AFAC 2026 Financial AI Finals Draw Nearly 20,000 Contestants, Spotlighting New Talent Criteria

According to a report by QbitAI, the competition featured two tracks: a challenge track and a startup track. Contestants said this year's tasks were significantly harder than in previous years, as they involved noisy real-world data, sparse feedback, limited computing resources, and long-text answers that had to be accurate while controlling token consumption and finding traceable evidence. The design was intended to filter out people who rely on memorized algorithms and parameter stacking, pushing them to understand business pain points and the value of problems.

The talent criteria behind the contest were echoed by industry and academic representatives. Yao Quanming, an associate professor at Tsinghua University's Department of Electronic Engineering, said he now asks prospective PhD students whether their research would be swallowed by large models within six months. In his view, pure execution skills have largely been automated by AI agents, while intrinsic drive and the ability to make things happen remain scarce. Li Bei, head of Alibaba Cloud's university cooperation program, noted that benchmarks have been raised overall by AI; contestants are not only competing with peers, but also with a threshold elevated by AI absorbing historical experience. He argued that human value has shifted toward judging problems, understanding business, and taking responsibility for results.

The challenge track's first-prize winner in one category, Ding Ding, exemplified this shift. He said he proposed hypotheses and discussed verification criteria with AI before submitting results. AI handled the time-consuming execution while he decided which limited submissions were worth prioritizing. AI can validate hundreds of hypotheses, but the judgment of which one deserves priority still requires human taste and insight, he said. Ding credited AI assistance with enabling him to compete as an individual and reach the top.

Xu Wanqing, head of Ant Group's wealth and insurance business unit Wealth AI Lab, said his team urgently needs two types of talent: those who can break through conventional thinking and find better methods, and those who can genuinely use AI to solve problems and turn innovative ideas into products. He said such people must have both ideas and strong hands-on ability. Chen Zhaoqun, a senior algorithm expert at Ant Group, summarized a three-stage progression for algorithm professionals: the pursuit of excellence, the ability to handle noise in real business data, and systematic problem-solving skills. He said he saw these qualities in some contestants who distilled general capabilities into vertical agents or used code as a vehicle for self-evolution.

SGInnovate Director Jae Annie Tay described rare talent as 'AI plus X,' where AI is a horizontal capability across vertical industries such as finance, robotics, or cybersecurity. The winners of the fourth challenge, Zhang Nianhao and Wang Yang, said they applied their AI-for-research methodology to AI-for-finance, finding the two fields shared many fundamentals. Their main effort remained on the AI framework itself, with domain knowledge quickly supplemented by AI.

The competition was also closely linked to recruitment. More than 15 Agent internship positions were opened to AFAC participants after the finals, covering insurance, wealth, investment, and platform architecture at Ant Group. Contestants entered a dedicated referral channel, and the entire process from screening to interviewing was expected to take about one month. Several participants said they received interest from business line executives and academic supervisors on site.

The startup track, meanwhile, explored whether contestant capabilities could grow into stable business organizations. This year's finalists included companies that had already raised funds and secured orders, as well as early-stage teams and one-person startups. Shi Baixin, a tenured associate professor at Peking University, cautioned that technology can be cutting-edge but products cannot be aloof. In financial AI, a demo running on anonymized data does not mean it can be plugged into a financial institution's production system. Ecosystem support was therefore central, with the competition connecting more than 10 incubators, early-stage investors, and AI communities, and collaborating with Shanghai's 'Chuang·Zai Shanghai' competition for three consecutive years. Winning projects can receive further support such as policy assistance, financing roadshows, and incubation space.

The organizing committee said it regularly tracks past winners. A 2024 team from Sichuan University of Media and Communications later established Sichuan Emei Digital Co., Ltd. after gaining exposure through the event. A 2025 winner, Shanghai Ciling Technology, advanced to the second round of 'Chuang·Zai Shanghai' and secured its first overseas order in March 2026.