Lessons from an Early AI x Fintech Pioneer | Doma - Andy Mahdavi

Foundation Capital hosted our first Fintech x AI PortCo Summit in June 2023 to help executives across the Foundation Capital Portfolio answer the question “What should I be doing about AI?”
Founders and leaders from over 35 Fintech companies convened to share their learnings and gather perspectives on this essential question.
Andy, CTO of Doma, reflected on the knowledge he’s gained from over six years of leading the startup’s data science and ML efforts. Doma is reinventing real estate transactions by dramatically improving the title and closing process. This complex workflow requires analyzing large amounts of unstructured data. Well before the advent of generative AI, Doma met this challenge by developing proprietary ML models, starting with Bert-based precursors to GPT.
Doma’s primary advantage lies in its unique data assets. These include millions of historical transactions, over 100 public data sources, and centralized operations for human-in-the-loop exception handling. Fine-tuning of models on business-specific data in this “last mile” is essential for financial use cases, which demand extremely high precision. Given the large amounts of sensitive financial data and money involved, there’s no room for mistakes in this industry.
Andy categorized the learnings of Doma’s ML team into two areas: risk/predictive modeling and automation:
1. Risk/Predictive Modeling: Andy emphasized that regular engagement with regulators is important to secure approval for innovative techniques. He also warned against depending on data managed by legacy competitors.
2. Automation: Keeping current with the latest ML research through regular journal reviews has been instrumental for the team. However, Andy cautioned that introducing partial automation into products can create unanticipated challenges, especially if the handoffs between automated and manual processes are not clearly defined. He also noted that forming internal ML teams can lead to significantly higher cloud and engineering costs compared to offshore outsourcing. While in-house teams may deliver better results in the long run, founders should carefully evaluate this trade-off.