2023 is shaping up to be a pivotal year for artificial intelligence, with the rise of generative AI poised to transform industries and reshape workforces. The latest McKinsey Global Survey on AI confirms the increasing market penetration of tools like DALL-E and ChatGPT and outlines how workers and employers see the technology affecting work patterns in the coming years. 

One-third of survey respondents say their organizations are using generative AI regularly in at least one business function. Respondents across regions, industries, and seniority levels say they are already incorporating generative AI tools into their business.

Marketing, sales, product development, and customer service operations are the top functions using generative AI. Common use cases include drafting text, personalized marketing, identifying trends in customer needs, and using customer service chatbots.

The organizations achieving the most value from AI—what the survey calls “AI high performers”—are using generative AI in product development and supply chain management. These leading companies are also twice as likely to see generative AI’s top benefit as creating new revenue streams rather than reducing costs.

Workers and organizations expect generative AI to change work. Three-quarters of respondents expect AI to cause significant or disruptive change in their industries within three years, with tech and financial services companies predicting the most change. Disruption, however, does not necessarily imply displacement. As shown below, the disruptive impact is much more likely to manifest as reskilling than in job loss. Nearly four in 10 firms expect to reskill over 20 percent of employees in the next three years to help current workers adapt to AI. The Forbes Technology Council recently identified 20 new business roles—from prompt engineering to AI ethics to AI output verification—that will add to business headcounts.

The survey also suggests that many organizations are unprepared for the widespread use of generative AI. Since McKinsey’s previous survey, the same 55 percent or so of organizations have adopted AI, typically in just one to two functions like product development and customer service. And just 23 percent of respondents say that at least 5 percent of their organizations’ earnings last year is attributable to their use of AI—essentially the same as the previous survey. These results suggest that AI use remains limited in scope and there is significant room to capture value through improved efficiency, productivity, and labor substitution.

In a couple areas, AI implementation is experiencing significant lags. Though hiring for AI-related roles has become somewhat easier in the past year, the demand for data scientists, data engineers, and machine learning engineers continues to outstrip supply, and talent development and recruitment remains challenging.

Lastly, just 21 percent of firms have established formal policies for governing employee use of generative AI, creating something of a Wild West of AI use in offices. As the chart below shows, many organizations are worried about AI risks, such as generating inaccurate information, cybersecurity, and intellectual property infringement, but they are largely unprepared to address them. This pattern will undoubtedly lead to some misuse and challenges that we can expect to hear about in (sometimes) amusing and (usually) breathless media coverage of the always-forthcoming-never-quite-here AI apocalypse.

The overall message of this and other recent reports is three-fold. It remains very early days for generative AI; the big effects are still over the horizon. Second, contrary to some of the worst fears about employment impacts, in the short- and medium-term, the biggest effects are likely to be felt in upskilling rather than in worker displacements. Finally, the most important longer-term impact seems to be that AI will create more, higher-value, and higher-skill job opportunities than it destroys. It’s not an easy path but with proper support from employers and government, the benefits of AI are likely to significantly outweigh its costs.