Accounting for the AI Moat - Key Takeaways from Our Webinar with Digits CEO

Our discussion with Digits reinforced our Moat & Journey framework around software's AI positioning. With LLM one-shot accounting accuracy plateauing, deep domain expertise, proprietary data, and workflow ownership likely prove critical to sustaining software moats and monetizing AI. We hosted Jeff Seibert, Co-Founder & CEO of Digits, an AI-native accounting platform, for Episode #4 of our 'AI in Practice' webinar series. Key takeaways: (1) Domain-specific AI continues to outperform frontier LLMs - Digits' latest study found standalone LLM accuracy in accounting classification has plateaued at ~80%, agentic models improved to ~89%, still trailing Digits' purpose-built architecture at ~98%, attributed to its proprietary "layer cake" architecture of deterministic, accounting-specific ML models built over six years and embedding each business's historical accounting data and conventions. (2) Systems of record may capture more value than the underlying models, as foundation models commodit