Beyond the Hype: Real-World AI for Lean IR Departments
Dear Colleague,
Look, I understand. Your CEO just asked about “leveraging AI” after reading a think piece on the plane, but your entire IR budget barely covers a capital markets platform and annual report printing. Welcome to the club. The good news is you do not need a six-figure software contract to start using AI meaningfully in your day-to-day IR work. I have been testing free AI tools for the past 18 months, and here is what actually works when you are running IR on a tight budget. Just a disclaimer: be sure to use only public information and double-check any numbers these AI tools provide.
Implementing AI-driven sentiment analysis The reality check
Forget about fancy sentiment analysis dashboards. With free tools, you do this manually, but you do it more effectively. I spend about 30 minutes after each earnings call running sentiment checks on our recent communication, and it has caught issues before they became problems.
Here is what I actually do
After earnings calls, I copy the transcript or text into ChatGPT or Claude (or whichever you prefer). Here’s my go-to prompt flow:
Step 1: Do an initial sentiment scan
Prompt: "Analyze the sentiment of this earnings call transcript. Identify: 1. Overall tone (bullish/neutral/bearish) 2. Any defensive or uncertain language from management 3. Aggressive or skeptical questions from analysts 4. Topics that generated the most pushback [Paste transcript]" The AI typically highlights phrases I missed—like when our CFO said “we’re cautiously optimistic” three times in five minutes. That repetition signals anxiety to investors, even if the words sound positive.
Step 2: Do the competitive comparison
Prompt: "Here's our transcript. [Paste transcript]. Now compare it to this competitor transcript on the same topics: [paste competitor]. What are the key messaging differences? Where do we sound weaker or less confident?" Regarding positioning and other topics discussed in the sector, this is useful information. Recently, I noticed we were vague about pricing power, while our competitor was very specific. We adjusted our communication and, accordingly, our investor presentation.
Investor targeting Step 1: Investor targeting
For targeting, I use a different approach. I maintain a simple spreadsheet of our institutional holders and meeting notes. Once a quarter, I run this through AI:
Prompt: "I have notes from investor meetings. Help me segment these investors by concern themes: [Paste sanitized meeting notes - remove names/confidential info] Group by: 1) Growth investors focused on TAM (total addressable market) expansion, 2) Value investors concerned about margins, 3) ESG-focused, 4) Activists or governance-focused, 5) Short-term traders vs. long-term holders" The AI identifies patterns I might miss manually. Last time, it noticed that four different “growth” investors all asked about the timing of international expansion, indicating we needed clearer guidance on that topic.
Limitations Let us be real. Free tools have daily limits (Claude Free: about 30 exchanges; ChatGPT Free: rate-limited during peak hours). I reach these limits regularly. My workaround is to rotate between tools. For example, on Monday mornings I use ChatGPT, in the afternoon I switch to Claude if needed, and use Gemini as a backup. It is not elegant, but it works.
Automating earnings call transcription and Q&A preparation The transcription piece
Here is how I will save you money immediately: stop paying for transcription services for internal preparation. Our provider already transcribes our earnings calls for distribution, but for internal strategy sessions or management practice calls, I use free tools.
My workflow: 1. Record the practice session (Zoom free tier lets you record) 2. If under 25MB and you have ChatGPT Plus ($20/month—I pay this myself because it’s worth it), upload directly 3. For free versions: Use a free tool like Otter.ai (600 minutes/month free) or Transkriptor.ai (which also transcribes different languages) to transcribe first 4. Then paste into AI for analysis Q&A preparation is where this gets powerful Two weeks before earnings, I start building our Q&A prep document. Here’s my systematic approach:
Step 1: Compile analyst questions from last quarter
Prompt: "Here are analyst questions from our last earnings call. For each question, identify: 1. The underlying concern or information need 2. Whether our answer directly addressed it 3. Follow-up questions we should anticipate this quarter [Paste Q&A section]" Step 2: Predict new questions
Prompt: "Based on recent news about [supply chain disruptions/interest rate changes/competitor earnings], what questions should we expect from analysts about our business? Our company: [2-3 sentence description of business model and current priorities]" This is surprisingly accurate. Before our last call, Claude predicted six questions, and four variations actually came up.
Step 3: Stress-test management answers
When management drafts answers, I run them through:
Prompt: "Here's a potential answer to an analyst question about [topic]. Analyze: 1. Does this actually answer the question or dodge it? 2. What follow-up questions does this invite? 3. Rewrite this to be more direct and quantitative 4. Flag any phrases that sound defensive or evasive Answer: [paste draft answer]" I did this with our CEO’s explanation of a delayed investment launch. The AI flagged that he never actually provided a new timeline, which would have prompted analysts to ask for more details. We revised it to include specific milestones.
Step 4: Build a master Q&A doc
I create a document with three sections:
• Expected questions with prepared answers
• Tougher “red flag” questions with talking points
• Topics we absolutely cannot discuss (competitive/legal reasons)
Then I use AI one more time:
Prompt: "Review this Q&A prep document. Are there obvious questions missing? Are any answers inconsistent with each other? Flag any internal contradictions." Last time, it noticed that our guidance implied one growth rate, while our market share commentary implied another. It seems we updated one answer but not the other. That would have been embarrassing on the call.
