Summer Davos 2025: AI Ecosystems, Workforce Bottlenecks, and the Future of Work
From trust in robots to glue work, learning velocity, and AI-powered management, here are 14 shifts to watch.
This year, I had the privilege of attending the World Economic Forum’s Summer Davos in Tianjin. It was a special moment: Workera was named a 2025 Technology Pioneer for our work at the intersection of AI agents and the workforce. I was honored to represent the team and share how we’re helping organizations build AI-ready talent.
Back home in San Francisco, I spend most of my time deep in AI agents, tools, and infrastructure. But in Tianjin, the conversation shifted from technology to impact!
Below are 14 key takeaways, drawn from panels, private sessions, and hallway conversations, organized across three key themes:
- Part I: Barriers to delivering value from AI
- Part II: Talent & workforce readiness
- Part III: AI transforming work itself
Part I: Barriers to delivering value from AI
1. It’s not companies competing anymore, it’s ecosystems
During our AI panel, Zhou Yunjie (chairman and CEO of Haier) pointed out that “It’s not companies competing anymore, it’s ecosystems.”
He emphasized the rising importance of ecosystems, whether formed as conglomerates, partnerships, or organized at regional or national levels.
A full-stack AI ecosystem has essential pillars like energy, compute, foundation models, AI applications, and talent & workforce readiness. Once you frame it this way, it’s easier to identify the vertically integrated AI ecosystems. The U.S., for example, functions as its own ecosystem with multiple sub-ecosystems within it:
- Microsoft: Nuclear plants, Azure, OpenAI, Copilot, MS 365
- Google: Fusion power, GCP, Gemini, Google Suite
- Elon’s stack: Tesla, Grok, X
On geopolitical lines, there are of course two clear leaders: the US leads in data centers, models, and apps; China leads in energy, robotics, and is rapidly advancing in models and apps too.
2. Talent remains the biggest gap in most ecosystems, even the leading ones.
According to Workera data, only 28% of non-technical enterprise employees demonstrate Gen AI literacy when assessed.
So, many countries are shifting focus from infrastructure to talent:
- Bahrain, as Minister Noor AlKhulaif shared, is partnering with AWS to upskill public-sector employees in cloud and AI.
- Armenia, according to Minister Narek Mkrtchyan, is launching national programs to prepare its workforce for the digital era, including collaborations with tech partners.
- US building AI-ready workforces through Workera (partnering with Marine Corps), and investing early through the new executive order on youth AI education; establishing a national K–12 curriculum, teacher training, apprenticeships, and a Presidential AI Challenge.
Even the best AI stack is useless if your people can’t (or don’t) use it.
3. Robots are coming fast, but their form is still debated
A major theme this year: the integration of AI agents into physical systems.
Should robots be humanoid or utilitarian? The argument for humanoid design is that our world is built for humans (factories, kitchens, tools). But many argued that form should follow function. Think about how strange it would be if your smart vacuum cleaner was shaped like a human.
Regardless of shape, one thing is clear: robotics is accelerating. But we’re still waiting for foundation models that can reason and act in the physical world.
4. Trust will be a real bottleneck for robotics adoption
On our AI panel, Wang Xingxing, CEO of Unitree, shared a story: when a robot steps on someone’s foot, people panic. But when a person does the same thing, they don’t. Same action, very different reaction.
This is more than a tech challenge, it’s psychological. As robots enter public and professional spaces, emotional design, cultural norms, and workforce training (knowing how to work with the machine) will matter just as much as the robot’s performance. Change management will be critical.
5. AI tools are getting better, but workforce adoption is lagging
There’s clear mass adoption of products like ChatGPT and DeepSeek. But beyond that, usage drops off quickly. Even tools like Cursor and Claude Code (which are second nature in San Francisco) are still mainly used by early adopters in technical roles.
We’ve built some of the most powerful workplace tools ever. But most employees haven’t integrated them into their workflows. Execs everywhere are realizing: “it’s really hard to get people to change how they work.”
The tools are ready. The workforce isn’t. That’s a real bottleneck.
Part II: Talent & Workforce Readiness
6. The workforce readiness gap is wider than we think
Saadia Zahidi and Till Leopold from the WEF team shared findings from their 2025 Global Jobs Report, showing just how big the challenge is and how much learning velocity and workforce movement will be needed in the years ahead.
- +78 million jobs expected by 2030
- 39% of current skills becoming obsolete
They gave a useful breakdown for a representative group of 100 workers:
- 41 would not need training by 2030
- 29 would be upskilled in their current role
- 19 would be upskilled and re-deployed
- 11 would be unlikely to receive the necessary upskilling
7. Most teams think they’re ready. They’re not. And it starts at the top
One of the biggest blockers to learning isn’t motivation, it’s unknown unknowns. To give you a concrete number, 71% of professionals misjudge their skill level on Workera before taking their objective assessment. Most teams are operating with a false sense of readiness, and that slows down every learning initiative.
