Generative AI models such as GPT-4 have fundamentally changed which jobs are vulnerable to automation, moving beyond routine manual work to tackle complex cognitive tasks that require human reasoning. This shift is forcing business leaders to reconsider workforce planning, employee skills, and how work itself gets done.

How Has ChatGPT Changed What We Think About Job Automation?

For decades, business leaders assumed artificial intelligence would primarily affect jobs involving manual, routine, and repetitive work. That assumption has been upended. When OpenAI released GPT-3 and later GPT-4, it became clear that large language models (LLMs), which are AI systems trained on vast amounts of text data, could do something unexpected: they could learn new tasks without being shown examples first.

"Before, many assumed artificial intelligence would mostly affect jobs where their responsibilities were manual, routine, and non-cognitive. With GPT-3, we realize that these foundational language models are zero-shot learners. Without having seen any examples of a given task, these models can generalize very well to new tasks and produce human-level output. That is quite remarkable, because this creates the opportunity for these generative AI models to be put in the domain of non-routine cognitive tasks, where complex human reasoning was required," explained Christopher Blackburn, senior data scientist in Human Capital Solutions at Aon.

Christopher Blackburn, Senior Data Scientist, Human Capital Solutions at Aon

This capability means that professions built around generating, analyzing, and summarizing large amounts of information, such as law and consulting, now face genuine uncertainty about how AI will reshape their work.

What Does Lower Task Cost Mean for the Future of Work?

When technology dramatically reduces the cost of performing certain tasks, the economic outcome is not always job loss. Instead, demand for those tasks often increases, which can actually expand the market. Consider software development: as AI tools make writing quality code faster and cheaper, companies may invest in more software projects, creating demand for more developers rather than fewer.

"I think the best way to think about these technologies is that it dramatically lowers the costs of doing certain kinds of tasks. When that happens, one possible outcome is that the demand for those kinds of tasks goes way up. I think we may see that actually in software development. As the cost of producing good code goes down, I think we may see the quality of software that people use go up and the areas in which software applied expand," said Muir Macpherson, partner in global human capital analytics at Aon.

Muir Macpherson, Partner, Global Human Capital Analytics at Aon

This pattern has historical precedent. When automated teller machines (ATMs) were introduced in 1978, many predicted the end of bank teller jobs. Instead, the number of tellers in the United States remained stable over the following 30 years because tellers became more productive and could focus on customer service and sales rather than routine cash handling.

How to Prepare Your Workforce for AI-Driven Change

• Build AI Literacy Across All Levels: Employees need foundational knowledge in how AI works, including basic probability and statistics, as well as understanding the ethical implications and limitations of these technologies at scale.
• Establish Data Privacy Protocols: Companies must create clear policies about what information employees can and cannot input into AI systems hosted by external providers like OpenAI, protecting sensitive business data from exposure.
• Create New Roles and Skill Pathways: Organizations should develop job categories that didn't exist before, such as prompt engineer positions designed specifically to test and optimize language model capabilities for business applications.

The reality is that workers are already experimenting with AI tools on their own, often without their employers' knowledge. Many employees are discovering ways to automate parts of their jobs using ChatGPT and similar tools, which means AI literacy will become a core competency that workers bring to the workplace whether companies plan for it or not.

Business leaders are experiencing what Muir Macpherson calls "shock and amazement" at how quickly these tools have progressed. Developments over just six to nine months have caught executives off guard, and the pace shows no signs of slowing.

The challenge ahead is not simply whether AI will eliminate jobs, but how businesses will manage a period where AI capabilities outpace organizational readiness. In fields that generate and process large volumes of text, such as law and consulting, the net impact remains uncertain. One plausible scenario involves AI generating initial drafts of documents that humans then review and refine, with AI tools also used to summarize and analyze those same documents, creating a hybrid workflow that didn't exist before.

What's clear is that generative AI has democratized access to powerful AI capabilities for businesses of all sizes. This accessibility means that companies across industries, not just tech giants, now have tools to drive innovation and optimize operations. The question is no longer whether AI will change work, but how quickly organizations can adapt their strategies, training programs, and job structures to thrive in this new environment.