A cross-section of industry experts gave their insights into AI use in business at Summer Davos 2025.
AI agents are complementing human workers – putting the onus on managers to optimize the capacities of both.
Women need to be better integrated into AI in the workplace, both in their involvement with the technology and in the underlying data.
본 내용은 세계경제포럼이 2025년 6월 27일 홈페이지에 게재한 내용을 옮긴 것입니다.Ian Shine
Senior Writer, Forum Stories
Almost 80% of companies say they are now using AI in at least one business function. However, statistics on whether they are using it to its full potential – or know how to do so – are not as readily available.
Have you read?
- Why China’s AI breakthroughs should come as no surprise
- How emerging markets are rewriting the future of finance with AI
- AI Agents at work: How to select the right type of AI agent for your organization
Moving beyond the question of simple AI adoption into deeper questions about how the technology can be used most effectively has been the focus of several sessions at the Annual Meeting of the New Champions in Tianjin, China. The topics of governance, fairness, use cases and the AI workforce all came up, with speakers from a broad cross-section of industries sharing their experiences of learning to work with this new technology.
AI agents as colleagues and teachers
The arrival of AI agents as co-workers, and their potential impact on the skills workers and managers need, was explored in the session Building an Agentic Economy. Wang Guanchun, Chairman and CEO of Laiye, focused on how AI is changing job roles, and how leaders need to adapt to this shift to stay relevant.
Wang Guanchun, Chairman and CEO of Laiye, focused on how AI is changing job roles.Image: World Economic Forum
“Traditionally, people would think of a good manager as a good people manager with strong communication skills. Now the role of management is changing. Very soon, I think the valuation metric for a good manager will be: How many digital workers can you manage? That’s a different skill set. It’s about how you can prompt your agents to do the best work they can do.”
For Kian Katanforoosh, Founder and CEO, Workera, this change in the skills people need is not a one-off event triggered by the arrival of AI, but will keep evolving. “In AI, skills can become outdated in three months. Companies [have to] stay at the edge. That’s the real moat – can you execute really fast on that? Hire the right talent, hire for AI readiness, hire for learning mindset. In every one of our interviews, we ask: what have you learned in the last 90 days and what is your learning strategy? Helping your team be plugged into the major AI hubs is very important.”
AI’s ability to empower staff in their learning – about AI or any other topic – was highlighted by Van Vu, Co-Founder and CEO of ELSA. “One of the most eye-opening lessons I’ve had was that learners feel a lot more comfortable learning with an AI agent because AI doesn’t judge you and people don’t feel embarrassed. That’s one of the biggest challenges in learning with humans – that you make a lot of mistakes and you feel embarrassed, and you start pulling back. And teachers are uncomfortable giving feedback because they don’t want you to feel bad. AI takes away all of that.”
Finding applications for AI – or ‘AI+’
In The Time of AI+ session, the focus was more on the search for use cases to take AI to the next level in terms of producing tangible benefits for businesses.
“AI+ is AI with applications,” said Zhu Min, a member of the Senior Expert Advisory Committee at the China Center for International Economic Exchanges. “Application, application, application is really the key issue in the next few years. I expect to see in China in the coming 18 months – given the scaling of a pool of engineers, given the massive industry everywhere, and the 1.4 billion consumers – more than 100 different types of DeepSeek-type software across all sectors. That will fundamentally change the nature, and the tech nature, of the whole Chinese economy.”
Peter Koerte, Chief Technology Officer and Chief Strategy Officer at Siemens, the “killer application” for AI right now is in software development.Image: World Economic Forum
For Peter Koerte, Chief Technology Officer and Chief Strategy Officer at Siemens, the “killer application” for AI right now is in software development. “We’re talking about large language models and, think about it, computer programming is a language. AI really is good at programming. We’re seeing better productivity by about 20-30%.”
This has helped Siemens develop what it calls an “industrial co-pilot”, able to help its clients in a range of sectors. “It’s a co-worker that works with you in setting up machines, programming machines, and it’s really saving significant costs, 30-40%, and that’s starting now to scale. About 200 customers worldwide are using it and it’s really ramping up in China, because what we see there is an ageing population going out of the workforce, a lot of knowledge getting lost, and the only way to capture that is AI. If we can capture all that knowledge, put it into AI, it lowers the barrier for you in terms of being a trained technician.”
