Exclusive interview with Kairui International's rotating CEO, Zeng Cheng: It's no longer the case that "all AI positions are popular." The competition for AI talent is shifting from general capabilities to scenario implementation.

Recently, UBTECH has once again driven a surge in AI talent recruitment by launching a global hiring drive for “Chief Scientists of Embodied Intelligence” with annual compensation ranging from RMB 15 million to RMB 124 million.

What is the current state of AI talent recruitment? What trends will emerge in the future? What pain points exist in the recruitment ecosystem? In an interview with a reporter from Securities Times recently, Liu Cheng, the rotating CEO of Korch Talent, said that compensation hiring of more than tens of millions per year is not an industry norm; it typically appears only in a small number of top-tier companies and is often limited to individual cases during specific windows. This move precisely shows that competition for AI talent is shifting from general capabilities to scenario deployment. When embodied intelligence reaches a critical turning point, what companies are competing for is no longer just the talent itself, but the small number of key people who can truly drive technological adoption and define the future competitive landscape.

She also expects that in 2026, the heat level in AI industry-chain recruitment will structurally continue and will not be the case that “all AI positions are hot.” Instead, it will be “the ones that should be hot will be even hotter, and the ones that shouldn’t will naturally cool down,” entering a new stage of “rational prosperity.”

Pay premiums are evident for three types of AI roles

Reporter from Securities Times: What recruiting trends in the AI industry chain have you observed right now?

**Zeng Cheng: **According to real-time monitoring from Korch Talent’s data analytics platform, the demand for the AI industry chain is indeed staying strong, and AI talent needs show three relatively clear changes. First, companies are increasing investment significantly in core algorithm and model engineering talent, with optimization for vertical industry models and upgrades to multimodal capabilities. Roles such as large-model algorithm engineers, algorithm researchers, and engineers who can enable model deployment and performance optimization have long remained at a high level of demand, and recruiting difficulty is also relatively higher.

Second, as embodied intelligence and humanoid robots enter large-scale validation, related cutting-edge roles quickly become recruiting hotspots. For example, in the VLA/L4/world model direction, embodied intelligence algorithm engineers, multimodal fusion algorithm experts, and talent in robotic intelligent control—previously these role needs were scattered, but now they have become a priority target for companies, and salary premiums are also clearly evident.

Third, as AI accelerates deep penetration into physical industries, especially with the implementation of agents, demand growth on the industry application side is being driven. Companies are more inclined to hire composite talent who both understands technology and understands business—for example, agent development engineers and AI solutions architects. At the same time, AI product managers and AI product solution experts who can transform technology into commercial value and make precise insights into users’ needs across different scenarios are also becoming scarce key positions in the market.

In addition, as AI applications deepen further into companies’ core business scenarios, companies are also placing noticeably greater emphasis on model reliability, data quality, and business security. This, in turn, continues to drive ongoing increases in the heat level for roles such as data governance, AI security assessments, and compliance reviews.

Reporter from Securities Times: Has there been a clear increase in compensation levels for recruitment across the AI industry chain?

**Zeng Cheng: **Overall, compensation across the AI industry chain is not rising across the board. The core salary increases are concentrated in scarce tracks and core positions. For top talent who switches jobs, salary increase ranges are generally concentrated between 20% and 30%. Companies also show greater pay elasticity for key technologies and leading-role positions.

The roles with truly obvious premiums mainly fall into three categories. First is the direction of multimodality and embodied intelligence, especially composite talent who combines algorithm, systems, and control capabilities. The compensation premium for related core roles is significant. For example, senior experts in large-model algorithms have an annual salary in the range of RMB 1 million to RMB 2 million. Senior AI Agent technical engineers have an annual salary between RMB 400,000 and RMB 700,000.

Second is the direction of model engineering and large-scale deployment. Simply put, it refers to engineers who can take models from the lab and genuinely deploy them into real business, running stably in production environments. Demand for this talent is strong, and salary growth is also particularly prominent.

Third is “technology + industry + product” composite roles, such as AI product managers and solutions architects. These people need to understand both technology and industry/business, and also be able to interface with commercial requirements—so their compensation levels continue to rise. For example, compensation for senior AI product managers can reach RMB 800,000 to RMB 1 million.

AI industry-chain recruitment demand stays high, but growth rates stabilize

Reporter from Securities Times: Do you predict that the recruitment heat level for the AI industry chain in 2026 will continue, or will it become more stable, or cool down? What is the basis for your judgment?

**Zeng Cheng: **I believe the heat level in AI industry-chain recruitment will structurally continue through 2026. Overall it will remain high, but the pace of growth will stabilize, and there is potential to enter a new stage of “rational prosperity.” Whether in China or in major global economies, AI has already been placed in the position of core competitiveness. Continued investment of policy, capital, and industrial resources determines that this will not be a short-cycle hot trend. From the perspective of technology itself, artificial intelligence is still in the early stages of generational evolution. Multimodal large models, embodied intelligence, and AI for Science have already achieved some preliminary results, but there is still a long way to go before true maturity. As long as technology continues to evolve rapidly, demand for high-quality talent will not stop.

At the same time, AI is accelerating its penetration across all industries. In the past, it was mainly concentrated in highly digitized fields such as internet and finance, but now it is accelerating into physical industries including manufacturing, energy, agriculture, and healthcare. Every traditional industry’s upgrades toward digitization and intelligentization behind the scenes will create sustained and stable talent demand.

But from a trend perspective, it will no longer be that “all AI positions are hot.” Instead, it will be “the ones that should be hot will be hotter, and the ones that shouldn’t will naturally cool down.” For both companies and talent, this is actually a good thing.

Reporter from Securities Times: A good thing for both companies and talent—how should this be understood? How would you evaluate the current recruitment ecosystem for the AI industry chain?

