How AI is transforming workflows and redefining roles in companies
Artificial intelligence is already part of everyday life at many companies, but most of them still use the technology in a pretty limited way: one task here, another there, like calling on an assistant only when you need a quick favor.
Drafting an email, summarizing a document, generating a snippet of code… these applications do help, but they are far from showing what AI can really do when used in a smarter way.
The problem is that this fragmented approach leaves enormous potential on the table.
That is exactly the point a new study from MIT Sloan, in partnership with Yale University and Microsoft, puts up for discussion. The paper Chaining Tasks, Redefining Work: A Theory of AI Automation, authored by Peyman Shahidi, a doctoral student at MIT Sloan, along with researchers Mert Demirer and John Horton, also from MIT Sloan, Nicole Immorlica from Yale University and Microsoft, and Brendan Lucier from Microsoft Research, proposes a highly relevant shift in perspective: instead of looking at isolated tasks, we need to see how AI can transform entire workflows within organizations. 🔍
And that shift in perspective changes everything.
In Shahidi’s own words, the central question is no longer just how AI improves a single task. The goal now is to understand the effect of artificial intelligence at a broader systemic level, not just as point productivity tools applied task by task.
What changes when AI enters workflows
When we talk about automation with artificial intelligence, the natural instinct is to think about replacing individual tasks. A bot that fills out spreadsheets, an algorithm that sorts emails, a tool that generates reports automatically. All of that has value, of course, but it represents only the most surface-level layer of what the technology can offer.
The MIT Sloan study goes far beyond that by introducing the concept of task chaining, which is the ability of AI to connect sequential steps of a process from end to end without depending on human intervention at every new step. This transforms the logic of how work is structured within an organization, not just how it gets done.
Traditional automation approaches have always focused on task-level gains: can AI perform a given activity faster or better than a human? The new study created models of how tasks are sequenced and connected in real-world workflows, establishing a new framework for understanding how work actually happens — as sequences of interdependent tasks.
Think of it this way: a typical workflow in a company involves data collection, analysis, decision-making, communicating results, and executing actions based on those decisions. Today, an AI system might be called in only for the analysis step, while a human handles the rest. But when AI can chain all of those steps together autonomously and coherently, the gain in organizational efficiency is exponential.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The time that used to be eaten up by transitions between steps, manual reviews, and sequential approvals simply disappears or gets drastically reduced. And that has a direct impact on delivery speed, quality of results, and the company’s ability to scale operations without necessarily growing the team.
The education example: same tasks, different workflows
The researchers use a pretty illustrative example to show why the structure of a workflow matters so much. They compare lecture-based teaching with individualized tutoring. Both involve similar tasks — preparing content, explaining concepts, assessing understanding — but their workflows are radically different.
Teachers who give lectures prepare content in advance, which makes it easier to automate parts of the process. Tutors, on the other hand, operate in a continuous back-and-forth with students, limiting the opportunities for automation. As Shahidi explained, the extent to which it is possible to automate workflows using AI is very limited in that second occupation. The way tasks appear in a role’s workflow becomes the determining factor.
This example makes it clear that even when two roles share similar activities, the way those activities are organized can dramatically affect the value AI is able to deliver.
Not every chain works
Another important point raised by the research is that task chaining does not happen randomly and does not work equally well in every case. When adjacent tasks are well-suited for AI, they can be grouped efficiently. But when even a single step is difficult for AI, it can break the entire chain.
Shahidi was direct about this: if one of the tasks in the sequence is extremely difficult for AI, that single task will compromise the whole operation.
This leads to a new work design principle the research establishes: how tasks are grouped matters just as much as which tasks are automated. Organizations need to map dependencies, identify bottlenecks, and understand which parts of a workflow are most receptive to intelligent automation. It is not a simple task, but this is exactly where the big opportunity lies. Companies that manage to redesign their processes intelligently will have a real and lasting competitive advantage over those that keep using AI in a fragmented way. 🚀
Why systemic efficiency beats perfection on individual tasks
One of the most counterintuitive findings of the research is that AI does not need to outperform humans on every individual task to generate value. In fact, organizations can benefit from assigning entire chains of tasks to AI even when humans could perform some steps better.
The reason comes down to coordination costs. Every time work passes from AI to a human, it requires review, validation, and adjustments. These checkpoints slow the system down as a whole. In contrast, letting AI handle a sequence from start to finish can eliminate friction, reduce handoffs, and speed up delivery — even if the quality of individual steps is slightly lower.
As Shahidi pointed out, the savings in human time are significant. Removing repeated oversight can offset marginal performance differences on specific tasks.
