How AI is reshaping workflows and redefining jobs

Most organizations have approached artificial intelligence as a tool for boosting productivity at individual tasks, such as drafting emails, summarizing documents, or generating code. But new research suggests that this task-by-task mindset may be limiting the true value of AI.

A new research paper, “Chaining Tasks, Redefining Work: A Theory of AI Automation,” argues that AI’s biggest impact comes from how it reshapes entire workflows — specifically, how tasks are sequenced, grouped, and handed off between humans and machines. 

“The central question is no longer just how AI improves a single task,” said Peyman Shahidi, a PhD candidate at the MIT Sloan School of Management and co-author of the paper. “We’re trying to understand AI’s effect at a broader system level, not just as spotty productivity tools at the task level.” 

The paper is co-authored by Mert Demirer and John Horton of MIT Sloan, Nicole Immorlica of Yale University and Microsoft, and Brendan Lucier of Microsoft. 

AI value emerges at the workflow level

Traditional approaches to automation have focused on task-level gains, such as whether AI can perform specific activities faster or better than a human. The new research created models of how tasks are sequenced and connected in real-world workflows to create a new framework for how work actually happens, which is as sequences of interdependent tasks.

That shift matters because even when two roles involve similar activities, the way those tasks are arranged can dramatically affect how much value AI can deliver. The researchers highlight lecture-based teaching and tutoring as an example. Both involve similar tasks, but their workflows differ. Teachers, for example, prepare content in advance, making it easier to automate parts of the process. Tutors operate in a continuous back-and-forth with students, limiting opportunities for automation.

“The extent to which you can automate your workflows using AI is very limited in that second occupation,” Shahidi said. “How these tasks appear in an occupation’s workflow becomes important.”

That’s where the concept of task chaining becomes critical. Rather than using AI for isolated steps, organizations can link together multiple tasks so AI executes them as a continuous sequence.

Not all chains are equal, however. When adjacent tasks are well suited to AI, they can be bundled effectively. When even one step is difficult for AI, it can break the chain, Shahidi said. “If one of them is super hard for the AI, that single task is going to undermine the entire operation,” he said.

This finding leads to a new work design principle: How tasks are clustered matters as much as which tasks are automated.

AI Essentials: Accelerating Impactful Adoption

Attend Online

Register Now

Why system-level efficiency beats task-level perfection

One of the most counterintuitive findings in the research is that AI doesn’t need to outperform humans at every individual task to create value. In fact, organizations may benefit from assigning entire chains of tasks to AI even when humans could perform some steps better.

The reason is coordination cost. Each time work passes from AI to human, it requires review, validation, and adjustment. Those checkpoints slow the overall system. In contrast, allowing AI to handle a sequence end to end can eliminate friction, reduce handoffs, and accelerate output — even if the quality of individual steps is slightly lower.

“You’re saving on human time cost,” Shahidi said, noting that removing repeated oversight can outweigh marginal differences in performance. 

This reframes how leaders should evaluate AI: They should focus less on whether it excels at each individual step and more on whether it improves the efficiency of the entire workflow. It also reinforces the importance of task adjacency. When AI-friendly tasks are clustered together, they can be executed in a single flow. When they’re scattered or interrupted by tasks that AI struggles with, the benefits diminish.

Redesigning work and expectations for AI

Historically, job roles have been defined by bundles of tasks that are most efficient for a human to perform. AI changes that equation by reducing the cost of certain activities and enabling new combinations of work, Shahidi said.

For example, if AI can automate several routine tasks within a job role, employees can take on additional responsibilities — often, more judgement-based or higher-value work. Over time, this can redraw how work is distributed across teams and functions.

For business leaders, this shifts AI adoption from a pure technology decision to a broader organizational design challenge. It also requires patience: Many companies expect rapid returns from AI investment, but research suggests that meaningful gains often emerge only after organizations have adapted their workflows and built sufficient capability.

“Up until reaching that threshold, the costs of adopting AI dominate the gains,” Shahidi said. Only after that point does restructuring work around AI begin to deliver measurable benefits. 

Organizations that treat AI as a plug-in tool may see incremental improvements, while those that rethink how work is structured — grouping AI-compatible tasks, reducing unnecessary handoffs, and redesigning workflows — are more likely to unlock its full potential.

“It’s not about how I’m going to introduce AI in my existing workflow,” Shahidi said. “It’s about how I can redesign my workflow in such a way that is more AI-friendly.”

Read the research: “Chaining Tasks, Redefining Work: A Theory of AI Automation”
 

Peyman Shahidi is a PhD candidate at the MIT Sloan School of Management. He studies market design and labor economics, with a focus on the effects of AI on labor markets and online platforms. 

Mert Demirer, PhD ’20, is the Ford Foundation International Career Development Assistant Professor and an assistant professor of applied economics at MIT Sloan. He works in the field of empirical industrial organization, and his research focuses on firm productivity, firms’ use of digital technologies, antitrust issues, and the productivity effects of generative AI.

John Horton is the Chrysler Associate Professor of Management and an associate professor of information technologies at MIT Sloan. His research focuses on the intersection of labor economics, market design, and information systems. He is particularly interested in improving the efficiency and equity of matching markets.

Nicole Immorlica, ’00, ’02, PhD ’05, is a professor of computer science at Yale University and a researcher at Microsoft. She studies the design and operation of sociotechnical systems, drawing on tools and modeling concepts from theoretical computer science and economics to understand and influence behavior patterns in various online and offline systems, markets, and games. 

Brendan Lucier is a senior principal researcher at Microsoft Research New England in the Economics and Computation group. His research lies at the intersection of microeconomic theory and theoretical computer science, and he is especially interested in the ways that users interact with (and through) algorithms and how it informs market design.