The simplest way to look at artificial intelligence (AI) is perceiving it as a means to improve your business’s service offerings. For example, a software as a service (SaaS) company hires a machine-learning team to improve their music recommendations. This increases revenue. The executive shows the board. The team chalks up a win.
Under this paradigm, it follows logically that AI creation is viewed as a software investment in your financial model. This investment requires skilled people, capital and time for development. It naturally competes against other projects based on expected returns.
As a technology entrepreneur, I have worked with a variety of companies building AI solutions. These solutions applied the latest technologies and solved material business problems. However, I have often felt these businesses could have achieved more if they tweaked their understanding of AI.
Understanding AI
I believe that the software investment understanding of AI is reductionist. It adapts an old way of thinking to a fundamentally new thing. AI affects your business as a factor of production -- land, labor, capital, AI. The good news is that you can acquire AI capabilities for your business. The bad news is that acquiring valuable capabilities is quickly becoming extremely competitive.
AI is competitive and complementary with labor, and it creates efficiencies in capital allocation and expenditure. In the past, when a business applied for a loan, it required people and capital to be approved. Today, at a smart bank, it requires people, capital and AI. AI should be viewed more like a commodity like steel. It has a cost, implementation time and expected useful lifespan. The correct question is “how much AI do we need?,” not “what AI should we build?”
The good news is that AI can improve your business, as a factor of production, without any effort from you. For example, AI enables loans that allocate capital, skill-set matchmaking and new ways to rent office space. The bad news is there’s a secret race going on for AI resource ownership. And in my experience, very few businesses are participating. In the short term, this will result in two classes of enterprise: those that own AI assets and those that do not.
Defining AI Assets
What is an AI asset? The first type is a compounding dataset. This is a dataset that, in conjunction with a machine-learning model, continually yields complementary data. This creates a competitive moat. These datasets are important because they prevent other companies from effectively making similar products. I draw a distinction between general big data and a compounding dataset. The distinction of the latter is that it grows in importance exponentially rather than purely in size.
A second AI asset is specialized AI computing equipment. This is a quickly moving space and companies are making massive investments. Many companies use specialized GPUs (graphics processing units) for computation. They also acquire peripheral equipment, which is important for performance -- fast storage, for example.
Leaders are investing in customized AI chips that will give them a permanent performance edge over their competition. Google’s T(ensor)PUs are a good example. Graphcore is an example of an AI unicorn that develops specialized chips. It has raised $200 million with Microsoft as a strategic investor. Performance is important because it reduces research and development time and broadens which AI models can be put into production.
My last example of an AI asset is AI automation. Aptly named, data “pipelines” transfer data, format it and test models. This is a more efficient way to carry out the AI development process. However, it still requires a fair amount of manual work.
More recently, companies have started to focus in two areas. The first is deployment solutions. These consist of easier ways of getting AI into the field. Deployment solutions standardize scalability and model accessibility. Often, these solutions advertise “one-click” deployment or containerization that makes life easy for machine-learning engineers.
The second area, AutoML, is beginning to be adopted on the cutting edge. AutoML includes model architecture search, automated data cleaning and creation (generative adversarial networks, or GANs), and the aforementioned deployment solutions. These form a cohesive solution that eliminates much menial data science work.
Businesses with these three AI assets have the ability to harness a new competitive edge. This fundamentally changes the business landscape in terms of your capability to offer new services, be smarter than your competitor and be faster than your competitor.
Harnessing AI’s Competitive Edge
Business leaders need to decide whether they want to be, as I call them, “AI leaders,” “AI competitive” or “AI sustained.” Each category requires a different amount of AI investment. For example:
• An “AI leader” strategy might require the company to pay for leading AI talent. The AI talent focuses on a difficult problem domain and publishes their findings to establish authority after a successful solution is found.
• “AI competitive” companies might take a skilled team and leverage modern tools to keep pace in the market.
• An effective “AI sustained” strategy could be to apply an existing engineering team’s skill set to solvable machine-learning problems.
Once the appropriate strategy has been determined, businesses should look first to their core business model and work outward to find the AI assets they wish to acquire. This may seem risky. After all, change to a profitable business line always carries risk. However, what looks like an offensive strategy is actually a defensive one. It prevents the company from being overtaken by high velocity, AI-enabled disruptors.
As a business, ask the following questions to determine your strategy:
• What data drives your decisions?
• How commoditized is that data?
• If you could automate the single most important part of your business, what would happen?
• What if your competitor automated the equivalent part of their business?
• How much capital would you need to accomplish a meaningful AI initiative?
Ultimately, AI assets will enable your business to thrive in the new economy. No business is immune to these fast-moving changes and I encourage all businesses to define who they want to be in the new economy.
