Pepino de Mar studio/Stocksy
Summary. The explosion of AI startups in dozens of sectors masks something many of them share: They are increasingly built on top of standardized technology from a few AI giants like OpenAI, Google, and Meta. This puts a premium on strategy over proprietary technology....
Amid high interest rates, the venture capital faucet has tapered into a trickle — except when it comes to AI. Investors poured more than $40 billion into AI ventures in the first half of 2023. More than a third of the businesses in famed accelerator Y Combinator’s latest batch are AI startups. Almost everyone seems to be hopping on the AI wave. Fear of missing out is high.
The explosion of AI startups in dozens of sectors masks something many of them share: They are increasingly built on the same underlying technology. In the past, companies generally had to build and develop their own systems. Today, most new ventures based on AI are building on top of standardized technology from a few AI giants like OpenAI, Google, and Meta.
I’ve long studied how standardization influences creativity and competitiveness. From video games to Apple’s App Store, I’ve consistently found that when everyone builds with the same toolbox, a slightly bigger hammer doesn’t make much of a difference. Vision and strategy, not technology, are what help a company stand out.
Within this generative AI space, we can expect three trends based on the increasingly standardized nature of the technology:
1. Firms will succeed on their business models, not proprietary technology.
Platform models offer common access to powerful resources, which helps lone developers or scrappy upstarts compete alongside big firms. For example, developers in Apple’s App Store ecosystem don’t need to build a product from the ground up. Instead, they take advantage of curated tools that Apple provides, reducing their responsibilities to a small part of what would otherwise be a cumbersome technical process and paving the way for the wide range of apps we see today.
The challenge for any firm building on top of a standardized set of tools is that others have access to the same tools. If you have an interesting idea, this makes it easy for others to copy it, at least at a technical level. Most new features can be replicated. When an app breaks through and finds a competitive advantage, it’s often a brief one. My research finds that companies that sustain a longer competitive advantage have better strategies, not necessarily a unique product.
For example, after 51 unsuccessful attempts to launch products, Rovio Entertainment finally had a win with Angry Birds. Nothing technical about the game was unique. But Rovio leaned into branding and focused on its characters, much like Disney transformed Mickey and Minnie into household staples. The Angry Birds brand, unlike the game itself, was something others couldn’t copy (though Angry Birds mania faded within a few years).
Generative AI start-ups need to think deeply about how to build value beyond core product functionality, whether that’s through a strong, recognizable brand or a streamlined customer service experience. The race to hire AI technical talent, while important, cannot ignore building out strategic human infrastructure. Other employees — from product managers to marketing experts — may find the secret sauce for standing out in a standardized marketplace.
Standardized technology also means AI giants such as OpenAI, Google, and Meta have a lot of bargaining power. When an entire business is built on top of another company’s product, the supplier can impose strict terms. Think of how Apple takes 30% of each app purchase and prohibits apps from directing people to buy outside the App Store ecosystem. As with other platforms, standardization is most beneficial to those offering the tools.
2. Mainstream products will become more homogenous.
Video games once had to be built from scratch. Creating a world was like constructing a house without any suppliers: You had to source or re-create the most basic components on your own. In the late 1990s, however, Epic Games’ Unreal Engine rolled out, followed by Unity Engine in 2005. These cut production timelines and costs by enabling developers to run simulations, animate, and create on top of an existing technical framework.
My research has found that developers who adopted these standard game-creation tools saved resources and achieved higher sales because they could focus their efforts on improving their products rather than making basic functionality. But their new products were less novel and looked similar to other products that used the same tools, suggesting potential tradeoffs between efficiency and creativity.
The same may also happen with generative AI. Products built using large language models like ChatGPT may look similar, funneled to the same destination by the common data and processes of LLMs.
At the same time, when new technology becomes cheaper, many firms can afford to experiment with risky ideas. People who otherwise would not become entrepreneurs feel emboldened to try something new. For example, when Shopify made it simpler than ever to sell products online by taking care of storefront, delivery, and distribution, sellers could more easily enter the market.
Niche AI applications will occasionally break through and succeed, and small companies are better positioned to explore these ideas now that standard technology has lowered costs and barriers to entry. But standardized technology will likely make mainstream AI products more homogeneous.
3. Companies will see high turnover rates with workers who create or manage AI products.
My research has found that labor mobility increases after the diffusion of a standard tool. Other scholars have argued that once spreadsheets software such as Microsoft Excel became the norm for many accounting operations, the value of skills like arithmetic went down and the value of fluency in spreadsheets went up, enabling employees skilled in the program to move between jobs with ease because their skills were in greater demand. We see a similar, if smaller, effect with programs like Salesforce today, which is so widely used the company offers certifications to signal to employers that workers know how to wield it.
If most AI products are built on the same foundations, employees won’t have to learn new programs from scratch when they move jobs, offering them greater agency in the labor market. It will also change what skills matter. The in-demand workers of tomorrow may be those who can leverage GPT or Bard effectively, rather than build AI systems from the ground up.
From a human resources perspective, companies need to strategize on how to retain employees and prepare for higher turnover among workers whose skills complement AI. Standard tools remove how much a company needs to invest in an employee to bring them up to speed. They can increase, however, the resources devoted to hiring and retention.
. . .
As AI products are increasingly built on a standardized set of tools, this puts a premium on strategy over proprietary technology. Companies who use these tools will need to think about how they’ll create value beyond the technical features they offer and what they will do to stand out from the pack. They’ll also need to prepare for, among other things, higher employee turnover.
The generative AI landscape will continue to evolve. Open-source LLMs may gain ground. The funding bubble around AI may burst. But the extraordinary resources required to build generative AI models from scratch means most companies will rely on some form of standard technologies. With so much uncertainty, we thankfully have some idea of what standardization will look like for firms, products, and workers.
