Entrepreneur in Residence, CFO Agent
About Us
Zoth is a stablecoin-led neobank for the Global South and the agentic economy, delivering yield, payments, and card services to non-crypto-native users on institutional-grade infrastructure under a regulated framework. To date, Zoth has reached a peak TVL of $75M, processes $40M+ in average monthly payment volume, and has generated $5.2M in revenue.
The CFO AI
The CFO AI is an autonomous treasury and finance agent. It moves stablecoins, funds accounts, reconciles balances, and executes payments on behalf of businesses, taking routine actions on its own and routing high-risk actions to a human for approval
The product question at the center of this role is not a feature backlog. It is a boundary: what the agent is trusted to do alone, and what it must never do without a human in the loop.
The role
This role owns the autonomy boundary. The core product decisions are exactly what the agent may do autonomously versus what requires approval, and the guardrails that make that boundary safe. Success is measured by how much the agent handles unattended without a trust failure, not by how many features ship.
What you will own
- Autonomy and approval policy. Define, for every class of action, what the agent executes autonomously versus what it escalates for human approval, and why.
- Guardrails and human-in-the-loop design. Design the checks, limits, confirmations, and fallbacks that keep autonomous actions safe and reversible.
- Agent quality and evaluation. Build the evals and quality bar that tell you whether the agent is safe to widen, and where it fails.
- Prototyping against APIs. Ship working prototypes against real payment, banking, and wallet APIs to pressure-test decisions before engineering commits.
- Risk and compliance partnership. Work with risk and compliance to make the autonomy boundary defensible, not just functional.
Must haves
Product experience. 3 to 5 years in product management, including at least 1 year building AI or LLM-powered products hands-on.
Guardrails and HITL fluency. Demonstrated judgment on what an agent should do autonomously versus with approval, and how to design the guardrails and human-in-the-loop around that.
Shipped B2B or fintech, end-to-end. Owned a B2B or fintech product from concept to production in payments, invoicing, treasury, banking, or accounting.
Web3 or crypto exposure. Worked at or built for a crypto company, with a working understanding of wallets, stablecoins, and compliance.
| Requirement | What it means |
|---|---|
| Product experience | 3 to 5 years in product management, including at least 1 year building AI or LLM-powered products hands-on. |
| Guardrails and HITL fluency | Demonstrated judgment on what an agent should do autonomously versus with approval, and how to design the guardrails and human-in-the-loop around that. |
| Shipped B2B or fintech, end-to-end | Owned a B2B or fintech product from concept to production in payments, invoicing, treasury, banking, or accounting. |
| Web3 or crypto exposure | Worked at or built for a crypto company, with a working understanding of wallets, stablecoins, and compliance. |
Strong plus
- Hands-on agentic experimentation. Built and iterated on agentic workflows and multi-agent architectures, using agent skills, plugins, and frameworks such as the Claude Agent SDK, OpenClaw, Hermes, or similar.
- Prompt and context engineering. Fluent at shaping agent behavior through prompts, tools, and context, not just specs
- Builder-PM. Ships own tools and prototypes without waiting on engineering.
- Cost and latency awareness. Understands the token cost, latency, and reliability tradeoffs of LLM and multi-agent systems.