Startup Stories - Mixergy #2288 She had people do AI’s work
Dec 17, 2025
Helen Hastings, founder and CEO of Quanta, a cutting-edge accounting platform, shares her journey from engineering at Affirm and Google to revolutionizing bookkeeping. She discusses how she initially had humans perform tasks that AI later tackled. Helen reveals insights from shadowing bookkeepers and highlights the pain points of current systems. Quanta builds a new ledger from scratch, emphasizing real-time data and leveraging a hybrid model. She also unveils Prism, a feature that provides natural-language reports, transforming accounting for software companies.
AI Snips
Chapters
Transcript
Episode notes
Humans Taught The AI To Read Memos
- Early on Quanta had humans read memos and receipts to understand context like AWS vs Amazon purchases.
- Those same language-understanding tasks later migrated to LLMs as models improved.
Shadow Real Users Early
- Do extensive user research by shadowing bookkeepers and finance teams to learn real workflows and edge cases.
- Use LinkedIn warm intros and ask each contact who to talk to next to compound meetings quickly.
Manual Data Moves Are The Big Bottleneck
- Much bookkeeping pain comes from manual data import, CSVs and logging into many tools.
- Automating direct integrations and augmenting bank data removes the largest time sink.