At the "Crypto 2026: From Cryptocurrency to Smart Economy" forum in Hong Kong, industry leaders discussed the evolution of AI agents from conversational to executional roles. Laughing, Head of Asia-Pacific at KiteAI, highlighted this shift as a long-term trend, despite challenges in trust and data security. Jerry from Mentis noted issues like fuzzy memory and token-wasting loops in current AI agents, while Ricardo from Predict.fun emphasized the rarity of reliable, profit-generating trading agents. Michael from Brevis pointed out that AI agent processes often lack transparency, making verification difficult. He suggested zero-knowledge proof technology as a solution to ensure compliance without compromising privacy. Anita from Sentient discussed efforts to integrate AI and blockchain to support the digital economy. The forum identified programming, prediction markets, and API marketplaces as promising areas for AI agent application, with future potential in content generation and agent-to-agent interactions.