But this lack of self-awareness doesn’t just come from individuals. It often starts at the top. Executives feel pressure to sound AI-ready. Boards expect strategy. Customers want innovation.
So leaders pretend. And when leaders pretend, teams do too.
The better approach? Admit what you don’t know. Take the same learning programs as your team. Share your results. I’ve seen firsthand how leadership humility and vulnerability builds trust and dramatically accelerates upskilling in AI transformations.
8. Generational patterns are shaping AI adoption
My friend Omar Bawa put it well:
- Boomers use AI like search
- Millennials use it like an assistant to complete tasks
- Gen Z uses it like an operating system
That’s why reverse mentorship (where younger employees teach others how they use AI) is quite interesting. And why, cutting junior hiring entirely might not be a viable long-term strategy.
Part III: AI Transforming Work Itself
9. AI shines in structured roles, but struggles with complexity
AI shines at well-defined tasks, which many junior roles are built around. Here’s a simple illustration:
- Junior engineers write scripts
- Mid-level engineers build features, connecting scripts.
- Senior engineers build entire products, connecting features. They do lots of glue work: resolve ambiguity, integrate systems, manage dependencies
Glue work is still hard for AI. That’s why adoption skews toward juniors: it’s about task structure, not age.
In fact, I’d expect that the productivity boost from AI will be inversely correlated with how many distinct tasks a job entails, i.e., how much glue work is has.
You see the same pattern in finance and professional services when comparing an analyst, consultant, and partner. Partners are often strong at glue work because they’ve built up a wide range of experiences over time.
10. We need to teach glue work (or systems thinking) earlier and faster
We’re exploring onboarding programs that replicate 1-2 years of systems thinking in just a few weeks. That means surfacing unknown unknowns, grounding learning in context, and helping new hires understand how the pieces fit together.
Imagine if a consulting firm could give the level of context a partner has to an analyst within their first 2 months. That’s what companies should aim for.
11. AI may narrow the performance gap
Carl-Benedikt Frey shared a study showing that AI boosts the performance of low and average performers more than top performers.
The floor rises faster than the ceiling.
This reframes how we think about performance distribution. The roles of training, support, and workflows will need to evolve accordingly.
12. AI will take over skills interviews, and that’s a good thing (there will be no coming back)
I’ve interviewed and been interviewed many times. But when it comes to skills assessment, I don’t think I’m that good at it. I know Workera’s agent, Sage, is already better than me.
Humans are biased and inconsistent. AI might still have bias today, but it’s fixable. You can calibrate, test, and audit a single AI system far more efficiently than training 1,000 human interviewers. And that’s the point:
With 1,000 interviewers, you get 1,000 different standards. With AI, you get one and the ability to refine it fast.
Yes, it’ll feel strange at first. But if it’s more fair, more accurate, and more scalable. Then, it’s the right move for building a meritocratic, skills-based workforce.
13. Management and org design will change more than you think
Jensen Huang reportedly manages 40+ direct reports. I’ve discussed with companies how a 1:100 manager-to-IC ratios, supported by AI, would work.
That’s possible not because humans changed, but because AI now supports core talent management tasks:
- Interviewing and measuring skills with less bias
- Knowing every project happening across the company
- Mastering the latest content (internal and external)
- Instantly connecting you to the right expert when you need help
If you’re a manager, you’ll soon stop trying to decide the details of what your team should learn or constantly search for the right content. In a year from now, your role will be about setting a high bar, defining clear goals, and letting AI guide your team on how to get there. You’ll scale your capacity, and they will move faster.
If you’re an IC, most of your learning will come from AI, not your manager. Relationships at work will shift toward building human connection and a sense of belonging, rather than being the primary source of skill development.
14. Transparency is key to talent matching
Ying Ni, CEO of Adecco China, emphasized how mismatches between candidates and employers often stem from a lack of transparency.
He expressed excitement about AI tools that give candidates a realistic, immersive preview of what a role entails, before they ever apply.
For example, imagine a sales candidate stepping into a 10min AI simulation that mimics a typical day: joining a virtual standup with teammates, responding to a simulated customer objection, and navigating CRM tools under time pressure. Paired with insights into team norms, communication styles, and decision-making culture, these experiences help candidates self-assess fit before day one. You could even do it during the assessment process.
I think this will lead to better matches and is a win-win.
Final Thought: Tech Acceleration, Learning Velocity & Workforce Movement is the Mandate
The companies that win won’t just move fast on tech, they’ll move fast on people. If you’re not building learning velocity, workforce movement, and adaptive systems, you’re not ready for this new work era!
Thanks for reading. Would love to hear your thoughts in the comments. If you’re exploring these shifts or trying to close your own workforce gaps, feel free to reach out, or learn more at workera.ai.
