In the United States, Tarka made waves with its AI agent for LinkedIn designed to help sales teams. The Raleigh startup’s tool acts like a human salesperson - researching prospects, starting conversations, and booking meetings automatically. Unlike spam email systems, Tarka’s agent works within LinkedIn’s platform while maintaining the user’s personal brand voice. The company already has 200 waiting customers and aims to hit $1 million revenue this year.

Chinese startups showed strong competition in the AI agent space. Zhipu AI launched GLM-Z1-Rumination, claiming it outperforms OpenAI’s paid tools at multi-step tasks like data analysis. In a real-world test, the AI helped a Xiaohongshu user gain followers and make money within days. Meanwhile, Butterfly Effect’s Manus AI drew attention for reportedly beating OpenAI in web research tests, though users complained about errors during trials.

The push for job automation accelerated as consulting firm McKinsey revealed an AI agent that can evaluate project proposals in 2 days instead of 20. While humans still check the AI’s work, this shows how agentic AI could replace parts of office jobs. A recent survey found 32% of workers believe AI will reduce job opportunities.

Startups face technical challenges in making reliable AI agents. Tarka struggles with LinkedIn’s rules against automation, while Manus’s mixed performance shows the difficulty of creating “thinking” AI. Chinese companies like Zhipu are trying to compete by offering free tools, hoping to attract users before monetizing later.

Privacy concerns grew as security experts warned some AI agents might expose user data through rushed development. Multiple countries have already banned certain Chinese AI models over security fears. With AI agent spending predicted to hit $7.6 billion in the US alone this year, the race between startups shows no signs of slowing.