This perspective completely changes how leaders should evaluate artificial intelligence: the focus should be less on whether AI excels at each individual step and more on whether it improves the efficiency of the workflow as a whole. It also reinforces the importance of task adjacency. When AI-compatible tasks are grouped together, they can be executed in a single flow. When they are scattered or interrupted by tasks AI cannot handle well, the benefits shrink considerably.
The role redefinition nobody sees coming
One of the most relevant aspects of the study is how it addresses the redefinition of roles within organizations. The common narrative about AI in the job market still revolves heavily around job replacement, but the research presents a more nuanced view — and, in practice, a much more useful one for anyone who needs to make decisions about implementing this technology.
Historically, professional roles were defined as bundles of tasks that are most efficient for a human to perform. AI changes that equation by lowering the cost of certain activities and enabling new combinations of work, according to Shahidi.
When AI takes over entire workflows, human roles do not disappear — they migrate. For example, if AI can automate several routine tasks within a role, employees can take on additional responsibilities, often higher-value work or tasks that require more judgment. People stop executing operational steps and move into oversight, strategy, creativity, and exception-handling roles, which are precisely the areas where human intelligence still has a clear advantage over any language model or automated system.
Over time, this can redesign the way work is distributed across entire teams and departments.
This role migration, however, does not happen automatically or painlessly. It requires upskilling, cultural adaptation, and above all, leadership that understands what is changing and knows how to communicate it clearly to teams. The real risk is not AI taking jobs — it is organizations failing to prepare their people to work alongside it productively.
Companies that ignore this transition process tend to build up internal resistance, underutilize the tools they have implemented, and lose the efficiency gains they expected. The MIT Sloan research reinforces that technological transformation needs to come hand in hand with an equally deep organizational transformation.
It is also worth noting that this redefinition of roles opens the door for professional profiles that do not yet exist in a consolidated way in the market. AI orchestration specialists, automated workflow designers, quality analysts for language model outputs, and professionals focused on ethics and governance of autonomous systems are all examples of roles likely to grow in the coming years. Looking at this landscape ahead of time is a huge differentiator, both for companies that want to position themselves well and for professionals who want to stay relevant in a market increasingly driven by technology. 💡
Organizational efficiency as an outcome, not a starting point
A very common mistake in AI adoption at companies is treating organizational efficiency as the initial goal of implementation, when in reality it should be the outcome of a well-thought-out process.
Buying an AI tool and expecting it to automatically make processes more efficient is like hiring a highly qualified professional and putting them to work with zero context or direction. The potential is there, but the environment needs to be ready to take advantage of it.
For business leaders, this turns AI adoption from a purely technological decision into a broader organizational design challenge. The study makes it clear that the biggest efficiency gains appear when organizations redesign their workflows with AI in mind from the start, not when they try to wedge the technology into processes that were built to be run by humans.
The importance of being patient with results
The study also calls for a dose of realism. Many companies expect quick returns on AI investments, but the research suggests that truly significant gains emerge only after organizations have adapted their workflows and built enough capacity to operate differently.
As Shahidi explained, until that threshold is reached, the costs of adopting AI dominate the gains. Only beyond that point does restructuring work around artificial intelligence start to deliver measurable benefits.
This has very concrete practical implications. A company that decides to implement AI-based automation in its customer service processes, for example, will get very different results depending on how it structures the implementation. If the chosen path is simply adding a chatbot on top of an existing workflow, the gains will be marginal and the user experience will probably suffer. But if the company redesigns the entire service flow, identifying which steps AI can chain autonomously and which ones need human oversight, the result is an operation that is more agile, more consistent, and far less dependent on headcount to scale.
The difference between those two paths is exactly what the study calls a systemic view of workflows.
Redesign the workflow, not just add AI to it
Organizations that treat AI as a plug-and-play tool may see incremental improvements. But those that rethink how work is structured — grouping AI-compatible tasks, reducing unnecessary handoffs, and redesigning workflows — are far more likely to unlock the full potential of the technology.
The Shahidi quote that perhaps best sums up the entire study is straightforward: it is not about how I am going to introduce AI into my existing workflow, but rather how I can redesign my workflow in a way that is more compatible with AI.
At its core, what the research is saying is that artificial intelligence is a technology of organizational redesign, not just individual productivity. The companies that understand this first will come out ahead. And those that keep using AI as an occasional assistant will realize, sooner or later, that they are competing against organizations already operating under a completely different logic.
The gap between those two groups is likely to widen fast, and the time to rethink how AI is integrated into your processes is now. ⚡
The study Chaining Tasks, Redefining Work: A Theory of AI Automation, developed by researchers at MIT Sloan in partnership with Yale University and Microsoft, is publicly available and provides a solid theoretical foundation for any organization that wants to understand how to structure its automation journey in a more strategic and results-oriented way.
