While Artificial Intelligence industry is undergoing a rapid development today, it also remains a disorderly and complex field where value creation still poses a significant challenge for both executives and investors.
In general, sustained progress in an emerging field can be made viable only by establishing a strong and diverse ecosystem of players that are motivated to continue to push forward and deliver value to their partners and customers even during the periods of cooldowns. To quote Ron Adner of Tuck School of Business, “you’re only as good as your ecosystem” (HBR, 2016).
Over the years, a number of different instruments have been developed to frame the thinking and help identify the existing gaps and investment opportunities in a particular industry, ranging from Porter’s five forces to Gartner’s hype cycle for emerging industries. However, the ever-accelerating pace of the world today forces much faster business decision making, which in turn creates a need among business leaders, entrepreneurs, and investors for a set of tools capable of delivering actionable insights.
To address this need, we’ve developed a framework allowing to evaluate the level of ecosystem maturity of any emerging technology field and helping to identify key growth catalysts and blockers.
Conceptually, any technology space can be visualized as a series of “building blocks”: it begins with the areas that are farthest from the end customers, and then continues to the areas that can be regarded as downstream, until we get to products and services that are being directly purchased by the end consumers (end products, or implemented solutions):
As to be expected, most spaces typically begin with the research stage
It then progresses towards an area encompassing hardware that enables compute or other essential functionality, and then the basic “fundamental pieces” of software (such as frameworks like TensorFlow in case of AI, or virtualization for cloud computing) allowing for more efficient creation of products at the subsequent stages
From there, it separates into two pillars: the first includes end products that are made to be sold off-the-shelf (many, but not all, of those, would be B2C), while the second consists of platforms, solutions built on top of those platforms, and finally fine-tuned versions of those solutions that are then being implemented (this pillar is predominantly B2B)
Across all stages, there are 3 broad categories of commercial players involved: foundational product companies, specialized product players, and “implementers”:
It’s worth noting that precise definitions of “foundational” vs. “specialized” are somewhat fluid and are likely to evolve with the ecosystem: in the earliest days of a particular technology space, it is quite possible that there would be no foundational players until some of the startups or large technology companies don’t evolve to fill that role
Similarly, at the beginning most specialized players are likely to be startups, but this will likely change as the industry matures, with some specialized players eventually turning themselves into foundational companies, while the rest will instead carve out specific niches for themselves and stay in those or expand to some adjacent markets
Finally, sometimes the players of one type can try to fill the role of the other(-s) – e.g. specialized product companies might get involved with implementation, or the implementers might try to develop their own platforms; in the long term, though, such efforts would be unsustainable, unless the company pivots to become a different type of player
The level of maturity of each “building block” influences the overall level of maturity of the ecosystem, while its ability to progress towards maturity is defined by a combination of two indicators: the alignment of expectations across stages and the availability of external funding at each stage:
The idea is quite simple: the alignment of expectations allows for the creation of sustainable businesses at each stage, while the availability of external funding can allow for much-needed experimentation and building of all necessary components at the early stages
In the early days of any ecosystem, the alignment of expectations across at least some of the stages is likely to be imperfect, in which case the development of the ecosystem has to be supported through external funding until such alignment improves
Conversely, for more mature ecosystems, external funding becomes less critical, as by then a good alignment of expectations across all stages should allow the businesses to become increasingly self-sustainable
We’ve studied a number of approaches both the largest technology companies and startups take to build their businesses in AI space. Combining those examples with a number of other sources (venture capital investment data, industry analysts’ reports, expert opinions, etc.), and analyzing those through the lens of ecosystem maturity framework, we’ve reached the following conclusions:
At upstream stages, the industry today enjoys a good level of alignment of expectations, and the same is largely true for the off-the-shelf products
However, AI ecosystem remains fairly immature on the B2B side, especially when it comes to downstream providers building specific solutions and implementing those
The key reason for the relative immaturity of the ecosystem at the implementation stage today lies in the lack of alignment of customer expectations
For business executives looking to leverage AI technologies in their current businesses or build AI-infused products, it is crucial today to focus on the areas that represent the most substantial blockers on the path of overall ecosystem maturity, even if those areas don’t always appear to be the most attractive ones.
Moreover, as it is the investors who support the formation of the ecosystem with their money until it could allow for sustainable business models to emerge, it is in their interest to incentivize the management to prioritize the opportunities that might bring the ecosystem to maturity faster.
Sample proof points
AI ecosystem overview:
The number of AI papers published has increased more then eight-fold since 1996 (3 times the rate of growth for research papers in general, and 1.5x that of CS papers); in some subfields (namely, machine learning, neural networks & computer vision) the growth in number of papers published proved to be even steeper
More generally, the number of jobs in AI has been steadily rising, with the number of job postings increasing more than 2.5 times in 3 years, according to Indeed.com, while the number candidates failed to keep up so far (potentially exacerbating talent shortages going forward)
AI frameworks, and TensorFlow in particular, have experienced some of the fastest growth in adoption for new technologies, and are almost universally loved by developers
Current state of startups’ ecosystem:
The investment in AI space has been ramping up for years now; in the U.S., it has grown from $1,147 million invested across 207 deals in 2013 to $9,334 million across 466 deals in 2018
Moreover, the slowdown in the number of deals in 2018 (466 deals vs. 533 deals in 2017), coupled with the continued growth in overall investments (thanks to larger late-stage rounds) suggests the industry should be nearing maturity, at least from the investment perspective
The number of acquisitions has also increased more than five-fold since 2013 (22 acquisitions in 2013 vs. 115 in 2017)
However, we are seeing a lack of large exits, despite significant investments in the space: as of February 2018, there were only 2 exits at a valuation higher than $1 billion (Roche Holding acquiring Flatiron Health for $1.9B in February 2018, and Ford buying Argo AI in 2017), and neither has there been any substantial number of public offerings among AI startups so far
The list of the most prolific acquirers in the U.S. consists almost exclusively of big tech firms, which is quite consistent with the lack of large acquisitions or IPOs so far: most of the acquirers are purchasing these startups at a fairly early stage, and thus are buying them for their founding teams, or products, rather than acquiring already established businesses
Broader AI adoption:
However, according to another report from PwC, only 4 percent of the surveyed technology executives in 60 countries had successfully implemented AI, with many experiencing challenges in implementing AI within their organizations
Security issues, privacy concerns, lack of appropriate skills, and lack of understanding among employees about the technologies being adopted are commonly cited as key barriers to successful adoption of AI
It’s also telling that, according to Gartner, when asked about the expectations around AI implementation, most CIOs today tend to emphasize cost optimization and efficiency improvements, with only 4% focusing on using AI to gain additional business/capture new customers (which typically is a more powerful lever to drive technology adoption).