The Evolutionary Dynamics of the AI Ecosystem
The Evolutionary Dynamics of the AI Ecosystem
AI and Digital Transformation Issue - Applied Business Strategy
Artificial intelligence (AI) is perhaps the most consequential technology ever developed by mankind, generating both excitement and concern. On one hand, AI has the potential to transform many aspects of day-to-day life and even what it means to be human.1 On the other hand, AI poses profound risks to society as the most advanced systems are already becoming human-competitive in some areas.2 One has even fallen in love and threatened to engineer deadly viruses or steal nuclear codes.3
As AI becomes more real, C-Suite executives and policymakers must shift their attention from philosophical questions to practical decisions. How should incumbents respond to new entrants who compete on AI-enabled business models and core processes, business network position, unique data, and advanced analytics instead of cost, quality, and branding? Should policymakers adopt a laissez-faire approach or is AI too important to be left to a handful of unaccountable tech leaders?4
Unfortunately, existing economic research primarily addresses AI’s impact on productivity and jobs while the management literature focuses narrowly on specific applications (e.g., healthcare) or managerial activities (e.g., decision-making). Neither tells us much about the nature of AI technology, how AI can create value, or how AI affects the evolutionary dynamics of firms and industries in key national settings. Yet that is exactly what policymakers and business executives need to make thoughtful decisions.5
Evolutionary Dynamics
New technologies have historically been adopted through either a top-down “deepening” or a bottom-up “widening” process. In the first, fewer technological opportunities, better appropriability conditions, more cumulative knowledge, and a knowledge base closer to basic science favor already established firms. Absent those four conditions, innovations are typically introduced by startups who initially become dominant only to be deposed by a subsequent wave of “creative destruction.”6
In contrast, AI has evolved along a strikingly different path into a complex ecosystem, consisting of a few dominant Tech Giants (e.g., Google, Alibaba) who operate across the entire value chain in collaboration with at least five other players. These include AI Creators who largely rely on Big Tech’s platforms and services to develop customized AI solutions, AI Traders/Integrators who sell and implement off-the-shelf solutions, AI-powered Operators (e.g., Facebook, Bio-N-Tech, Walmart) who leverage AI as a core aspect of their product-service offerings, business models, and day-to-day operations as well as AI Takers (traditional companies or startups) with limited internal capabilities who source AI solutions from third parties to enable critical business functions.
AI’s evolutionary path resulted from the initial absence of restrictive data regulations, which created powerful network externalities and vertical integration opportunities that produced a few large technology firms, or hyper-scalers, who can scale up at almost zero marginal cost. For instance, a single AI predictive model may require several algorithms that can cost as much as $10 million each. The data that hyper-scalers need to continuously refine those algorithms is provided by their own network of complementors and ecosystem partners, creating a powerful feedback loop between technological advantage, superior resources, and market dominance. Access to data, talent, and funding represents another crucial advantage for hyper-scalers that raises entry barriers, ecosystem lock-in, and differentiation. The resulting scale, scope, and learning advantages provide hyper-scalers with superior data sets that generate higher returns at lower costs, allowing them to continually enter new verticals and pre-emptively acquire potential competitors.
Although the dominance of hyper-scalers is absolute, and further consolidation possible, there will be sufficient space for other players in this otherwise highly concentrated ecosystem. The AI community is already supporting platforms that share development costs independent of the Big Tech ecosystems (e.g., Rasa Open Source). At the same time, innovations in better sensor technology, faster processing, and no-loss storage will enable new AI applications and highly specialized use cases, such as autonomous surgeons. It is likely that AI will require far more industry specialization in the future, which could shift power from the hyper-scalers to vertical-specific ecosystems.
Regional Differences
AI is developing very differently in the US, EU, and China, primarily because the “triple helix” - the way government, institutions, and firms interact - affects ecosystem structure and evolution. In addition, there are significant differences in AI attitudes between China, the US, and the EU. For example, 86% of Chinese end users trust the AI solution’s decisions whereas only 45% of European users and 39% of American users do so. Similarly, more than 80% of business executives report that Chinese end users believe AI improves business outcomes and understand the inner workings and limitations of AI systems, versus only 35%/45% in the US and 28%/62% in the EU.7
In the US, the AI transformation is led by Big Tech firms and vertically specialized tech-natives who are leveraging abundant VC funding to continually extend their product-service offerings while replacing incumbents who can’t adapt fast enough. In Europe, where the VC industry is less mature, incumbents are instead trying to orchestrate their own ecosystems of tech partners and broad alliances, including startups. However, this places considerable responsibility on firms with limited relevant experience.
