Latest / Tech IPO Conversations with Fexingo: Public Software Companies and Capital Markets / How Post-IPO Companies Are Using AI to Write Their Earnings Scripts
Transcript
- Lucas: So we are three weeks away from peak earnings season for Q2 2026, and I’ve been noticing something that’s quietly reshaping how post ipo companies talk to the Street. More and of them are using generative AI to write their earnings scripts. Luna: Wait — you mean the actual words the CEO reads on the call? Not just the charts and numbers? Lucas: Exactly. The narrative portion. I’ve been looking at transcripts from the last few quarters, and there’s a pattern. Some calls have this peculiar uniformity in phrasing — almost too polished, too consistent. One analyst I talked to called it ‘ai ese.’ Luna: Interesting because Roblox, ticker R-B-L-X, just jumped 15 percent in the last five days. Do we know if they’re using AI for their scripts? Lucas: So Roblox hasn’t confirmed it, but their last two earnings calls had unusually symmetrical sentence structures. I’m not saying it’s a smoking gun, but compared to their earlier, more freewheeling calls, it’s a clear shift. And they’re not alone. I’ve heard that at least three of the companies in the ARK Innovation ETF are experimenting with ai written sections. Luna: That makes sense from a risk-management perspective. If you’re a post ipo company, every word you say on an earnings call can move your stock. Having AI draft it might reduce the chance of a verbal misstep. Lucas: Right, and there’s a specific case that really highlights the trade-off. NVIDIA, ticker symbol N-V-D-A, which is down about 2.5 percent over the past five days, has been using AI to generate their prepared remarks for the last two quarters. I’ve confirmed this with someone on their IR team. Luna: Wow. So the company that makes the chips for AI is using AI to talk about its own earnings. That’s… meta. Lucas: It is. And the interesting thing is, when they first did it, analysts barely noticed. But by the second quarter, a few pointed out that the language lacked the typical Jensen Huang spontaneity. One said it felt like reading a Wikipedia entry about NVIDIA rather than hearing from the CEO. Luna: So investors actually prefer the human version, even if it’s less polished? Lucas: That’s the critical question. I think it depends on the audience. Retail investors might appreciate the clarity and consistency. But sell-side analysts — they want the off-script moments. They want the CEO to say something that isn’t in the press release. Luna: Right, because that’s where the real signal is. A scripted call can be gamed. But an unscripted answer to a tough question — that reveals conviction. Lucas: Exactly. And there’s data backing that up. A study from last year by an academic at Stanford looked at earnings call transcripts over a decade. They found that calls with more spontaneous language — more verbal disfluencies like ‘um’ and ‘uh’ — actually had stronger stock performance in the following quarter. Luna: Because it signals authenticity. The market can tell when you’re reading a script. Lucas: Precisely. So the post ipo companies that are turning to AI for scriptwriting might be solving a short-term risk but creating a long-term trust problem. And I think this is especially acute for companies that went public in the last three years — they’re still building their credibility with investors. Luna: So what’s the alternative? Keep the AI but use it differently? Lucas: That’s what some companies are doing. Palantir, ticker P-L-T-R, which is down about 1 percent in the last five days, has been using AI to generate bullet points and data summaries, but the narrative flow is still written by humans. They’re essentially using AI as a research assistant, not a replacement. Luna: That seems like the sensible middle ground. You get the efficiency, but you keep the human voice. Lucas: And I think that’s where the industry is heading. Because the SEC is starting to pay attention. I’ve heard that the Division of Corporation Finance is considering whether ai generated earnings commentary needs additional disclosure. If the Street can’t tell if a statement is from the CEO or a language model, that’s a material issue. Luna: That would be a fascinating regulatory development. Imagine having to say, ‘Portions of this call were generated by artificial intelligence.’ Lucas: Exactly. And that could really change the calculus for these companies. Because right now, the main motivator is speed and cost reduction. You can draft a full earnings script in minutes instead of hours. But if you have to label it, the efficiency gain might not be worth the perception hit. Luna: So for a post ipo company, what’s the right move today? Lucas: I think the right move is to use AI for the data-heavy sections — the financial highlights, the operational metrics — but keep the strategic narrative human-written. And definitely do not let the AI answer analyst questions. That’s where the real damage can happen. Luna: Have any companies actually tried ai generated Q&A? Lucas: Not that I’m aware of, but I wouldn’t be surprised if someone experiments with it in the next year. The technology is certainly capable. But the risks would be enormous. Imagine the AI generating a specific forward-looking statement that the company can’t back up. Luna: Yeah, that’s a lawsuit waiting to happen. And it would completely undermine the purpose of the call — to build trust. Lucas: Right. So the real question is not whether AI can write an earnings script — it clearly can. The question is whether that script serves the long-term relationship between the company and its investors. And I think the early evidence suggests that a fully ai generated script is a net negative. Luna: So what do you think the landscape looks like three years from now? Lucas: I think we’ll see a hybrid model become the standard. AI handles the boilerplate — the safe, factual stuff. Humans handle the narrative, the tone, the strategic messaging. And companies that try to go full AI will face a backlash from analysts and institutional investors. Luna: That sounds a lot like the early days of robo-advisors in wealth management. Everyone thought algorithms would replace human advisors, but instead they became tools that made human advisors better. Lucas: Great analogy. And I think that’s where we’re headed. By the way, speaking of things that make our work better — if you’re getting value from these conversations, we do have a small way you can support the show. It’s buy me a coffee dot com slash fexingo. Luna: Yeah, it’s a way to keep this ad-free and focused on exactly the stuff we just talked about — no sponsors, no interruptions. Lucas: Exactly. And we really appreciate it. So back to the AI scripts — I think the next frontier is the earnings call itself becoming a two-way ai human interaction. Luna: You mean the analysts using AI to ask questions and the company using AI to answer? That would be bizarre. Lucas: It sounds dystopian, but some hedge funds are already using AI to generate questions in real time based on the prepared remarks. So we might end up with AI asking AI questions, and no human actually talking on the call. Luna: And then the SEC would have to step in. At that point, the entire purpose of the earnings call — to provide a genuine exchange of information — is lost. Lucas: Exactly. So for now, the smart post ipo companies are treating AI as a co-pilot, not an autopilot. They get the efficiency, but they keep the human judgment. And that’s the lesson I think will stick. Luna: Well said. And a good reminder that as much as technology changes, investors still want to hear from a person they can trust. Lucas: Absolutely. That’s the core of it. So next earnings season, listen for the ‘ums’ and ‘uhs.’ They might be the most valuable signals on the call.