Work in the AI Era

For decades, organizations have lived with a basic design constraint: important decisions often stay centralized because information is dispersed, expertise is uneven, and coordination is difficult. Frontline employees may hold the richest local knowledge, but that has rarely been enough. Without broader information, interpretive support, and coordination mechanisms, decentralization can come at a cost. Hierarchy persisted not simply because leaders preferred control, but because the system often required it.

This paper asks whether AI begins to loosen that constraint. As AI reduces the cost of accessing, processing, and interpreting information, it may alter one of the classic tradeoffs in organizational design: the tradeoff between centralizing authority to economize on expertise and decentralizing authority to make better use of local knowledge. In that sense, the project is not mainly about automation. It is about whether AI changes the underlying logic that has long shaped organizational architecture.

We study this question in two field settings where the stakes of coordination are high and local knowledge matters deeply. In home care, scheduling is not just a logistical problem. It also involves continuity, relational fit, caregiver availability, and sensitivity to the human realities of care. In mental health care, scheduling similarly requires judgment about client needs, therapeutic progress, and timing. In both settings, the traditional assumption was that these decisions needed to remain centralized. Our interventions test whether AI-supported frontline workers can take on greater decision authority without sacrificing decision quality.

The broader argument is that AI may matter less as a tool for replacing people than as a tool for redesigning the organization itself. When embedded in the hands of employees rather than imposed solely from above, AI can function as a form of distributed expertise. It can narrow skill gaps, reduce informational frictions, and make it more feasible to move authority closer to where the best local knowledge resides. That possibility has consequences far beyond scheduling. It speaks to a larger future in which management is no longer defined primarily by the upward flow of decisions, but by the design of systems that allow judgment, information, and coordination to operate differently.

At the same time, the paper does not treat empowerment as automatically beneficial. Expanding decision rights can improve ownership and motivation, but it can also create pressure if employees experience greater authority without sufficient support. That is why the paper examines not only whether AI-supported decentralization improves decision quality, but also whether it changes how employees experience their work. The question is not simply whether AI allows authority to move. It is whether organizations can redesign that authority in a way that actually works.

At its core, this project is about a much larger possibility. If AI can reduce the structural reasons organizations have historically relied on centralized control, then the future of AI at work will not just be about doing the same things faster. It will be about building organizations on a different logic.

Employee creativity is essential for innovation and business success, but ideas only create value when employees are willing to share them. That makes it a central challenge for managers to create systems that encourage idea sharing while also allowing ideas to be reviewed fairly and efficiently. We examine whether AI can help with that challenge by evaluating employee ideas rather than requiring managers to review every submission themselves. This matters because paying employees for each idea they submit can increase idea generation, but in practice it is often difficult to implement because it places a heavy review burden on managers. Before organizations adopt systems like this, though, it is important to know whether employees will respond positively to being evaluated by AI.

Our results suggest that they do. When employees are rewarded for each idea they submit, AI leads them to share more ideas, and those ideas are more useful on average. We do not see the same benefits under tournament-style incentives, which suggests that AI works best when it is paired with the right incentive system. Most importantly, some of the very best ideas are especially likely to emerge when AI evaluation is combined with per-idea rewards. Overall, the results suggest that AI can do more than make idea evaluation easier to scale. It can also help organizations bring out more of the creativity already within their workforce.

When AI becomes part of the workflow, traditional incentive systems may no longer direct effort in the right way. Some dimensions of performance are easier for AI to support, while others still depend more heavily on human judgment, creativity, and discernment. That makes the central challenge not just technological, but managerial: how should firms motivate people when the value of human effort is changing?

This paper explores what that shift means for performance evaluation and incentive design. In doing so, it moves the conversation beyond AI capability alone and toward a broader question of how organizations can make human–AI collaboration more effective.