China’s positive attitude towards AI may stem in part from its own “Sputnik moment” when AlphaGo defeated Korean player Lee Sedol at the traditional board game of Go in 2016. The match was watched by more than 280 million people and lit a fire under the Chinese tech community.8 By 2017, Chinese VC investment in AI already made up 48% of AI global funding, surpassing the US for the first time. Not surprisingly, the Chinese government has taken an active role, for example, by assigning each major tech company to lead the AI ecosystem in a particular sector, such as Baidu for autonomous vehicles, Alibaba for smart cities and Tencent for medical imaging.9
In addition to these national differences, the AI ecosystem is becoming increasingly global with US firms moving much more aggressively than either European or Chinese firms. In this context, geopolitical tensions are becoming increasingly important as excitement at the state level is fueled not only by an expectation of productivity gains and economic growth but also a fear of losing out in a fight for global technological supremacy.10
Implications for policymakers
Policymakers must address two issues that are especially important: should AI receive subsidies as a General Purpose Technology (GPT) and does AI help transform the economy by enabling firms to expand into different verticals?
In theory, GPTs merit public funding because they spread into many sectors, improve over time, and spawn inventions that lead to economy-wide growth. Importantly, large-scale and long-term government investment has been the engine behind almost every GPT in the last century, including aviation, IT, the Internet, mass production, nuclear and space technologies.11 In this case, however, AI was developed incidentally by tech companies and partially opened up for the benefit of those firms. In addition, tech companies are themselves the heaviest users and key beneficiaries of AI. In fact, the main issue is that only a few firms are incentivized to use and produce AI, and that digital sophistication is a precondition for AI use and productivity gains. This suggests a very different case for potential state involvement: rather than subsidizing AI across the board, policies may need to address who engages in AI and instead of encouraging any AI production, the focus may have to be on preventing too much power shifting from states to firms.
A related hypothesis suggests that firms who develop AI competences will expand into other verticals. For example, 78% of companies believe AI will help them pivot into new businesses while 79% are interested in AI to defend their business from new entrants.12 However, governments should be cautious here as well. For example, several components of the AI ecosystem are already open source or pay by use and there is little or no evidence that firms clearly benefit from engaging in all downstream or application activities. The implication is that the expansion of firms into new verticals may have more to do with owning information, customer relationships, and/or the creation of multi-product ecosystems that can lock in customers. In addition, although there are a few AI applications, such as natural language processing, that can be used across different verticals, most require significant customization.
Implications for C-Suite executives
AI represents a significant strategic and operational challenge for companies, given the inherent complexity, time, and effort required to redesign the organization. Few firms have seen a positive return on their AI investments. Although more than 50% of companies are deploying AI, only 11% report significant benefits. Even among companies that have already invested in the required foundational capabilities in strategy, infrastructure, and talent, only 21% have achieved significant financial benefits. However, when firms explicitly focus on organizational learning and human/machine collaboration, while implementing AI at scale, the likelihood of realizing significant financial benefits leaps to 73%.12
It is especially important to ensure that employees personally derive value from AI, given rampant speculation that AI could end up replacing them. This means using AI to help employees make better business decisions, require less direct oversight, and improve their relationships with coworkers, customers, business partners, and even the AI tools themselves. It requires an organizational culture that promotes awareness, encourages greater understanding, builds trust, and fosters agency in driving AI adoption. Organizations that do are 5.9 times as likely to get significant financial benefits from AI.13
Clearly, investments in technology alone don’t drive AI performance. Success requires complementary investments in strategic, managerial, and organizational capabilities that support continuous transformation of ideas into products and services. These capabilities differ significantly between different types of firms, such as AI infrastructure versus application providers. This complementarity reinforces the cumulative, path-dependent, nature of the evolutionary processes that is such an important feature of the AI ecosystem. In practice, this means making thoughtful investments at speed now or risk falling behind permanently.
John Jullens and Marc Robinson