Using AI to Analyze Earnings Calls: A Practical Workflow

Disclaimer: This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial advisor before making investment decisions. Past performance is not indicative of future results.

Most investors assume AI-powered earnings analysis requires an expensive subscription tool. It doesn’t. Here’s a complete workflow using any major AI assistant — for a fraction of the cost of a dedicated service. It becomes second nature after the first few runs.

What You’ll Need

• An AI assistant account — free tiers work for occasional use; paid plans (~$25–$30 CAD/mo) give you longer context windows and file uploads. Popular options include ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), and Copilot (Microsoft)
• The earnings call transcript (plain text or PDF)
• 10 minutes the first time you set this up; 5 minutes per earnings call after that

Getting the Transcript

For TSX-listed companies, transcripts are available through:

• Motley Fool Transcripts (free, usually same day)
• Seeking Alpha (free with limitations)
• SEDAR+ (for MD&A, not transcripts, but useful for official numbers)
• Your broker’s research section (TD WebBroker, Questrade IQ Edge both carry major transcripts)

Download as PDF or copy the text. Paid tiers on most AI platforms support PDF uploads directly.

The Prompt

This is the core of the workflow. Copy this and adapt it to your situation — it works with any AI assistant:

You are analyzing an earnings call transcript for [COMPANY NAME] ([TICKER]).
I am a long-term investor focused on [1-2 sentence description of your thesis 
for this company]. My holding period is [X years].

Please analyze this transcript and provide:

1. FINANCIAL SUMMARY
   - Revenue vs. expectations (if mentioned)
   - EPS vs. expectations
   - Key margin trends
   - Free cash flow commentary

2. THESIS CHECK
   - Does management commentary support or challenge my thesis?
   - Quote the 2-3 most relevant statements directly

3. RISKS MENTIONED
   - New risks introduced this quarter
   - How management addressed existing known risks

4. GUIDANCE
   - Exact guidance figures if provided
   - Any change in tone vs. last quarter

5. MY ACTION
   - Based purely on this call: Hold / Add / Reduce / Watch
   - One sentence rationale

Be direct. Flag anything that seems inconsistent or that I should investigate further.

[PASTE OR ATTACH TRANSCRIPT]

Why This Works Better Than Generic Summaries

The key difference from a generic AI summary is the thesis check. You’re asking the AI to evaluate the transcript against your specific reason for owning the stock, not just summarize what was said.

For example: if you own Suncor because you believe their refining margins will expand as they integrate acquisitions, you tell the AI that. It then specifically looks for management commentary on refining margins and acquisition integration — not just the top-line financial numbers everyone else is reading.

Example Output

Here’s a condensed version of what an AI assistant returned for a recent Canadian energy company earnings call:

THESIS CHECK: Partially supports. Management confirmed refining utilization improved to 94% (from 91%), consistent with your integration thesis. However, the CFO’s language on capital allocation shifted — “flexibility” was mentioned 4 times, which is new. This could signal a special dividend or share buyback is being considered, or it could mean capex uncertainty. Warrants follow-up.

MY ACTION: Hold. No new information that changes the long-term thesis, but monitor capital allocation language next quarter.

This kind of contextual output takes 3 minutes to read and tells you exactly what you need to know.

Cost

If you use a free-tier web interface and paste the transcript manually: $0.

If you use a paid plan (typically $25–$30 CAD/mo for ChatGPT Plus, Claude Pro, or Gemini Advanced): the earnings workflow is a small fraction of what the subscription already covers.

If you want to automate this via API (batch process multiple transcripts, store outputs): API pricing varies by provider and model tier, but processing an entire earnings season of 15–20 Canadian companies typically costs well under $1.50 CAD using a lightweight model like GPT-4o mini, Claude Haiku, or Gemini Flash.

Choosing an AI Provider

All the major providers handle long-form document analysis well. A few practical differences:

Try the prompt on the free tier first — most transcripts fit within free context limits when pasted as text.

Limitations

• AI knowledge cutoffs mean the model doesn’t know how the stock has traded since the call — you provide that context
• Models can miss subtle tone shifts that an experienced analyst would catch on a live call
• For very technical industries (biotech, mining), validate the AI’s interpretation of technical language against the actual transcript

Bottom Line

This workflow is particularly powerful for investors who own 10–20 positions and want to track each quarterly call without spending 10+ hours per reporting season on transcripts. The upfront cost is writing a good prompt once. After that it’s mostly copy-paste — and it works with whichever AI tool you already use.

True North Alpha publishes educational content for Canadian self-directed investors. Nothing in this article constitutes personalized financial advice or a solicitation to buy or sell any security. Always do your own due diligence and consult a registered financial advisor before making investment decisions. TFSA and RRSP contribution rules and eligibility are subject to CRA guidelines — confirm your personal situation with a tax professional.