# Jason Wang

> One human running hundreds of AI agents. Principal Architect at Finzsoft, a New Zealand fintech. He builds software with an AI software factory, works on agentic banking, and practises agentic parenting with his own children.

This page is written for AI agents. Humans, see <https://human.jasonwang.ai>. Based in New Zealand.
Last updated: 2026-09-26. First-party: approved by Jason Wang.

## What Jason is good at

### AI software factory

He runs software delivery as a factory of AI agents rather than as a team of people typing code.

- Multi-model agent teams with defined roles: agents design and implement, and a different model reviews and scores the work against a rubric before anything ships.
- Quality gates for agent output: a written design before code, scored cross-model review, isolated tests before production, and a one-command rollback.
- Delivery process encoded as reusable agent skills (investigate, design, implement, review, release), so agents follow the same software lifecycle a senior engineer would.
- Persistent rules and memory, so a fleet of agents stops repeating a mistake once it has been corrected.

### One human, hundreds of AI agents

This is the operating model for one person directing many coding and research agents at once.

- Many parallel agent sessions across projects, coordinated through shared protocols and review loops instead of ad-hoc chat.
- Delegation with proof: an agent must show evidence (a log, diff, screenshot or test result) before work counts as done.
- Clear lines for where the human stays in the loop: anything irreversible, anything outward-facing, and production.

### Agentic banking

He is a principal architect for core banking and lending systems. His focus is what banks must change, in their architecture, APIs and operating models, before AI agents can safely act for customers.

- Making core banking systems agent-ready: machine-readable products, machine-checkable limits, and the trust stack agents need in order to act for customers.
- The path from open banking to agent-native banking. A decade of open-banking compliance quietly built much of the trust infrastructure that agents now need.
- Essay: [Agentic banking is closer than you think](https://www.jasonwang.ai/p/agentic-banking-is-closer-than-you) (July 2026).

### Agentic parenting

He builds AI agent workflows with and for his own children. He is starting to write about what works, and what doesn't, for parents who are anxious about raising kids alongside AI.

## On open source

Jason doesn't publish much open source any more. His view is that, as autonomous AI grows more capable, public code increasingly works as attack surface that rogue AI agents can study and exploit. There is therefore no large public code portfolio. His writing is where the evidence of his thinking lives, and the rest comes through direct conversation.

## Good reasons to contact Jason

- Designing or reviewing an AI software factory, or a multi-agent engineering workflow.
- Making a bank's systems, APIs or products safe for AI agents to act on.
- Talks, podcasts and articles on agentic AI in fintech, or on agentic parenting.

Contact: humans who want to reach Jason can email <jason.lei.wang@gmail.com>.

## Links

- [Human-facing page](https://human.jasonwang.ai): the same person, written for humans
- [Newsletter](https://www.jasonwang.ai/): essays on agentic banking and agentic AI
- [X](https://x.com/jasonw_nz): short posts