One of the most popular sections of my book, AI in Practice, is the part where I explain how I used AI each day. (The book is divided into 30 practice sessions, one for each day. Each day contains a practical example of how I used AI.)
One of the most common questions I get from people wanting to learn AI is “What can I do with it?” Giving these examples helps spark ideas.
While this isn’t strictly about how I use AI for writing, I think it’s still relevant. Anything that AI does to free up time to write is helpful. So learning how to use it to simplify daily tasks can be the best way to use AI for writing.
I got the idea for today’s post in a roundabout way. Today’s post by Paul Krugman linked to a post by Mike Konczal, in which he explained how terminal AIs were saving him hours per day of research and coding. He used this time to dive deeper into the numbers and improve the quality of the research.
I thought this paralleled many of the ways I use AI—learning, researching, and writing. It saves me time by doing low-value work, so that I can focus on higher-value work that ultimately creates a better product.
Tech stocks and AI are driving the stock market right now. The S&P 500 is a weighted index, which means some stocks carry far more influence than others. The Mag 7 tech stocks (AAPL, NVDA, MSFT, GOOGL, AMZN, META, TSLA) represent roughly a third of the entire index, and NVDA alone accounts for about 7.5%. So when NVDA announces earnings, it’s a big deal. Where NVDA goes, the market follows.
NVDA announced earnings yesterday afternoon (February 25, 2026). Like all good executives, CEO Jensen Huang and CFO Colette Kress painted a rosy picture of their financials—and parts of it were certainly impressive. Revenue up 73% year-over-year is a big move for any company, let alone one already valued at nearly $5 trillion. But I wanted to know: was the reality as good as the hype?
Companies are required to file a report with the SEC—either a 10-Q for quarterly filings or a 10-K for annual ones. Since this is the end of NVDA’s fiscal year, they filed a 10-K. That’s the unvarnished version of their finances, and it doesn’t always tell the same story as the polished earnings call. Transcripts and press releases are easily found online. Here are a few links:
I downloaded the PDFs and pulled them into Claude, which has become my go-to AI. I have a special Claude instance that I’ve set up for financial analysis. If you want to try this yourself, you might start by telling your AI that it’s an “expert financial analyst, specializing in tech stocks, AI, and data centers.” That gives it context and sharpens the analysis.
After dropping the files in, I prompted:
Do an analysis of the transcript and press release from the NVDA earnings call of February 25, 2026, and compare it to the corresponding 10-K filing (in the files provided). Does the earnings call accurately reflect their true financial picture? Are there hidden risks that weren’t discussed on the call? What are the implications for the overall market?
Within minutes, Claude had done what would have taken me hours. It provided a thorough comparison of the earnings call narrative against the actual filings, flagged areas where the call’s optimistic tone didn’t quite match the risk disclosures buried in the 10-K, and identified potential red flags worth watching in the coming quarters. It even laid out the broader market implications—what NVDA’s results might signal for the AI supply chain, hyperscaler spending, and competing chipmakers.
I’m not a financial advisor, so I won’t go into the specific findings here. But this is a simple exercise you can do on your own. And if you don’t fully understand the output, just prompt:
I don’t understand. Explain it to me like I’m twelve.
That’s the real power of this. You don’t need to be a financial analyst to do financial analysis. AI handles the heavy lifting—reading hundreds of pages, cross-referencing numbers, spotting inconsistencies—so you can focus on the part that matters: making better decisions with better information.
Think about what I didn’t need to do this. I didn’t need a finance degree. I didn’t need a Bloomberg terminal. I didn’t need to spend hours reading through hundreds of pages of SEC filings, cross-referencing numbers between documents, or deciphering accounting jargon. I just needed the documents and a well-crafted prompt.
This is what Konczal was getting at in his post. He described how terminal AI tools compress what he called “the setup and robustness-checking phase of knowledge work.” The hours he used to spend wrangling data and doing first-pass analysis now take minutes, which frees him to focus on the part that actually requires human judgment—figuring out what’s really going on in the numbers. As he put it, the human attention that used to go to plumbing can now be repurposed to asking, “What is actually going on here?”
My NVDA analysis was the same idea applied to a different domain. The AI did the tedious work of reading, comparing, and cross-referencing. I did the thinking about what it all meant.
And here’s the thing that connects this back to writing: every hour I save on tasks like this is an hour I can spend writing. That’s the real leverage. AI doesn’t just help with the thing you’re asking it to do—it gives you back time for everything else.
If you want to try this kind of analysis on your own, here’s a quick framework:
Find the source documents. For any publicly traded company, the SEC filings (10-K or 10-Q) are available at sec.gov. Earnings call transcripts are widely available on sites like Seeking Alpha and The Motley Fool.
Set the context. Tell your AI what role to play. “You are an expert financial analyst specializing in [industry]” goes a long way.
Ask a specific question. Don’t just say “analyze this.” Ask it to compare the earnings call to the filing, look for discrepancies, or identify risks that weren’t discussed.
Follow up. If something in the output doesn’t make sense, ask for clarification. “Explain it like I’m twelve” is a perfectly valid prompt.
You don’t need to be an expert to get expert-level analysis. You just need to know what questions to ask—and that’s a skill that gets better with practice.
Which, come to think of it, is the whole point of the book.
As a thank-you to my paid subscribers, I’ve created a ready-to-use Claude Project template available behind the paywall below. You can download and import. It includes:
A system prompt that configures Claude as an expert financial analyst, with instructions for how to approach earnings calls, SEC filings, and press releases
A set of starter prompts you can use out of the box or customize for any publicly traded company — not just NVDA
Follow-up prompts for digging deeper, including risk analysis, competitive positioning, and plain-language explanations
Just download the file below, and follow the instructions inside to set it up in Claude. Once it’s configured, all you need to do is drop in the documents and start asking questions.
Disclaimer: I am not a financial advisor, and nothing in this post should be construed as financial advice. The analysis described here is for educational and informational purposes only. Always do your own research and consult a qualified financial professional before making any investment decisions.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.