Understanding how to use data
Gong Yingying, Founder and Chair of Yidu Tech, an AI-driven analytics firm that works with the healthcare sector, told The Time of AI+ session that 20% of her workforce is already made up of AI agents. She expects this to increase significantly. “Our guys have to be able to work with our AI agents, in the office, and be super-adaptive to this new work environment, and be extremely focused on final use cases.”
How the company got to this position is a lesson in thinking ahead and laying the groundwork. “We spent 10 years just understanding big data, processing it and making it computable,” she said. “We have now helped our clients process about 6 million [health] records … and we have seen very exciting results. A lot of doctors in top hospitals have a very heavy workload … and they don’t get enough time to know and understand their patients. The AI copilot we launched three months ago – of the top 150 hospitals, about 50 have adopted the platform. The most popular use case is pre-clinical information-gathering. The patient talks to the AI before seeing the doctor, and it gathers the relevant information. Or AI can read 20 years of medical records and create a one-pager for the doctor.”
Kenny Lam, Asia-Pacific CEO of hedge fund Two Sigma, said that his company is increasingly looking at how to use AI trained on specialized, localized sources of information and that can keep drilling into these niche areas – rather than drawing on a single global source that may be too broad to offer the laser-focused insights investors need.
“We are focused on the scientific process [behind AI],” he told the session. “The core to the scientific process is iterative, meaning we want to find a way to apply AI in a micro sense, but in such as deep way that we get feedback and can iterate constantly.”
Ethics and access to AI
Women make up 25-30% of the AI workforce worldwide but hold only 15% of senior roles, the Women in AI session heard. With Lam among those highlighting that “the war for talent has intensified” in the AI space, what should business be doing to better nurture and make use of women’s AI skills?
“Companies should be more proactive in providing training to women and helping women adapt better to AI and AI-related industries,” said Yang Jingjing, Chairman of the Board at Shanghai Generative Artificial Intelligence Ecosystem Development. “In July, at the World AI Conference, we will have a forum on female role models in the AI industry. We hope to tell more stories about female role models and build more communities to help women in this industry build up their confidence and a better sense of belonging.”
Have you read?
Angela Wang Nan, Chairman and President at software firm Neusoft, said some universities in China have seen very strong growth in female applicants for IT degrees, leading to a 50:50 gender split in enrolment in some cases. Similar trends have been seen with AI majors, which have seen an increase from around 14% women enrolment when they began, to about 40% today, she added.
“This represents not only the willingness of girls to join the IT industry in STEM-related roles, but also society’s perception that girls can succeed,” Wang Nan said.
For Meirav Oren, Executive Chairwoman and Co-Founder of Versatile, women working in AI can attract others to the sector by “just being out there and encouraging super-capable, bright, young and even older women to just apply for these positions”, rather than being put off by the fact that they might not check all the boxes on the job spec.
Gender composition by GenAI process from ‘Gender Parity in the Intelligent Age’ publication, March 2025.Image: World Economic Forum
Making AI reliable
Gender equality in AI is not just an issue for the workforce, but for the AI models themselves. “When we develop AI applications for industry – healthcare, automotive, garment usage – we need to be thinking about demand from females,” Wang Nan underlined. “This means diversity among AI developers really means a lot, and in the data itself. We need really balanced data to train the algorithm to make it more inclusive and serve better the demands of industry.”
John Lombard, CEO of NTT DATA Asia Pacific, agreed, telling the Building an Agentic Economy session that “a key element is the auditability and traceability of these models. Bias or gender bias in these models can be devastating for a brand. Governance needs to be put in place before we embark on these projects, because there’s so much opportunity, so much promise, but if we get it wrong it would be very damaging to our organizations, brands and societies.”
For Koerte of Siemens, speaking to The Time of AI+ session, data quality can only improve if the data itself is shared as widely as possible. “The more data we have, the better the algorithms become. What that means is data has to flow freely around the world. So when we talk about China, when we talk about Europe, when we talk about the United States, it’s not helpful if we create data siloes and say this is just yours. It really helps if we can pool the data together. The more we can share the data between countries, between companies, the better the quality of the models themselves.”
