Reading earnings with AI: how to understand quarterly results in minutes

Why earnings are hard to read

A quarterly release can run to dozens of pages. Then there are the investor slides and the conference call transcript, where management answers analysts' questions. Reading everything takes hours, and almost nobody does it.

The result is that many people stop at the headline. But a company can beat profit estimates and still fall because it lowered its outlook for the next quarters. Or miss estimates and rise because margins are improving. What matters is often in the details.

This is where AI is genuinely useful: reading and summarising long texts is one of the things it does best.

The golden rule: give AI the document, do not ask it for the numbers

If you ask an AI assistant "how did this company's quarter go?", you risk getting old or made-up numbers. Assistants do not always have up-to-date data and sometimes get figures and dates wrong.

The fix is simple: download the official release from the Investor Relations section of the company's website (or the call transcript) and paste or upload it into the chat. That way AI works on the real text and you can check every number.

Step 1: the 10 line summary

Prompt to copy:

I am uploading the official quarterly results release of [company].
Summarise it in at most 10 lines: revenue, profit, margins, outlook for the next periods (guidance) and the 2-3 most important facts.
For every number, tell me the page or sentence it comes from. Use only data from the document, do not add numbers that are not there.

Step 2: what changed compared to before

A number on its own says little. What matters is the comparison with the same quarter last year and with what the company had promised.

Prompt to copy:

Still from the document I gave you:
1. How did revenue, margins and profit change compared to the same quarter last year?
2. Was the outlook (guidance) raised, confirmed or lowered?
3. Which segments or regions did better and which did worse?
If the document does not contain a piece of information, write "not present" instead of estimating it.

Step 3: what management is really saying

On the conference call, management chooses its words carefully. AI can help you notice changes in tone and recurring themes in analysts' questions.

Prompt to copy:

I am uploading the conference call transcript of [company].
1. What are the 3 topics analysts asked about most?
2. Are there vague or evasive answers? Quote the exact sentence.
3. Which risks or difficulties did management mention, even in passing?
Always quote the original sentences, do not rephrase them.

Step 4: devil's advocate

As with any trade idea, the last step is to look for the weak spots in your own reading.

Prompt to copy:

After reading these results my impression is: [write your impression].
Play devil's advocate: which elements of the document go against my impression? What should I check before drawing conclusions?
Do not tell me whether to buy or sell.

Mistakes to avoid

• Asking AI for numbers from memory instead of giving it the document.
• Stopping at profit and revenue without looking at the outlook.
• Trusting a number without finding it in the original document.
• Turning an earnings read into a buy or sell signal: it is information, not a decision.

For a complete method to follow before every trade, read How I prepare a trade with AI. To keep track of what you learn, read The trading journal with AI.

This article is for educational and informational purposes only and does not constitute financial advice. Trading involves a high risk of losing capital.