This weekly update shows how startups around the world are racing to build and use AI agents - smart computer programs that can work on their own to complete tasks.

In India, a major transformation is happening in customer service. Indian startups like LimeChat are creating AI chatbots that can answer customer questions without human help. These AI agents are so good that they can handle up to 80% of customer conversations. This means call centers need far fewer workers than before. India has a huge technology industry worth $283 billion, and this change is affecting thousands of jobs. However, the startup companies say their AI agents provide better and faster service to customers. They are also starting training programs to teach workers new skills in AI and automation.

On October 16, 2025, Zurich Insurance Group announced the winners of their Innovation Championship, where startups competed to show the best AI agent technology. The company held a special Innovation Festival that brought together over 1,000 people including startup founders, technology experts, and investors. Ten startups won prizes for their innovative AI solutions.

One impressive winner was AgentricAI, which created an AI agent called Clara. Clara can automatically process travel insurance claims, which normally takes a lot of time when humans do it. The smart part is that Clara still keeps human experts involved to make the final decisions. Another winner, Interloom, built a platform that helps AI agents learn from the daily work that humans do. This makes it easier for human workers and AI agents to work together as a team.

Wangari Global won for creating an AI system that helps with financial reporting. Their system uses multiple AI agents working together to create reports and find problems faster than humans can. WordLift makes AI tools that help websites show up better in search results. Even Zurich's own employees created a winning project called ZuriFault, which uses AI agents to make faster decisions about car accidents.

Three older startups that won previous competitions also got recognition for growing bigger. Lisa Insurtech has been working with Zurich since 2022 and now processes over one million health insurance claims every year using AI. These success stories show that AI agent startups can grow from small experiments into major businesses.

However, not everyone believes AI agents are ready for widespread use. Andrej Karpathy, a famous AI expert who helped create OpenAI, shared his honest opinion in a podcast. He said current AI agents just don't work well enough. The technology has too many problems - agents aren't smart enough, they can't do enough different types of tasks, and they can't remember things you tell them.

Karpathy explained that it will take about ten years to solve these problems. One expert named Quintin Au from ScaleAI explained why agents struggle. Every time an AI agent does something, it has about a 20% chance of making a mistake. If an agent needs to do five things to finish a job, there's only a 32% chance it will get everything right. The mistakes pile up quickly when agents try to do complex work.

Karpathy believes humans and AI should work together, not that AI should replace people completely. He wants AI tools that show him how they work and ask questions when they're not sure about something. This way, humans can learn and stay involved instead of just watching AI do everything.

Meanwhile, healthcare companies are pushing forward with AI agents despite the challenges. A survey of 605 healthcare and life sciences company leaders found that 44% of companies using AI have already added AI agents. The survey was done by Google Cloud between April and June 2025. These companies most commonly use AI agents for technical support, with 53% using them for this purpose. Security and cybersecurity come second at 49%, and research work is third at 46%.

One healthcare technology company called IKS Health shows how multiple agents can work together like a team. They created a system where different AI agents handle different parts of the medical billing process. First, a scribe agent takes notes while a doctor talks to a patient. Then, a coding agent checks those notes and adds the correct medical codes. If the patient needs insurance approval for treatment, a prior authorization agent steps in to handle that paperwork. At the end, a human expert reviews everything to make sure it's correct. Company leaders say this system makes things better for patients, doctors, and office staff.

Healthcare companies are investing heavily in this technology. About 74% of leaders said they increased their AI spending, and 48% are moving money from other projects to invest in AI agents. They're betting that AI agents will save time and money while improving patient care.