To keep pace with the AI era, these six core capabilities are indispensable.

To test whether employees are ready to embrace AI, we should not only look at the training completion rate, but focus on business outcomes at the core. If employees are required to continuously learn and pile up new skills along with the development of AI, it will only drag down rather than improve the team's readiness. In the final analysis, cultivating employees' capabilities that can evolve synchronously with the maturity of technology is the foundation for unlocking the true value of AI.

Nowadays, every CEO has a question in mind: How on earth can we truly create value from AI investments?

Many enterprises choose to expand AI training courses to answer this question — for example, offering prompt-engineering related courses, or expanding authorization for multiple AI tools. However, such practices often overwhelm employees and fail to achieve the expected results for enterprises. In addition, with the rapid iteration of AI capabilities, the growth rate of market demand for AI-related skills has far exceeded the supply of available talents.

The root of the problem lies in: enterprises regard AI skills as a constantly expanding checklist. A better idea is to focus on cultivating a small number of core capabilities that can iterate together with AI technology. Gartner divides capabilities into two major sections and extracts six core competencies: collaborative work orchestration capability, and supervision and governance capability. Only by concentrating resources to polish these capabilities can enterprises enable employees to grow synchronously with AI.

Collaborative Work Orchestration Capability

Collaborative work orchestration capability refers to how employees cooperate with AI to achieve business results, including the following three items:

1. Problem Definition Capability

In the early stage of AI transformation, enterprises mostly use AI to speed up original business processes. However, the quality of output from AI tools depends entirely on the quality of input information. If people cannot clearly state what problem to solve at the very beginning, the final result will inevitably be unsatisfactory.

The key to getting the expected results is to be able to clearly define the scope, state the problem, clarify the goals, and set boundaries, so as to guide AI to produce effective solutions.

We have seen that some leading enterprises no longer offer general AI popularization training, but instead the human resources department and information technology department jointly organize cross-departmental seminars, where everyone collectively sorts out and defines business problems.

For example, the human resources and IT team of a global law firm customized seminars for different business units. Employees bring their own business problems at work and use AI to find solutions, including a large number of pain points commonly encountered in daily local business. Starting from problems rather than technologies, employees can develop solid problem definition capabilities. With the support of colleagues and IT personnel, AI solutions can be iterated and implemented faster and more efficiently.

2. Systems Thinking Capability

When AI starts to reconstruct work processes, simply defining problems needs to be upgraded to systems thinking: understanding how each independent module interacts and connects with each other in the overall system. A survey conducted by Gartner on 3029 employees around the world in July 2025 shows that in an AI-empowered work environment, the performance of employees with systems thinking has increased by 1.6 times. Because this capability enables people to deal with complex situations instead of only optimizing isolated individual tasks.

Managers can establish a fixed mechanism to allow employees to immerse themselves in understanding the ripple effects brought by interconnected systems, so as to implement systems thinking.

U.S. Venture, an energy distribution enterprise, mainly deals in renewable energy, traditional energy products, lubricants and tires, and also provides transportation-related data analysis. The management of this enterprise found that if employees are only allowed to use AI sporadically, only small benefits can be obtained; to achieve real growth, business processes must be reconstructed. The human resources department arranges internal employees to follow front-line sales staff on the job, to witness the real obstacles of the team, cumbersome workarounds, and bottlenecks in the approval process.

Before introducing AI, sort out the entire business process together with front-line employees. This not only helps enterprises find real business pain points worthy of solving, but also demonstrates systems thinking in person: allowing employees to collectively understand the logic of interlocking work links and decision-making steps when embedding AI into the business.

3. Strategic Foresight Capability

With the continuous development of AI, it will eventually help enterprises create new working modes, greatly expand business scale, and even subvert the original market. For example, in the next few years, sales staff will have AI agents to cooperate with them: screening priority follow-up leads, carrying out research and analysis, and independently cultivating potential customers.

To achieve this, systems thinking needs to be further upgraded to strategic foresight. When enterprises use AI transformation to create new products, new services or new business models, the focus of human capabilities should shift to predicting future scenarios and building adaptive solutions. The goal is not to accurately predict the future, but to establish a mindset of embracing changes and predict disruptive changes in advance.

Managers can embed "hypothesis deduction" scenario planning into the existing decision-making process to cultivate this capability.

