When I worked as an equity research analyst on Wall Street, I wrote my earnings reports before the earnings calls even happened. I think that would surprise some investors.
I didn’t know the final numbers, obviously, but the investment thesis, commentary, and general conclusions were usually inferable in advance. After the earnings report, I’d update my financial model and add color to anything notable before publishing.
Sometimes I’d spend less than an hour editing a report that looked like it had taken days to write.
Other analysts I worked with did the same thing. Our institutional clients expected research within hours of a company’s earnings announcement, and there wasn’t time to write a 15-page report from scratch.
But there was another reason this worked. After covering the same companies for several quarters, I generally knew what management was going to say. The earnings calls rarely changed my investment thesis. Most of the research had already happened.
Now that I use AI for my own investment research, I’ve been thinking about how much of that process individual investors can replicate. Particularly the preparation that happens before a company reports earnings.
What actually happens on an earnings call
If you’ve ever listened to an earnings call, you know how formulaic they can be. Management discusses quarterly results, highlights the areas performing well, acknowledges a few challenges, and reassures investors about the long-term outlook. It’s pretty predictable.
That’s because earnings calls are carefully prepared. Management has worked with investor relations and legal teams to decide what information to disclose and how to present it.
One thing I learned was to pay attention to the metrics management stopped discussing. For example, if a company had spent the previous four quarters highlighting the growth of a particular product and suddenly stopped disclosing the number, I’d want to know why. That omission might matter more than anything management chose to emphasize.
The Q&A is more useful than the presentation
The most interesting part of an earnings call is usually the Q&A, when analysts have an opportunity to challenge management’s assumptions.
A 2011 study published in The Accounting Review examined more than 10,000 earnings call transcripts. Researchers found that both prepared remarks and Q&A sessions conveyed information beyond what was already in the earnings release, but the Q&A was generally more informative.
That’s consistent with my experience. During prepared remarks, management gets to choose the topics. During Q&A, analysts can raise issues management might prefer not to discuss.
A 2026 study in the Review of Accounting Studies examined earnings calls from 2002 through 2017 and documented how executives redirect answers away from analysts’ questions.
Interestingly, this behavior wasn’t necessarily an indication of bad news. Management was more likely to change topics when business performance was strong, among other circumstances. So a non-answer isn’t necessarily a red flag, but it’s certainly something to pay attention to.
For example, imagine an analyst asks whether a company’s largest customers are reducing orders. Management responds by discussing the company’s strong pipeline of new business opportunities. That’s the kind of thing I’d flag for further research.
What our institutional clients actually paid for
I covered healthcare companies, primarily in diagnostics, laboratory services, and genomics.
Our clients were institutional investors. They had their own financial models, analysts, and investment opinions. They weren’t interested in paying us to summarize quarterly results they could read themselves. They wanted proprietary research that might help them anticipate business performance before everyone else. That led to some strange assignments.
I spent days calling STD clinics across the country. One of the companies I covered manufactured molecular diagnostic tests. I called clinics to understand testing volumes, available tests, and which diagnostic equipment they used. We used this information to develop independent estimates of demand.
I got my blood drawn three times in one day. I covered two of the largest laboratory testing companies in the country. A Silicon Valley startup was promising to revolutionize blood testing, and investors were concerned about the competitive threat. I visited three laboratories in one afternoon, documenting differences in the test results. (That startup was Theranos.)
I crawled underneath a DNA sequencing machine to read its serial number. We wanted to estimate how many sequencing machines a company had sold. During a laboratory visit, I got down on the floor to find the serial number on a new machine, hoping to estimate the company’s sales that quarter.
This is what some of the research behind those polished Wall Street reports actually looked like. A junior analyst crawling around a laboratory trying to read a serial number so a hedge fund could refine its revenue estimates.
The point of all this work was to develop an independent view of business performance. If we could estimate a company’s quarterly revenue more accurately than other investors, that information might be valuable. But even that wasn’t enough.
We needed to understand whether our expectations differed from those already reflected in the stock price.
Good earnings don’t necessarily mean a good investment
This is one of the most important things for individual investors to understand.
A company can report an excellent quarter and still see its stock decline.
