AI IPO Drafting Accuracy Hits 98-99%, Says Perplexity Labs

AI is Nailing IPO Drafts—But Here’s Why Humans Still Matter

Let’s be honest—IPO prospectuses are about as exciting as watching paint dry. But here’s the thing: they’re the golden ticket for companies going public. And until recently, drafting one meant weeks of sleepless nights for lawyers and bankers. Now? AI’s claiming it can get 98-99% of it right. That’s according to Aravind Srinivas from Perplexity Labs, who says accuracy has jumped from Goldman Sachs CEO David Solomon’s earlier estimate of 95%. That’s not just a small upgrade—it’s a total game-changer for Wall Street.

How AI is Shaking Up Financial Docs

Picture this: a team of exhausted lawyers, drowning in coffee and spreadsheets, trying to meet some insane deadline. That’s how IPO drafts used to happen. But AI—especially tools using natural language processing—is turning the whole process upside down. Companies like Perplexity Labs and Goldman Sachs are at the front of this shift, automating what used to be a manual nightmare. And honestly? It’s about time.

From 95% to Near-Perfect: What Changed?

Goldman’s Solomon first said AI could handle 95% of the work, cutting a six-person team’s workload from weeks to minutes. Now, Srinivas says better training data and context understanding pushed that to 98-99%. Think of it like going from an A- to straight A’s in a class where one typo could land you in regulatory trouble. The kicker? Early users say AI drafts only need minor tweaks—mostly for legal fine-tuning or executive summaries.

How Does AI Pull This Off?

Simple: it learns from tons of past filings, SEC rules, and prospectuses. The AI spots patterns, structures the doc correctly, and NLP turns numbers into something readable. But here’s the catch—humans still handle that last 1-2%. It’s like a dance: AI does the heavy lifting, humans add finesse, and the system gets smarter each time.

Why This is a Big Deal

• Speed: Weeks of work? Try minutes.
• Cost: Fewer lawyers billing insane hours.
• Consistency: No more fighting over formatting.
• Scale: Handle multiple IPOs without breaking a sweat.

But It’s Not Perfect

Look, AI still stumbles sometimes. Regulators might ask how bot-generated docs meet legal standards. And if the training data’s biased? That could cause problems. Then there’s the ethics angle—should investors know if their prospectus came from a machine? (My take: absolutely.)

Where This is Headed

Accuracy will likely hit 99.9% soon, with AI tackling annual reports, M&A docs, even earnings calls. Blockchain could make everything more transparent. And junior bankers? They’ll probably shift to overseeing AI instead of grinding through drafts. One thing’s clear—those all-nighters over prospectuses? They’re on borrowed time.

The Bottom Line

AI hitting 98-99% accuracy is huge—no question. But humans still bring the nuance and compliance know-how. As Srinivas and Solomon suggest, the future’s not about picking sides. It’s about AI and humans working together better than ever.