent GPT4 from speculating too wildly (i.e., hallucinating) is to simply ask it to give you a detailed citation of where the answer was found in the 10-K.

A few quick examples of context (along with sample prompts) in a 10-K that can help someone work deals:

• Business Overview: Understand the company's characteristics and priorities to align your offerings. [Give me an overview of the company’s business model.]

• Risk Factors: Recognize company challenges and risks to position your solutions. [Are any risks highlighted in the 10-K?]

• Operational and Financial Highlights: Use insights from the management section to address their needs with your products or services. [Can you identify any operational and financial highlights?]

• Market Segmentation and Geography: Focus efforts on the company's vital segments and regions. [What and where does the company focus from a business perspective?]

• Competitive Landscape: Use knowledge of competitors to emphasize your unique advantages. [Who are they company's biggest competitors?]

• Customer Relationships: Offer solutions that enhance customer satisfaction or retention. [What can you tell me about the company’s relationship with their customers as well as the sentiment of said customers?]

• M&A activity: Understanding a company’s acquisition activity in any period can shed light on areas of investment (ex. geography, technologies, product direction, etc.) as well as alert of opportunities related to transaction related dynamics (new teams to sell into, increase in headcount, consolidation of tools, ramping/onboarding new employees, etc.). [Wat can you tell me about any acquisitions and business combinations the company reported on?]

One of my favorite moves is to ask about risks highlighted in the 10-K and could potentially be addressed by the value prop of the software we build here at Endgame:

A few other fun things to note:

• While we simply leveraged GPT4 in this case, it’s also quite simple to build out a custom GPT with some pre-canned instructions that guide the responses. This is an easy way to speed things up if you’re doing this on the regular. Hit us up if you want to learn more about this technique!

• There are a number of custom GPTs that already exist that can analyze 10-Ks. They seem pretty solid—but sometimes come with some downstream costs for the underlying service.

• 10-K are provided to the SEC annually, and are made available via EDGAR.

• EDGAR’s API provides programmatic access to all filings (including 10-K) if you don’t feel like downloading each 10-K from a book of business individually for analysis.

GenAI truly does provide superpowers to reps that are willing to dive in. A 10-K is a great source, but keep in mind that the same approach can be taken for any relevant data (e.g., earnings call transcripts, CEO podcast transcripts, etc)—we’ll be touching on those soon.