Skip to main content

Generative AI: Boon or bane for the planet

Exploring the applications and risks of generative AI for environmental sustainability

Contact us for further information

The development of generative AI – and foundation models more broadly – has had a significant impact on the accessibility of AI to both businesses and consumers.

Historically, enterprise teams have been constrained by the complexity of AI, requiring extensive use case-specific datasets, specialised expertise in both AI and relevant industry sectors, and a Machine Learning Operations (MLOps) technology stack to manage models and data pipelines.

In contrast, generative AI feels valuable straight out the box, and business leaders are excited by the apparent simplicity of embedding AI into their ways of working.

This apparent simplicity, however, is available to all firms seeking to use generative AI, which begs the question: If everyone can do it, is it possible for organisations to gain a competitive advantage through generative AI?

In this post, we will discuss more precisely how recent developments in AI might impact the level of competition in different markets and the ways in which organisations can generate competitive advantages through generative AI. We will focus on two types of businesses:

  1. Organisations that use generative AI, and
  2. Organisations that develop generative AI models and products.

In doing so, we’ll demonstrate that there remains opportunities for both types of organisation to gain competitive advantages through generative AI, but that these opportunities differ.

Gaining a competitive advantage through tech adoption is hard

For organisations that use generative AI, the natural expectation is that those that quickly adopt the technology will gain lasting competitive advantages.

This could be, for example, through internal operational improvements which allow organisations to run at lower cost, and in doing so either improve their margins or capture market share through passing on price efficiencies to consumers. Alternatively, this could also be through embedding generative AI into existing products to improve customer value, passed on again through improved margins or market share. In both cases though, we’re assuming only one organisation is able to adopt the latest tech.

However, if generative AI is easy to use in commercial applications, everyone in the market will be able to buy and deploy the same models and off-the-shelf products, and in doing so generate the same customer value and operational efficiencies. In this scenario, improved accessibility has actually just lessened the competitive advantage that organisations can derive from AI, and unless there is a drastic mismatch in the speeds at which competitors can adopt the technology, there may not be a massive uprooting of the market.

Of course, this assumes that generative AI is indeed easy deploy and scale effectively within organisations, which is certainly not the case for now. There are currently a myriad of challenges that must be overcome before generative AI can be successfully deployed within commercial organisations, including technical performance, reliability, cost, security and ethical risks. This may enable early adopters to gain more substantial competitive advantages from generative AI in the shorter term.

We expect many of these challenges to lessen over time as the technology and wider generative AI ecosystem matures. However, in both scenarios, there are broader challenges that play a significant role in the value that an organisation can derive from new technologies such as generative AI, including strategy, operating models, internal processes, skills, culture, leadership, and more. These are not specific to generative AI, but relevant to all novel technologies, and are crucial in enabling organisations to build lasting competitive advantages.

This is not just about quick adoption, but also effective and efficient adoption. These lessons are applicable not only to generative AI, but with each new technology which comes to the fore.

First mover advantages for product developers

This competition dynamic doesn’t apply to organisations where AI models are a core aspect of their business. Here, market dynamics are different.

For organisations that develop novel generative AI models and products, and in particular where organisations own their models instead of plugging into Application Programming Interfaces (APIs), there is a lush green field of market opportunity to capture. Organisations that can move quickly are entering into markets that are not fully mature and where there may be significant first mover advantages, although this requires sufficient barriers to protect early movers from future entrants.

One assumption is that the prohibitive cost of training generative AI models will protect early movers. Large language models (LLMs) are not called large for nothing, and existing players have very deep pockets. Although AI organisations rarely speak openly about specific training costs, research suggests that the computational resources required to train some of the more recent cutting edge LLMs cost up to $10 million per model. This is obviously no small change, but it is also not beyond the reach of well-funded start-ups and larger organisations, especially if training costs decrease with cheaper hardware and more efficient training algorithms. Training costs alone therefore don’t appear to be sufficient to protect early movers.

Underlying training datasets are also unlikely to be a sufficient barrier to entry, at least for large foundation models that can be trained on public datasets. The traditional moats of large tech firms owning large amounts of user data don’t seem to apply here, although there is significant uncertainty in this area due to various on-going legal cases.

Perhaps the winning factor though will instead be simple customer inertia and convenience. The way we interact with generative AI today is arguably similar to search engines or smart speakers. The generative AI assistant is the starting point, and so it must be a user’s first instinct whenever they have a relevant question or task. A new competing entrant must have a significant and obvious advantage to break a user’s habit of choosing their default AI assistant, which can be difficult to demonstrate to uses when the relative performances of different generative AI models are hidden behind simple user interfaces. Search engine and smart speaker markets are made up of a few dominant players, and continuing this analogy, we might expect early winners to similarly stay on top.

Where does this leave us?

So, given the excitement around generative AI, just how disruptive will this technology be?

For organisations that develop novel generative AI models and products, there may be a sustained competitive advantage for early movers that results primarily from customer inertia.

For organisations that use generative AI, there still may be benefits in moving quickly and adopting generative AI in the short term whilst challenges persist, but to achieve a lasting competitive advantage, organisations must build sustained and effective approaches to innovation that enable them to not only adapt quickly, but also efficiently and effectively. And not just with generative AI, but with the next new technology too.

If you would like to find out more about the ways in which Deloitte can help your organisation build sustained advantages from new technologies including generative AI, please do get in touch.