From Siri and Alexa becoming household names to Apple recently announcing its new HomePod smart speaker, artificial intelligence finally seems to have made the jump from the world of futuristic science fiction straight into our own homes. The variety of tasks that AI creations are taking on is increasing. AI can vacuum our floors, create expensive works of art, and even debate the meaning of life.
With AI technology permeating the public consciousness, tech companies are eager to ride the hype and emphasize the AI aspects of their products. Everyone is claiming to be an AI company, showcasing their neural networks or their bots. But with a narrow and specialized focus, those companies are often missing the big picture. The companies that will rise to the top and stay there are full-stack AI companies.
AI is a holistic process
To build a meaningful AI system, you have to have a big data practice, a software engineering practice, and a user experience design practice. Only then can your AI evolve into a useful and practical tool that can communicate with other applications. You also need to create a feedback loop to determine whether the decisions that your AI system is making are actually helpful to the people using it.
I saw this for myself when I started my career as a knowledge engineer, focused on creating the backend rules and algorithms that give AI systems the ability to “think.” I quickly realized that in order to deliver results, knowledge engineering isn’t enough: I also needed to understand software engineering, the front-facing aspects of the trade that made our AI systems useful and usable to customers. If I wanted to be effective, I needed mastery over the complete, end-to-end solution.
Companies looking to develop successful AI will also need to master every level of the solution: big data, analytics, and user experience (UX). A bot or an algorithm can’t and shouldn’t just stand on its own.
Trend-driven AI misses the big picture
The market has witnessed enthusiasm for the AI trend before, and we can learn from it. Prior to 1990, venture funding poured into artificial intelligence. As early as 1986, university researchers in Munich successfully tested self-driving vans.
In the 1990s, though, the fickle market grew tired of AI. The bubble for AI products burst. Even mentioning the word in a company’s description would hurt a startup’s valuation. That doesn’t mean companies stopped working on AI advancements — they just stopped calling it that. They replaced the terminology with catchphrases like “big data” or “intelligent algorithms” that were still pieces of the AI stack but didn’t take into consideration the full picture.
This focus on just one facet of the science puts companies at risk of missing AI’s larger potential. In business, this translates to missed opportunities to build a lasting product, rather than a trend-driven one.
Moving forward with a full-stack approach
For artificial intelligence to truly become mainstream, it needs to work with and fit into the universe that people already live in. This is rough news for industry leaders who place AI at the center of its own universe, envisioning it to be a standalone source of light around which everything else must revolve. These companies tend to view a bot or an algorithm as the end game, not realizing that these products on their own are not actually solving problems.
People are connected to technology like never before, with end-points from our wrists to our thermostats to industrial assembly lines. That means AI finally has the network it needs to reach its full potential — if developers can successfully integrate it in order to improve and streamline the devices we are already using.
Until recently, only the largest companies and institutions had the technological capabilities to work with and analyze big data by reading, interpreting, and sending huge amounts of quality information. Today, though, cloud computing and advanced computing power make working with large datasets easier, more affordable, and more accessible.
Additionally, working on the user experience aspect of artificial intelligence has recently become within reach, given the proliferation of API building blocks for user-friendly packages like responsive mobile apps. Previously, these would have to be coded from scratch.
Why full-stack companies are still rare
While recent progress serves to level the playing field somewhat, it doesn’t mean an AI stack is easy to build.
Data, in particular, proves to be difficult for many companies hoping to join the full-stack club. It’s hard to acquire, it’s messy, it’s inaccurate, and it’s incomplete. Analysts can look at pools of data and come up with correlations and insights, but knowing how those insights apply to real-life problems is yet another part of the stack: user experience (UX). AI without UX design as a core competency will only have a tiny impact — namely, it will answer the question of “what does the data say?” but not “so now what do we do?”
Full stack is the way of the future
As the industry works to integrate artificial intelligence into every aspect of our daily technology use, a full-stack AI company will be best positioned to thrive. This is because companies that deliver a single AI component will become commodities for end users.
In contrast, full-stack companies will have the knowledge to see how their product fits into a complete solution. They will also be able to take advantage economically of the vertical integration that comes along with that knowledge.
John Price is the chief executive officer at Vast, a company that provides a platform to support big data experiences for automotive and real estate.