📰 Key Highlights

Fintech company Revolut announced a major upgrade to its crypto exchange Revolut X last Friday, allowing users to connect AI assistants directly to the platform. Currently supported AI tools include Claude, Gemini, Cursor, and OpenClaw. Users can operate these AI assistants through natural language commands to complete three core functions: market analysis, trading strategy backtesting, and actual order execution. This design lets non-technical retail investors drive quantitative trading workflows through conversation, without manually writing code or navigating complex interfaces. Notably, Revolut X adopts a risk-control mechanism similar to Kraken’s platform—all AI-suggested trading orders must go through manual review and confirmation by the user before being submitted, ensuring the final decision-making authority stays with the user and preventing the AI from autonomously executing unintended operations. This upgrade signals that traditional fintech platforms are actively integrating AI Agents into their trading ecosystems, but the “human-in-the-loop” confirmation mechanism also reflects the industry’s cautious attitude toward fully autonomous trading.

💬 JudyAI Lab Perspective

Revolut X has announced that AI assistants like Claude and Gemini can now be plugged directly into its crypto exchange, letting retail users handle market analysis, strategy backtesting, and actual order placement through natural language—marking the formal entry of traditional fintech platforms’ AI Agent integration into the commercial deployment stage.

The most noteworthy aspect of this case for AI builders isn’t “AI can place orders for users”—it’s the design philosophy chosen by Revolut X and Kraken: AI handles analysis and suggestions, but the confirmation button always stays in the user’s hands. The higher the risk and the more irreversible the operation, the clearer the human-machine boundary needs to be. This reminds us that the more autonomous an Agent becomes, the more important it is to design the “confirmation node” clearly—not just for compliance, but to help users build long-term trust in the system. Building this node into the product flow rather than bypassing it is the design decision worth learning from.

If you’re designing a product where “AI executes operations on behalf of the user,” you can ask yourself right now: Where is the confirmation node? Does the user clearly understand what decisions the AI is making for them?

📅 Source Information

🔗 Further Reading