Preparation for one-on-ones Before any investor meeting, I run this:
Prompt: "I'm meeting with [Investor Type: e.g., 'large-cap growth fund focused on tech']. Based on these notes from our last meeting [paste], our recent earnings results [summary], and current market conditions [brief context], what should I: 1. Proactively address 2. Be prepared to defend 3. Emphasize as positive developments 4. Avoid or handle carefully" This takes 2 minutes and makes me look incredibly prepared.
Balancing automation with personalized investor relationships Here’s the truth nobody tells you about AI in IR:
Investors can spot AI-generated content from a mile away. I learned this the hard way when I sent a "personalized" email that was 90% written by AI. The investor kindly called me out, saying, "This does not sound like you."
My rules for keeping it real:
1. AI is for drafting, not sending
I use AI to draft first versions of everything – emails, presentation sections, and Q&A responses. However, I always rewrite them in my own voice. My test is that if my CEO reads it and does not recognize my style, I revise it again.
Example workflow for investor emails:
Prompt: "Draft an email to a [long-term shareholder] following up after our earnings call. They expressed concern about [margin pressure] in our last meeting. Keep it professional but warm, 150 words max. Context: 1. Margins were down Q1 but we've implemented cost savings 2. We're seeing early positive results 3. Want to schedule a call to discuss in detail" Then I take that draft and rewrite it, adding specific details the AI does not know – such as referencing their firm’s recent research note or mentioning that we discussed this exact scenario last year. That specificity is what makes it feel personal.
2. Use AI for research, not relationship building
Before calls with new investors, I use AI to understand their portfolio and approach:
Prompt: "I'm preparing for a meeting with [Fund Name]. Based on publicly available information, they focus on [investment strategy]. What questions should I anticipate? What aspects of our business would align with their thesis? What would be red flags for them?" But during the actual meeting, put your phone down, take out your notebook, and give your full attention. I have seen colleagues scroll through AI-generated talking points during calls, and investors notice. They really do.
3. AI for volume tasks, humans for high-value interactions
I use AI heavily for:
• Scanning analyst reports for mentions of our company
• First-draft responses to routine questions
• Monitoring news flow for relevant topics
• Preparing routine disclosure documents
I never use AI for:
• Major investor communications during crises
• Sensitive board communications
• Personalized responses to concerned shareholders
• Relationship maintenance with top 10/20/50 holders
4. The “AI assist” approach to earnings prep
Our earnings preparation is a hybrid. I use AI to:
• Generate first draft of earnings press release
• Create multiple versions of guidance language
• Draft social media posts
• Prepare FAQ documents
However, our management team reviews everything multiple times, and I completely rewrite the shareholder letter myself. That document represents our voice, not an AI’s prediction of how we should sound.
What doesn’t work (learned the hard way): ❌ Using AI to write earnings guidance (too risky for accuracy) ❌ Letting AI interpret regulatory requirements (stick with lawyers) ❌ Automating investor email responses without review (killed three relationships before I learned) ❌ Using AI for anything involving confidential financial data on free platforms (huge security risk) ❌ Expecting AI to understand your company’s culture and voice without constant training
My final take Investor relations is still a relationship-driven business. AI is like having a highly intelligent intern who never sleeps – excellent for research, drafting, and pattern recognition, but you would not let an intern handle your most important investor calls alone.
The IROs who will succeed with AI are not those spending thousands on platforms, but those who identify where 15 minutes with ChatGPT or their preferred platform of choice can save 3 hours of work, then use those 3 hours to build real relationships with shareholders.
Start small. Choose one workflow from this article and try it for a month. If it works, add another. If your budget is limited, your time certainly is too. Use these tools to reclaim the time that truly matters: phone calls, relationship building, and strategic thinking – things no AI can do for you.
And when your CEO asks about AI again, you can honestly say you are already leveraging it. Just maybe do not mention that your entire tech stack costs exactly zero dollars.
Share your experiences and maybe even your favorite prompts!
Best, Muge
Your fellow IR Enthusiast!
Currently serving as the Director of Investor Relations and Sustainability at Galata Wind Enerji (GWIND.IS), Yücel brings a wealth of experience to the role, having begun her investor relations career in 2008 at Dogus Otomotiv (DOAS.IS). Her expertise in proactive strategies utilizing digital technology and AI, particularly in shareholder targeting, is instrumental in communicating Galata Wind's growth story. Traded on the Istanbul Stock Exchange, Galata Wind operates wind and solar farms in Turkey and is strategically expanding into Europe, targeting a capacity of over 1000 MW by 2030.
Yücel has recently published "The Investor Relations Playbook - Achieving Sustainable Success", a hands-on guidebook on investor relations operations with templates, checklists and how-to guides. The book is available in print in Turkish and in digital form in English.