← All work · 2024 · AI
ego: chat-based AI payments
A conversational interface for complex financial transactions — natural-language commands routed through LLMs into secure transaction protocols.
About the project
ego turns the long form of a financial transaction — recipients, amounts, conditions, schedules — into a sentence. An LLM interprets intent, asks for confirmation where stakes are real, and routes the result through secure transaction protocols. The point isn't novelty; it's that the slow parts of moving money are usually typing.
Challenge
Let people complete real financial transactions through conversation without losing the safety properties of a structured payment system. Natural language is forgiving; settlement is not. The two had to meet.
Solution
- 01Designed the conversation model — what the LLM proposes, what the user confirms, where structured forms intercept
- 02Integrated LLM intent extraction with payment-API surfaces
- 03Built confirmation and recovery flows for high-stakes commands
- 04Hardened the boundary between natural-language input and settlement
Results
- →Shipped a working chat-driven payment surface
- →Demonstrated LLM-routed transactions safely bounded by structured confirmation
- →Produced reusable patterns for natural-language financial UX
Next case study
GHash.IO — A high-performance Bitcoin mining pool that briefly reached 51% of global hashrate — and the operational and reputational response that followed