Summary
Daniel Kim and John Glasgow argued that legacy ERPs block autonomous finance because their data models were built for static monthly human workflows, not machine reasoning. They emphasized “propose, don’t post,” reconstructable evidence lineage, continuous close, and AI-readable finance data as prerequisites for useful finance automation.
Key Topics
5 key topics from Daniel Kim, John Glasgow and Anna Tutova at HumanX Amsterdam 2026. Thicker branches were mentioned more often in the talk.
The Autonomous Finance Office
Click a topic to open it
Key Takeaways
- Redesign finance workflows so AI proposes and humans approve postings.
- Build lineage controls that reconstruct model, evidence, and output history.
- Move repetitive spreadsheet work into AI-readable systems or warehouses.
- Use continuous close data to forecast and act before period-end misses.
- Add semantic layers and permissions before exposing finance data to LLMs.
“If you don't know why AI got the number right then it has no business touching your general ledger.”
Share
Up Next
BBC
WatchNext in agenda
Finance for Humans in an Automated World
Bianca Zwart, Biswa Sengupta, Carmel Crimmins
ai in financepersonalization
BSJ
WatchAlso on ai governance
The Missing Ingredient in Agentic AI: Context
Bianca Anghelina, Stefan Ostwald, Jeff Wilser
agentic aienterprise ai
SKG
WatchAlso on ai governance
Risk, Responsibility & Trusted AI in Zero-Tolerance Environments
Spencer Schaefer, Karen Silverman, Gabriel Banks, Barbara Junge
trusted aiai governance