We return to the AI theme, this time focusing on generative AI

Less than five years ago, we wrote on the disruption that firms would likely experience as AI matured and adoption grew. Back then, 29 sector research teams at UBS identified key AI-related opportunities and threats that were most relevant to their sectors. In light of the rapid adoption of generative AI, we asked our sector teams to revisit their views.

Identifying the sectors that will be most impacted by generative AI

In this Q-Series report, focusing exclusively on generative AI, we summarise the observations of UBS sector analysts on:

  1. adoption trends of generative AI in their coverage sectors;
  2. the companies they believe may be better or less well-placed to utilise the technology; and
  3. create a framework for investors to understand different sectors' relative exposures to the potential risks and opportunities posed.

UBS Evidence Lab found that on earnings calls since January this year, nearly 500 companies in 27 sectors have made over 3,500 references to generative AI, notably in Software & Services, Media, Commercial Services, Semis and Consumer Services (leisure, gaming and education services). Generative AI is set to automate or augment the roles of many knowledge workers. In all these sectors bar Semis, employee costs are high as a proportion of sales and a high proportion of employees work in roles prone to augmentation or automation from generative AI.

Assessing the potential impact of generative AI

To assist the sector teams in assessing the potential impact of generative AI, we provided them with:

  • Occupation profiles: Utilising US Bureau of Labor data on employment by industry, we considered which sectors have the highest proportion of occupations which might be exposed to automation – or augmentation through the use of generative AI. We found that language-based, creative, and computational activity is likely to be most affected.
  • Labour intensity: We also considered how labour-intensive each sector is. We used data from the largest European companies in a given sector as a proxy for the industry as a whole, as this is where data is most readily available. In total, we gathered data on 290 companies with a combined market capitalisation of $11.6tn and a total workforce of 17.8m employees. We found that industries where employee costs are significant as a proportion of revenue have more to gain from efficiencies related to automation/augmentation than those where labour costs are not significant.
  • Starting efficiency levels: We considered how profitable and capital-intensive each sector is relative to others. We found that industries with low margins or high capital intensity have more to gain from increased efficiencies related to exploiting generative AI, relative to those where the opposite holds.

Cost savings are seen as the most likely outcome from generative AI advances...

While the narrower scope of our new sectoral analysis relative to 2018 might imply a lesser impact, our analysts' bottom-up feedback suggests that many see a significant effect, with 18 of the 31 sectors and sub-sectors covered anticipating that generative AI will increase revenue. And while all but the Internet team believe that generative AI will help to reduce costs, 17 indicated it would potentially increase competition too; 10 teams see the potential for both higher revenues and cost savings, without increased risk of competition.

Open-source generative AI models are maturing rapidly and may lower the cost of adoption

We believe the barriers to organisations adopting generative AI models may fall over time. The rapid pace of progress with open source models suggests that barriers to the adoption of generative AI models will come down in time, although at least as of today the costs of developing and deploying these models can run to millions of dollars, or even more for the very largest ones, albeit the cost of running these models - once trained - in corporate environments may be more modest.

Adoption plans

UBS Evidence Lab has analysed what listed companies have been saying about AI in general and generative AI in particular, and which sectors are talking about it the most, this being a potential indicator of where the greatest effect might occur. UBS Evidence Lab searched for references based on various forms of "pure" and "applied" artificial intelligence (e.g. autonomous vehicles, virtual agents) since January 2015. The review encompassed over 300,000 transcripts in total, finding over 13,000 references each to machine learning and artificial intelligence, and over 37,000 to all forms of AI. Generative AI has generated over 3,600 references by over 500 companies in 27 sectors since the start of 2023.

What is Generative AI?

Generative AI is a type of artificial intelligence that can create various types of content, including text, image, video, audio, code, 3D models and synthetic data. It can create new and original content on demand, rather than simply analysing or classifying existing data. Generative AI models are trained on a vast quantity of unlabelled data and leverage different learning approaches, including unsupervised or semi-supervised learning, and use neural networks to identify the patterns and structures within existing data. The advent of Transformer neural networks in 2017 supercharged the pace of development in generative AI and large language models.

The benefits of generative AI include generating new and unique content and ideas, as well as reducing costs and increasing efficiency and productivity by accelerating manual and repetitive tasks, and personalising experiences by tailoring content to a specific audience. The shortcomings of generative AI include that there is no guarantee the output is accurate, sometimes models provide information that sounds plausible but is not true (known as hallucination), and the output may also infringe copyright, include biases, and violate data privacy.

Explore other articles you may find interesting