Welcome back to the Digital CPA Podcast, brought to you by CPA.com, the business and technology arm of the AICPA. We serve as the sounding board for the DCPA community, leveraging technology to transform practices. I’m Kacee Johnson, Vice President of Strategy and Innovation. Unfortunately, my co-host, Kalil Merhib, is on the road and unable to join us today. However, I’m excited to be joined by a true innovator and technologist, Jeff Seibert, founder and CEO of Digits. Welcome, Jeff!
Kacee, it’s great to be here! Thanks for having me. It’s great to have you on. I’m a big fan of Digits, and I’m excited for this conversation because I truly think you have one of the most innovative solutions in the space. Much like your earlier ventures, Digits is incredibly forward-thinking. What lessons from building Crashlytics and your time at Twitter have been most relevant as you grow Digits?
Great question. Back in 2011, I was in a completely different industry—the mobile developer tool space. We built Crashlytics to help app developers pinpoint within seconds exactly which line of code was causing a crash. We ended up being acquired by Twitter, and at that time, there were billions of smartphones worldwide. Essentially, Crashlytics was embedded in almost every active smartphone, helping developers monitor app crashes in real time. Later, at Twitter, I was leading product in 2015, and one key lesson I took from both experiences was the power of real-time data.
With Crashlytics, speed was everything—within two seconds of an app crash anywhere in the world, developers received a notification so they could diagnose and fix the issue. Twitter, of course, is famous for real-time engagement. We had internal dashboards tracking live trends—how many people were on the platform, what they were discussing, and the overall global zeitgeist. Then I looked at the financial side of the business, and it was the complete opposite. At Twitter, I had to wait two to three weeks to get our monthly financials. I once asked corporate finance for our budget projections, and they told me it would take three weeks to run the numbers! There were over a hundred people in corporate finance, yet real-time insights didn’t exist.
That was the lightbulb moment for me. The accounting profession has traditionally delivered perfect data—but weeks later. There wasn’t a solution for providing even directionally correct financial insights in real time. That’s why we launched Digits in 2018—our mission is to make small business finance real-time using modern technology.
Real-time data is critical in our profession, but so is accuracy. I saw on LinkedIn that Digits recently hit a 100% accuracy milestone on a particular aspect of its functionality. You and I have talked about how vendors typically guarantee accuracy in the 75–90% range. Did you anticipate your models reaching this level of precision so quickly?
Honestly, we’ve been really disciplined in our software engineering approach, and the progress has been fascinating. A couple of years ago, our model was only 50% accurate—it was right as often as it was wrong! Since then, we’ve built an intelligent system that continuously learns and improves. I had initially believed we’d cap out at 90–95% accuracy, given the complexities of accounting. But about a month ago, one of our models achieved 100% accuracy for a particular business for an entire month. Now, it’s not yet repeatable—it was a simpler business—but it opened my eyes to the possibility that 100% accuracy is achievable under the right conditions.
Broadly speaking, I think we can reliably hit the 90–95% range across the industry. That rapid evolution is amazing. Anytime we discuss AI on the AICPA Town Hall, where we get over 10,000 attendees, most practitioner questions revolve around data security, safety, and responsible AI usage. There’s also ongoing legislative attention in Washington, D.C. How does Digits approach responsible AI and data privacy?
That’s a fantastic question. We take data security and privacy extremely seriously. At Crashlytics, we had insights into every app on every device, and at Twitter, we had access to massive amounts of real-time information. When I left Twitter, I actually called for regulation of social media because of the sheer volume of data being collected. With AI, what concerns me most is the data used for training models. Unlike open-source software—where you can inspect the code—AI models are just billions of numbers. Even when companies claim their models are “open,” like Meta’s Llama, there’s no way for humans to truly understand how they operate. That lack of transparency is a major issue, especially in a trust-based profession like accounting.
So, I believe the key to responsible AI implementation is clear disclosure—firms should know what data is being used, where it’s going, and how models are making decisions. Transparency will be a huge factor in adoption and trust. Thanks for sharing your insights, Jeff. I know our listeners will find this incredibly valuable. We appreciate you joining us today! It was my pleasure, Kacee. Thanks for having me on!
And thank you to our audience for tuning in. We’ve launched a Generative AI Research Initiative with tons of free resources available at CPA.com/gen-ai. Let us know what topics you’d like us to cover next—email us through the feedback link in our show notes. Until next time, keep innovating and driving change!