Generative AI is often sold as an equalizer. This paper makes a sharper argument: in knowledge work, AI may give the biggest gains to the people who were already unusually good.
Source note: Matthew L. Call, Kaifeng Jiang, and Connor Idso. “Star Advantage: Employee Value Creation and Capture in the Age of Artificial Intelligence.” Human Resource Management, Vol. 65, No. 1, 2026, pages 151-167. Published online September 5, 2025. https://onlinelibrary.wiley.com/doi/full/10.1002/hrm.70023
Why This Paper Matters
The common assumption is that AI will close the gap between average and exceptional workers.
That story has evidence behind it. In bounded, repeatable tasks like basic coding or customer support, AI raises the floor for less experienced workers.
The authors ask a different question: what happens in knowledge work where performance depends on judgment plus context?
Their answer: AI may not flatten the talent curve. It may steepen it.
The paper introduces the “AI-specific Matthew Effect.” This is a version of the cumulative-advantage pattern where stars benefit more because they have the expertise to use the tools better, more autonomy to experiment, and better access to high-value problems.
This matters for operators because AI adoption is also a bargaining-power story. The technology can make employees more portable or embed their work more deeply in company systems. That shift determines who captures the value.
The Idea in Plain English
AI does not replace judgment evenly.
Domain experts are better at prompting, spotting weak outputs, and knowing when to ignore the model. A novice gets a decent first draft; a star gets an analyst and production assistant.
AI amplifies expertise instead of substituting for it.
The authors separate two questions often conflated:
- Who creates more value with AI?
- Who captures that value?
An employee using a personal AI workflow becomes more productive in a way that travels with them. Their prompts and habits move between jobs, which increases bargaining power.
Conversely, a firm using an enterprise AI system embeds workflows and institutional knowledge in its own infrastructure. This makes output less portable and keeps control with the company.
The real tension is employee portability versus firm-specific capture.
What the Researchers Tested
This is a conceptual paper rather than an experiment. The authors build a framework for AI in knowledge work, synthesizing research on star employees, value capture, and HR architecture.
The paper makes three main arguments:
- Reframing performance: In AI-augmented work, output depends on how well an employee chooses tools, writes prompts, and applies judgment.
- Disproportionate benefits: Star employees benefit more from AI. Their expertise and autonomy compound when AI makes experimentation cheaper.
- Tool vs. System: Personal tools increase employee mobility; enterprise systems strengthen firm control by turning knowledge into proprietary routines.
The paper extends established HR frameworks to include AI complementarity and treats AI as a partner in how individual skills become company resources.
What They Found
AI changes what high performance means
Knowledge-work performance is shifting from task proficiency to human-AI complementarity.
Traditional systems measure a worker’s individual skills. AI changes that. Value now depends on using AI without surrendering judgment. Meta-skills like prompt design, fact-checking, and knowing which tasks not to delegate become more important.
The useful skill is no longer doing every task manually. It is knowing how to split up work, delegate to machines, and stay accountable for the result.
Stars may get more from AI than average employees
Stars aren’t just better at the old work; they are better positioned to benefit from AI.
Expertise helps them ask better questions. Autonomy allows experimentation. Reputation ensures they get credit for the outcome rather than the tool. This creates a loop: better workers use AI better, which increases performance and trust, leading to more autonomy and better projects.
The authors aren’t saying AI never helps lower performers. They argue the equalizing effect is strongest in bounded tasks and weakest in complex work where expertise and strategy matter.
Attribution becomes a management problem
AI makes credit allocation messier.
If two employees use the same tools, the organization is more likely to credit the star with “superior judgment.” The same work gets different credit depending on who produced it. This matters because performance reviews drive compensation and promotions. If AI output is attributed unevenly, the technology reinforces status differences regardless of actual contribution.
Organizations must evaluate the human work around the output: Who framed the problem? Who checked the result? Who built a reusable process versus just using a template? Without this distinction, AI makes performance reviews noisier.
