The jobs AI can do — and those it shouldn’t

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Generative AI is a transformative technology that has the potential to redefine the nature of work. Understanding its role in the workplace, and what makes it different from past automation, requires a shift from what AI can do to what it should do.

Typical analyses of GenAI’s impact on workers focus on whether the technology can perform specific jobs. Such studies often break down a job and assesses the share of the constituent tasks that the technology can execute. For instance, common tasks for a customer service representative in a call centre include interacting with customers, recording interactions and resolving or escalating concerns. GenAI can handle these tasks, implying it could displace such workers.

But consider an occupation that might initially appear equivalent: an emergency service phone operator. The two jobs share many similar tasks. Should we expect them to face equal levels of risk of automation? The answer is more nuanced than technical capability alone. Beyond ethical considerations, automating such roles introduces complex trade-offs involving economics, task design, and operational interdependence.

We believe organisations should consider four pivotal questions when contemplating automation.

First, how complex is the task? Complexity is a key driver of both human labour and AI costs. Emergency service dispatchers solve a wide variety of problems, involving a level of complexity that outpaces the repetitive interactions of a customer service representative. In general, the more complex the task, the less likely it is to be automated, since humans are — for now — better than machines at handling increased complexity.

Second, how frequent is the task? The higher the frequency, the more likely it is to be automated. Machines have a clear advantage in maintaining speed over prolonged periods. Frequently repeated interactions with clients strengthen the economic case for AI replacement of customer service representatives.

Financial Times Global MBA Ranking

This article is from the 2025 MBA report and ranking

Third, how interconnected are the tasks? In providing a service or creating a product, many jobs are involved in a chain of interconnected tasks often completed by different workers and machines. What happens during the handoff between tasks is often overlooked. Fragmentation costs arise from inefficiencies and errors in the handoff process.

The initial task for a customer service representative involves conversing with the customer, while the final task is resolving their issue. When different workers or machines are involved, the handoff between these tasks can be costly. If the worker handling the final resolution did not interact with the customer initially, additional time would be needed to review all previously gathered information.

High fragmentation costs should discourage companies from dividing tasks between humans and generative AI, even if technically feasible. Automating the initial triage call in emergency services might seem cost-effective, but crucial information could be lost during the transition from AI to a human dispatcher.

Fourth, when executing a task, what is the cost of failure? Mistakes by emergency dispatchers pose significant risks, particularly in life-or-death situations. And GenAI can be less precise than some past forms of automation. 

These questions should guide companies considering automation and help explain why GenAI affects certain occupations more than others. Consider computer programmers, for example. Extensive, well-documented coding examples enable GenAI to provide effective solutions even for complex tasks. The high frequency and repetitiveness of many coding tasks is a good fit with GenAI.

Well before GenAI, programmers divided up large coding projects, and innovations such as distributed development platforms and modular design have reduced fragmentation costs. Safe testing environments keep the cost of failure low, as many errors in GenAI-produced code can be detected inexpensively. Within our framework, these features help explain why programmers, traditionally beneficiaries of automation, are facing increased disruption from GenAI.

The four questions above highlight what makes generative AI unique as an automation technology. As it evolves, GenAI is demonstrating its ability to manage complex tasks at high speed, making it more versatile than traditional automation. By offering a seamless interface and natural language processing capabilities, GenAI progressively lowers fragmentation costs compared with traditional automation. However, the uncertainty surrounding the output of GenAI potentially increases the risk of failure in a task.

Generative AI is a transformative technology with the potential to reshape labour markets. Its ultimate impact and its likelihood of adoption are shaped by the structure of tasks within a particular occupation. The complexity of tasks, their frequency, fragmentation costs, and the cost of failure, taken together, influence the balance between overt cost savings and hidden costs.