On June 10, the World Economic Forum unveiled the 2026 list of Technology Pioneers in Geneva. 100 startups, from 23 countries, 10 from China.
This matter is handled every year, but this year's 100 companies are somewhat different. The forum itself offered a judgment: these companies are not building AI applications, but rather "the infrastructure that enables AI agents to operate on their own." Let me translate that—they are issuing ID cards, opening bank accounts, and installing security systems for AI agents.
Two new companies represent a new track
Among these 100 companies, two made me go back and read them several times.
One called Skyfire, based in the US. Its business description is just one sentence: "Build the underlying infrastructure for identity verification and payments, enabling AI agents to autonomously conduct commercial activities."
Another one called Paid, based in the UK. "Building AI agent-oriented commercial infrastructure, covering pricing, billing, and renewal management."
In simple terms, one provides AI agents with an "ID card + bank account," while the other gives them a "checkout counter + subscription system." It is no coincidence that these two companies were both selected as Technology Pioneers by the World Economic Forum. This signals a recognition at the global level: AI agents are about to become independent economic participants, not just tools for humans.
Behind this signal lies a problem that will become increasingly urgent for enterprise CIOs: if AI agents will one day spend money, sign contracts, and manage subscription services on their own, is your IT governance system currently prepared for this scenario?
100 Technology Pioneers for 2026, from 23 countries AI agent infrastructure Identity verification AI payments Security compliance Enterprise integration
Not Just for Startups: Three Entry Paths for Chinese Enterprises
Among the 10 Chinese companies selected this year, three are directly related to "enterprise digitalization + AI." I will highlight them separately:
| Company | What do you do | Reference value for enterprises |
|---|---|---|
| JiuKe Information | Large Language Model + Robotic Process Automation, empowering enterprise digital transformation | Directly injecting LLM into RPA is not about "letting AI chat," but about letting AI run processes for you. |
| Yida Technology | AI Solutions for Trade and Intelligent Systems | Supply chain + AI: The actual essential need for foreign trade enterprises amid tariff fluctuations |
| DeepWise | Develop an automated machine learning platform for e-commerce and manufacturing | AutoML sinks into manufacturing, model training is no longer just the job of the algorithm team |
The commonality among these three companies is that none of them are building "general-purpose large models"; instead, they embed AI into specific enterprise business processes. JiuKe Information stuffs LLMs into RPA, Yida Technology uses AI to process trade documents and compliance, and DeepWise enables factories to run AutoML for quality inspection and production scheduling — ultimately, it's all about "AI in processes," not "AI for a chat."
This direction is consistent with the World Economic Forum's assessment of this year's entire Technology Pioneer cohort: the next wave of AI value lies not in the model layer, but in the infrastructure layer and embedded implementation within business processes.
AI needs an "identity layer," and enterprises need it too
If you look back at the enterprise AI news from the past month—Microsoft's release of the Agent Control Specification (ACS) at Build, Google's launch of AI Control Center, and IDC's prediction that there will be 13 billion AI agents globally by 2028—you'll notice these events seem to head in different directions, but they all point to the same gap:
When AI agents start working in enterprises, who will manage them? How do they prove "I am who I am"? If they spend a sum of money, who does that money belong to? If they access a system they shouldn't, who can immediately shut them down?
What Skyfire and Paid are doing is essentially addressing the underlying layer of these questions: creating a digital identity layer and an economic behavior layer for AI agents.
This is not irrelevant to domestic enterprises. If your company already has dozens of RPA robots running and several AI models handling approvals or quality inspections — do they currently have independent operational identities? Can the audit logs trace back to "which AI made the decision"?
Frankly speaking, the vast majority of companies currently have no answer. AI operations within business processes are mixed in with human operation logs, making it impossible to tell who caused an incident when something goes wrong.
My judgment: Build the governance framework before the infrastructure becomes entrenched
The 100 companies selected by the World Economic Forum this time are not meant for investors, but as a "weather vane" for policymakers and corporate decision-makers. Over the past 26 years, this list has included Google, Twitter, Airbnb, and Spotify.
It is already quite clear where the wind is blowing this year: the infrastructure for AI to evolve from a "tool" into a "participant" is being laid. The emergence of companies like Skyfire and Paid means that the issue of AI agents' "citizenship" has moved from academic discussion into the commercialization stage.
For enterprise CIOs and digital leaders, my advice is three things:
First, establish an identity audit system for AI operations now. Don't wait until an AI agent causes an incident to patch things up. Every operation executed by an AI should have an independent identity, independent logs, and an independent audit trail. This is the lowest-cost measure, and also the one that will be most expensive in the future if not done now.
Second, pay attention to the impact of AI agents' "payment capability" on your industry.If your industry involves a large number of B2B transactions, subscription services, and automated procurement—the day when AI agents can spend money and renew subscriptions on their own may come sooner than you think. Look at what Skyfire and Paid are doing, and consider whether your supply chain and customer channel systems are ready.
Third, don’t just focus on large models; Chinese companies in the infrastructure layer are worth attention. Companies like JiuKe Information, Yida Technology, and DeepWise AI, which embed AI into business processes, are doing the "last mile" work. Models can be swapped, but once process integration, data governance, and compliance frameworks are established, that is the true moat.