Suppose a company reports 20% revenue growth and beats Wall Street’s earnings estimates. But what if investors were expecting 30% revenue growth? Or the stock is trading at a valuation that assumes the company can maintain 30% growth for the next five years?
Suddenly, the results don’t look so impressive.
The opposite can also happen. A company can report declining earnings and see its stock rise because investors had expected an even worse result.
Analyst consensus estimates are a useful reference point, but they don’t necessarily capture everything investors have priced in. Valuation matters, too.
This is why I don’t think investors should pay too much attention to the initial market reaction after an earnings announcement.
The questions I’d rather answer are whether the company’s long-term prospects have improved or deteriorated, and whether the current valuation reflects those changes.
How I’d use AI to analyze earnings today
AI has made much of the research preparation I used to do on Wall Street available to individual investors.
You don’t need access to an expensive Bloomberg Terminal to review historical financial results. Companies publish their quarterly and annual reports through the SEC’s free EDGAR database.
Most also make earnings releases, presentations, and call recordings available on their investor relations websites.
Here’s the process I’d use.
Before earnings: establish your expectations.
I’d start by collecting the company’s recent financial filings, earnings releases, and several quarters of call transcripts.
Then I’d have AI compare management’s previous guidance with actual results and identify the operating metrics that matter most.
For example, I’d ask:
Review the last four quarters of this company’s earnings materials. Identify the three to five operating metrics most important to its long-term value. Track how those metrics have changed, compare management’s guidance with actual results, and identify any metrics the company has stopped disclosing. Cite the original documents for every finding.
This gives you a starting point before management presents the latest quarter.
I’d also establish what I expect to happen and what evidence would make me reconsider my investment thesis.
Otherwise, it’s too easy to rationalize disappointing results after the fact.
After earnings: focus on what changed.
Once the company reports, I’d upload the new earnings release and call transcript alongside the prior-quarter materials.
Rather than asking AI for a general summary, I’d ask it to compare the new information with our previous expectations.
Did revenue growth accelerate or decelerate? Did management change its guidance? Are margins improving for sustainable reasons? Has management’s explanation of its competitive position changed?
I’d pay particular attention to how management answered questions about the business’s biggest risks.
AI is useful here because it can compare several quarters of commentary without having to manually search through each transcript.
You can ask it to identify every instance where management discussed a particular competitor, customer group, product, or financial metric.
You can also ask it to find questions that management didn’t directly answer.
I would verify the underlying passages myself. A model might correctly identify a change in language while incorrectly interpreting its significance.
Finally: update the investment thesis, not just the quarterly forecast.
Suppose a company misses quarterly revenue expectations by 3%.
If the reason is a delayed customer order that will be recognized next quarter, the long-term impact could be negligible.
If the reason is a competitor introducing a superior product, that same 3% miss could be an early indication of something much more serious.
The financial result is similar. The investment implications are completely different.
This is where I think an investor’s own judgment matters most.
AI can help identify and investigate the differences. But I still want to decide for myself whether the new evidence changes the company’s long-term outlook.
You don’t have to compete with Wall Street every quarter
One thing I appreciate more now that I’m investing my own money is that I don’t have to predict next quarter’s results.
When I was in equity research, anticipating quarterly performance was an important part of the job. Institutional clients wanted information that could help them position ahead of earnings.
As an individual investor, I have a different objective. I’m looking for businesses that can become materially more valuable over the next five years. I don’t necessarily need to know whether they’ll beat earnings estimates next Tuesday.
I do need to know whether their competitive advantages are holding up, whether the growth opportunity is still attractive, and whether I’m paying a reasonable price.
Quarterly earnings are useful checkpoints to test whether the business is progressing as expected. They aren’t necessarily reasons to buy or sell.
AI makes those checkpoints much easier to manage. I can follow more companies, review more historical information, and investigate changes in their businesses without spending days reading transcripts.
But the preparation is still important. When I worked on Wall Street, I could write most of my earnings report in advance because I already understood the business and knew what I was looking for.
I think individual investors should approach earnings the same way. Know what matters before the company reports. Use AI to find what changed. Then decide whether the change actually matters to your investment.