Barney Hussey-Yeo in conversation with John Collison

Professional poker origins

Barney funded his university career playing online poker before GTO solvers existed, and now runs an AI poker agent that earns approximately $200,000 annually in rakeback.

Pre-LLM AI foundation

Cleo launched in 2016 using intent classification and supervised learning, predating transformers and GPT by years but anticipating the natural language processing revolution.

Core mission thesis

The company targets the 'wildly unoptimized' financial decisions of everyday people, focusing on the 60% of Americans with less than $400 in savings rather than traditional wealthy advisory clients.

Humor as financial tool

Early versions employed 30 female comedians to write copy and create 'roast mode,' using personality-driven humor to engage users who typically avoid financial discussions.

Financial health scoring

Cleo maintains a 0-100 financial health score for each user, optimizing its recommendation systems to improve long-term financial outcomes rather than just driving engagement.

Proactive agentic AI

Unlike reactive chatbots that require user prompts, Cleo operates as an 'always on' agent that proactively pushes advice and takes actions like moving money or securing credit on users' behalf.

Structured knowledge bases

Barney argues that vertical AI agents with structured memory schemas and deep domain expertise will outperform general frontier models for specific life decisions like personal finance.

Competing with vibe-based decisions

Cleo targets consumers making emotional or 'vibe-based' financial choices rather than those using spreadsheets, capturing value from the vast majority who lack financial optimization.

Automation vision

The ultimate goal is eliminating financial admin entirely—creating a system where credit, savings optimization, and investing happen automatically without users checking balances or transaction lists.