There's no longer any doubt that generative AI, based on models like OpenAI’s GPT-4, will impact our jobs.
Workers are already aware of this; 36% of them believe that their job will be replaced by artificial intelligence, according to a study conducted in 13 countries. It’s not just speculation. In recent weeks, reports of AI-related layoffs have been multiplying. A media company recently announced the elimination of 200 jobs (nearly half of its workforce), replaced by an artificial intelligence solution. The CEO of an Indian company that creates e-commerce websites, has announced the layoff of 90% of his customer service teams, replacing them with a chatbot. Business leaders aren't spared either. Almost one in two American CEOs believes that AI would be perfectly capable of carrying out all or part of their roles.
Clearly, GenAI will unlock substantial productivity gains—already certain functions within companies are witnessing productivity surge by over 50%. And though this has sparked anxiety about job losses, it depends on how business leaders intend to use that productivity.
Many businesses will reinvest the savings rather than simply eliminating jobs. Some banks, for example, have harnessed these gains to develop innovative customer outreach strategies, turning previously unprofitable services into profitable ones due to the reduced costs per customer interaction. And higher productivity can also translate into increased capacity where humans are the bottleneck. AI-assisted surgeons, for example, can prepare for operations more rapidly, reducing wait times for ailing patients and potentially increasing the number of people treated.
Beyond these announcements, researchers have tried to evaluate the likely qualitative and quantitative impact of GenAI on jobs. Research teams from Harvard Business School, MIT, Wharton, Warwick and the BCG Henderson Institute, for example, jointly conducted an experiment with more than 750 employees. They observed the impact of using GenAI for 18 so-called “creative” tasks (like coming up with product names, synthesizing documents, writing emails, etc.) and for two complex problem-solving tasks (making a diagnosis and suggesting actions). The results clearly show that undifferentiated deployment of GenAI is not the solution.
While 90% of employees using generative AI solutions improved their performance in “creative” tasks, and the quality of their output increased by an average of 40% (in terms of content and speed), the same was not true for complex problem-solving. For these more complicated tasks, the average performance of employees using GenAI dropped by 23%. Even more concerning, the performance gain on creative tasks comes at the cost of group creativity, with the variety of results decreasing by 41%.
This is indeed the main dilemma for companies: how to deploy the new generative AI to improve performance while maintaining creative capacity? How can companies differentiate themselves if generative AI standardizes offerings, ideas, discourses and interactions? Ultimately, how can a company differentiate itself if all competitors use the same models and the same data?
Nothing is certain at this stage, but there are some hints on how best to deploy the technology. Above all, be wise in choosing what to automate and what not to automate; decide on a task-by-task basis, not by job category.
Next, always consider the role of humans, understand the value they add over a machine and recruit accordingly. If sales agents are assisted by AI to answer technical questions about products, the best salespeople might be those with stronger relationship skills than is required today. Finally, set aside "AI-free" teams to maintain a core group of creatives.
Beyond the impact on specific tasks, GenAI can also help in the shifting labor landscape by bridging skill gaps. Tools like ChatGPT are democratizing job access for individuals with literacy challenges by providing conversational interfaces that simplify complex interactions. AI tutors can smooth and speed certification processes, lowering entry barriers, and aligning with diversity initiatives. Technical jobs with an aging workforce and high qualification barriers, such as elevator maintenance in some countries, are now more accessible to a broader demographic. These advancements position generative AI not only as a tool for productivity but also as a catalyst for a more inclusive workforce.
One thing is certain: the collaboration between humans and the new generative AI systems, harmonious or not, will be a fundamental challenge in the coming decade.