Zeng Cheng: I believe the current recruitment ecosystem for the AI industry chain is moving from early-stage extreme heat and strong emotions into a more rational, and more structurally driven, stage. On the one hand, talent demand is starting to return to a value orientation. In the past, the market did indeed have situations where people were hired just because they were related to AI. But now companies are becoming increasingly clear that what determines competitiveness is not the number of job openings, but whether the talent can support business deployment. This shift is driving recruitment away from “hype-driven hiring” and toward “capability-driven hiring,” which is a necessary course correction for the entire industry.

Talent structure is also upgrading, and composite capabilities are becoming the mainstream direction. Companies rarely hire for single dimensions anymore—only algorithm people or only business people. Instead, they need composite talent who not only understands the technical principles, but can also connect with industry scenarios and has product awareness. In a sense, this is also pushing talent from the traditional “T-shaped” structure toward a multidimensional “shaped like an eggshell” (兀型) structure. This is a long-term positive for improving the overall quality of talent across the AI industry.

Agile staffing models are shifting from being a supplemental option to becoming a strategic tool. This is something we have observed very clearly in the past two years. As AI technology iterates faster, it becomes difficult for companies to cover all high-end capability needs using traditional headcount establishment systems. As a result, more and more companies are bringing in key capabilities through project-based experts and independent consultants. On one hand, this reduces companies’ labor costs and trial-and-error risks. On the other hand, it provides experienced experts with more flexible and diverse career paths. For example, for a cross-industry company we serve that has entered the AI industry chain, based on our deep understanding of the founder’s target track—including the founder himself—we help the founder sort out business development directions and key talent needs through business and organizational diagnostics. We did not adopt the conventional approach of attracting and headhunting top industry talent, because in terms of time horizons and costs, that does not fit the track and the company’s actual situation. Instead, we help the founder break key modules such as product design, R&D, supply chain, and overseas marketing into project tasks, enabling him to quickly build a cross-domain expert team within three months, forming an agile organization of “core founder + external expert network.” This greatly shortens the product development cycle. The product is now preparing to land in overseas markets first, achieving a breakthrough from 0 to 1.

Shift “hunting for talent” toward a balance of “talent development + talent utilization”

Reporter from Securities Times: In a more rational, more structurally driven AI recruitment ecosystem, are there also risks that need attention?

**Zeng Cheng: **The current recruitment ecosystem is indeed becoming more rational, but there are also certain risks that need to be kept in mind. First, high-end talent is overly concentrated, making it “hard to find a general” for small and midsize enterprises. Top AI talent is monopolized by leading tech giants and star startups. As talent acquisition becomes harder for smaller companies, it may—at least to some extent—undermine the industry’s overall innovation vitality, and even lead to a “top-player dominance” structure.

Second, companies prefer “plug-and-play,” which compresses the growth space for junior talent. Many companies clearly favor senior talent with more than eight years of experience, while investing insufficiently in junior talent with 1 to 3 years of experience. At the same time, some companies lack a complete talent cultivation system. Even after hiring talent, they cannot provide an appropriate development platform, leading to consistently high talent attrition rates. If systemic cultivation mechanisms continue to be lacking, there may be a talent gap problem in the future.

Third, short-term profit-seeking mindsets are rising, creating risks of resource misallocation. Some companies and individuals overemphasize short-term salary returns while ignoring long-term capability building and value creation for the business. Once the market environment changes, it is easy to end up with a “high cost, low output” situation.

Reporter from Securities Times: What would you recommend in light of this situation?

**Zeng Cheng: **For the industry ecosystem, we recommend building a more open mechanism for talent mobility, encouraging talent from big companies to flow to small and midsize enterprises and traditional industries. Through models such as talent sharing and technical advisors, AI capabilities can be applied more broadly to enable real-world economic development. For companies, we recommend shifting from “hunting for talent” to placing equal emphasis on “talent development + talent utilization.” On one hand, companies can quickly acquire scarce capabilities through flexible staffing and independent consultants. On the other hand, they should increase internal training investment and build a composite talent cultivation system of “AI + business.” At the same time, they should validate in reverse: define roles using real business problems, maintain rational recruitment, and improve the talent cultivation and retention systems.

Companies that are hiring must clarify their needs before they start hiring. Many companies’ biggest misconception is: “Since other companies are hiring, I also have to hire,” but they have not thought through what problem this position is meant to solve. Is it a technical bottleneck? Does the product need a breakthrough? Or is it already at the critical stage of commercial deployment? If this question is not clarified, even after people are hired, it will very easily turn into the situation where “the people are expensive, but nobody knows what they should do.”

High-end talent doesn’t necessarily need to be “bought out” from the start. For extremely scarce talent at a very senior level, it is entirely possible to collaborate for a period through project-based and advisory-based arrangements first. This not only verifies capability and fit, but also reduces the risk of a company making a one-time large investment. While companies are frantically competing for mature talent, they also need to build a mechanism for identifying high-potential talent. Some people may not be able to “fight hard battles” right now, but they may have strong learning ability, good systems thinking, and enthusiasm for both technology and business. Once they are given the right environment, their growth speed often exceeds expectations.

For talent, you should build a “Π-shaped” capability structure. You must have a sufficiently deep technical vertical axis—within one direction such as algorithms, systems, or engineering. At the same time, you need to understand industry, business, and product horizontally, knowing what technical work ultimately aims to solve. Single-point capability is easy to be replaced, but connection capability will become increasingly valuable. Meanwhile, maintain a balance between hands-on execution and thinking: not only should you roll up your sleeves to write code and run experiments, but also step outside the technical realm to think about industry trends, user value, and the underlying commercial essence.