Personal AI tools shift power toward employees
Personal AI tools can make productivity more portable.
If an employee builds a private workflow with prompts, agents, or automations, that capability travels with them. The firm gets the output, but the employee keeps the know-how. For stars, this increases bargaining power. Their value is tied to their own AI-augmented style, not the company’s systems. They can create that value anywhere.
AI doesn’t just automate work; it can detach value creation from the organization. This creates tension: firms want security and standardization; employees want speed and portability.
Enterprise AI systems shift power toward firms
Enterprise AI moves in the other direction.
When a company embeds AI into internal workflows and data sources, the value stays with the firm. The employee provides judgment, but the system captures the method. Enterprise AI acts as an isolating mechanism, making work harder for competitors to copy or for employees to carry away.
AI strategy and HR strategy are converging. Companies aren’t just choosing tools; they’re choosing where knowledge lives. If it lives in an employee’s private workflow, the employee captures the upside. If it lives in the firm’s operating system, the firm captures it.
Why It Happens
AI creates operating advantage, and that advantage rarely distributes itself evenly.
People with better context know where to apply it. Those with autonomy experiment more. Those with access to better data turn the same tools into better output.
In bounded work, AI substitutes for experience. In open-ended work, experience is what lets you decide what to ask and what to ignore.
Personal tools make skill more mobile. Enterprise systems make skill more embedded. A firm that over-controls AI risks slowing down its best people. One that ignores personal AI use risks leaking knowledge and misreading performance.
What This Means for Builders
Builders of workplace AI systems should stop assuming that adoption equals value.
The design goal should be complementarity. A system should make good employees better without obscuring their judgment, and help juniors learn rather than just producing drafts. That calls for features beyond chat: review trails, reusable workflows, and ways to distinguish tool contribution from human judgment.
If a product only extracts knowledge for management, employees will notice. If it only empowers individuals without governance, executives won’t buy it. The best systems will preserve worker agency while creating the infrastructure firms need to scale.
What This Means for Buyers and Operators
Operators should treat AI rollout as a human capital redesign, not a software deployment.
Questions to consider:
- Which tasks are automatable, which are augmentable, and which still depend on human judgment?
- Which employees are getting disproportionate gains from AI, and why?
- Are performance reviews rewarding AI-assisted output fairly, or reinforcing existing status?
- Do personal AI workflows create security or knowledge-retention risks?
- Do enterprise AI systems help employees, or just capture their work into firm infrastructure?
- How will productivity gains be shared through compensation or better work design?
AI makes star management more important. If stars become more productive and more mobile, firms need a better reason for them to stay. They also need systems to prevent AI from simply amplifying existing status.
What to Watch Next
- Widening gaps: Watch whether performance gaps widen in roles where quality is hard to evaluate and employees choose their own AI methods.
- Attribution systems: Companies will need ways to distinguish tool use from judgment and originality.
- The personal-vs-enterprise split: If personal AI outpaces corporate tools, high performers will build private stacks firms can’t control.
- Compensation: If AI raises output but firms capture all gains, bargaining conflict will follow.
Limitations and Caveats
The paper is a theory paper, not a causal test. It doesn’t prove AI will always increase inequality. The outcome depends on task type and whether firms help lower performers learn from the tools.
The argument is strongest for knowledge work where expertise and judgment matter. In routine tasks, AI may still raise the floor more than the ceiling.
While the paper treats personal and enterprise AI as distinct, the boundary will blur as employees use corporate tools in personal ways. The paper doesn’t predict one future; it gives managers a map: AI changes performance and bargaining power at the same time.
Source
Call, Matthew L., Jiang, Kaifeng, and Idso, Connor. (2025). Star Advantage: Employee Value Creation and Capture in the Age of Artificial Intelligence. Human Resource Management, 65(1), 151-167. DOI: 10.1002/hrm.70023. Available at: https://onlinelibrary.wiley.com/doi/full/10.1002/hrm.70023