Clifford Chance, a global law firm, closely collaborated with management, partners, lawyers, knowledge experts and technical teams to promote the implementation of AI and embed AI into legal business processes. They jointly studied how AI will reshape customer expectations, legal service delivery modes, and the operating modes of law firms in the coming years. The law firm not only focuses on short-term productivity improvement, but also tracks key indicators such as readiness, implementation and usage, and employee mentality to support the long-term AI strategy and prepare for future changes.

Supervision and Governance Capability

Supervision and governance capability is the ability for employees to ensure that AI outputs are accurate, safe and credible, including the following three items:

1. AI Judgment and Risk Control Capability

In the scenario of AI-assisted work, employees must actively verify the reliability of results, resolve risks, and practice ethical AI governance. If you trust AI outputs without thinking and rely excessively on AI tools, it will easily lead to errors, algorithm bias or compliance problems.

Formal training and codes of conduct for the responsible use of AI are certainly essential, but in daily work, the team needs to reach a consensus: maintain skepticism and verification of AI-generated work content.

For example, when a global market research enterprise launched the AI document summarization tool, it established a supporting "verification and audit" process. In the initial stage of using the tool, employees need to act as fact-checkers to review their first few AI-generated briefings. They check the statistical data according to the standardized scoring form, and trace the cited content back to the original document. If there are unverifiable statements in the summary, they will make a record and feed it back to the development team to fine-tune the model. This practical training with reasonable suspicion helps employees quickly identify common hallucination problems and errors of AI, and turn unconditional trust in tools into conditional and controllable trust.

2. Human-Machine Mediation Capability

The deeper AI transforms work, the more complex the governance work will be. At this time, AI judgment and risk control will evolve into human-machine mediation capability. The improvement of automation will change the work content of human beings, and contradictions will easily arise between machine logic and human judgment. In an AI-empowered work environment, employees will increasingly need to interpret ambiguous scenarios and make measured complex decisions.

To implement and cultivate this capability, enterprises need to form a fixed mechanism: organize the team to carry out post-event reviews and scenario-based pre-deductions to discuss how to handle various complex exceptions.

For example: the AI system automatically cancels the subscription of a long-term cooperative customer because the customer has owed fees for many times, but fails to identify that this customer is being affected by a natural disaster.

Many medical institutions are planning to embed human-machine mediation into business processes. For example, clinical teams can enable a formal "manual veto log". In practice, when radiologists disagree with the AI diagnosis suggestions, they can immediately start the upgrade process: record this veto operation, attach standard reasons and timestamps. The interdisciplinary clinical ethics committee reviews these logs every week, on the one hand to verify manual decisions, and on the other hand to use these data to retrain and optimize the underlying AI model.

3. Ethical Leadership

The final form of the continuous evolution of human-machine mediation capability is ethical leadership. When AI gives birth to new work forms, human beings must be able to continuously control the design and operation of AI systems to ensure fair and responsible outputs. This is not only to verify the output content; people should take the initiative to participate in shaping a credible AI system.

Many enterprises are carrying out "red-blue team testing" on a regular basis: arranging employees to actively apply pressure tests on the AI system to dig out vulnerabilities. Some enterprises even set up incentives to encourage employees to act as internal ethical safety testers, actively look for algorithm bias, or deliberately interfere with the automated workflow before the system goes online internally and externally to find hidden dangers.

A case that many customers resonate with: a medical technology company is developing an AI system for screening cognitive decline in the elderly. The enterprise realizes that this brand-new AI product will bring huge compliance, trust and privacy risks. The human resources department, together with the heads of IT and product, set up an AI ethics committee within the product team, built governance constraint rules, carried out red-blue team drills, and trained employees to carefully examine every model decision, identify bias and ethical risks.

To test whether employees are ready to embrace AI, we should not only look at the training completion rate, but focus on business outcomes at the core. If employees are required to continuously learn and pile up new skills along with the development of AI, it will only drag down rather than improve the team's readiness. In the final analysis, cultivating employees' capabilities that can evolve synchronously with the maturity of technology is the foundation for unlocking the true value of AI.

CV Viverito | Article

CV Viverito is Senior Director Analyst in the Gartner Human Resources Research Department. He mainly serves chief human resources officers and human resource managers, helping them understand the impact of AI on the workforce, make arrangements in advance, and build a working mode in the era of human-AI collaboration.

This article is from WeChat official account "Harvard Business Review" (ID: hbrchinese), Author: HBR-China, 36Kr is published with authorization.