Contrary to popular belief, AI isn’t bad with numbers—it’s actually faster and more reliable than humans in many situations, especially when equipped with the right guardrails.

At Airkey, my cofounder (ex-PE & IB) and I (ex-Amazon) have built an automated workflow for business brokers & search funders to parse tax forms in seconds (1065, 1120, 1120-S, and Schedule C###-###-#### If you're curious about how we validate numbers and ensure accuracy, feel free to check out our more detailed blog post.

That said, let’s be realistic—AI isn’t crafting LBO models from scratch just yet. And if you’re simply dropping data into ChatGPT, the results might not be what you’re hoping for. But, as an engineer, I have a lot of tools at my disposal to improve the quality of outputs, such as orchestrating multiple chats, utilizing open-source libraries, and running deterministic algorithms.

I wanted to share this post to challenge some of the skepticism around AI in finance and start a conversation. I’d love to answer questions and share insights from the research I’ve done and the tools I’ve built.