AI business success can be enhanced through collaboration. The five pillars of implementation and collaboration describe the fundamental elements that any business in this space should be considering. Recognizing that mastery of all five components can be daunting, this article seeks to highlight the point that not each is a company’s strength; working with ecosystem partners to piece the puzzle together is important. Of course, while this approach can be applicable in other industries, the AI field is particularly interesting because it is so diverse. Companies can generally work together without a risk to their key markets. This becomes evident when looking at the vast array of players developing AI technology and innovative applications – everyone from Apple and Microsoft to John Deere and RR Donnelley.
From an academic perspective, artificial intelligence technology is uncovering amazing opportunities for research and advancement. It is equally exciting to see the fruits of this research come to life. In an academic or pure research environment, intellectually stimulating hypotheses and click-worthy advancements reign supreme. When it comes to operating a business, with goals focused on cost reduction, growth, revenue, and return on investment in technology, other variables quickly come into play. It is not terribly complicated, but there are five major components that factor into a business in the AI space:
- Data (the raw materials)
- Technology/talent (to process the raw materials)
- Intellectual property (to protect the technology)
- Capital (to fund the work)
- Collaborative design and delivery (the implementation)
The five pieces of the modern AI puzzle and what is needed for successful implementation
Anyone even dabbling in the AI world knows that data lies at the heart of artificial intelligence technology. One of the reasons that AI is back on the main stage is because our connected online lives have allowed us to collectively amass massive amounts of organized data. Now the machines can learn from this collected historical data and make meaningful predictions or pattern discovery.
Of course, data isn’t everything. It is best to think of data as the raw material that is necessary to make the end products. There are other key pieces to success in the AI field. The data (the raw materials) must be processed – that takes technology and talented scientists and engineers. The data needs to be organized, tagged and categorized. Much of the historical enterprise data available is not collected with the goal of applying AI technologies in the future. As a result, in a typical AI project, about 70% to 80% of the time is spent on cleaning and preparing the data that can then be fed into AI algorithms. The key to AI success is to carefully account for the budget and time required for data preparation activities. Failing to do this is a recipe for “garbage-in, garbage-out” scenario. Only once this is done can models be effectively built to consume the data.
Data is important but it is not the end of the story, rather the beginning which can decide the ending.
Technology and talent go hand in hand. AI is a highly technical field. Having access to technology and the talent to develop and innovate is critical to success. And there is more to this picture than just number crunching or processing data. How does data get into the system – what are the delivery and data handling technologies? How do results get delivered and how are those results used in a business process or end product? How are models adapted to market conditions or user experience feedback? The answers to these questions often require some significant legwork and innovative structuring.
The ability to move and process the raw material – the data – with new and innovative techniques provides competitive advantages in the market. In this respect, talent resources cover not only the technical development, but the creative adaptiveness to rework, retrain and redeploy models to fit market needs. The talent also provides a mechanism to adapt with agility to unexpected or newly discovered market conditions. When this agility is combined with technology that can be worked and adapted by its operators, momentum can be gained quickly.
Without a doubt, the technical and implementation innovation created with these elements adds tremendous value for a company. To maintain these advantages and protect key advancements, intellectual property is critical.
Due to the technical nature of the field and applications of AI, intellectual property (IP) protection and strategy play a central role. IP rights assign ownership of a technology to a person or company. Put simply, IP is important because the technology is important. Without protection around a new technology, other companies (who may or may not be competitors) can freely copy. IP, and patents specifically, provide value because they create limited monopolies – allowing one company to restrict others from doing their invention. But beyond that, they are also a mechanism for allocating transferable value to a technology. This is important. If I have a patent, I can share that technology with another for a fee. If I don’t, copying may be perfectly fine and legal (although unfortunate for the technology creator). And that is why it is important to understand the landscape of patents filed across the world – what is protected and where. The “where” component is important to pay attention to because patents are inherently geographic. A patent in the U.S. prevents others from copying in the U.S. A patent in China prevents copying in China. If I don’t file my patent application in Japan, for example, someone else can recreate my invention in Japan and I have no recourse. Figure 1 below illustrates the concentration of AI patent applications across the world. The most popular jurisdictions are the United States, China, Japan, Germany and South Korea (in that order).
IP protects technology and secures the value of that technology for its creator. Employing rights to protect this transferable value in technology aids in the generation or raising of capital – the next critical component.
This is probably the least intellectually complicated of the components. As we know, capital is the grease that allows the machines to move. Capital is the money that funds the research, product development, marketing, manufacturing, and everything else to get a product in the hands of users. Capital is the simplest because there are well established mechanisms for determining value. In fact, most other components are measured by capital value. The critical point with this puzzle piece is its availability – where does the funding come from, how free-flowing is capital to allow a company and its technologies to come to fruition. The best ideas have trouble getting off the ground without money to back them.
Collaborative design and delivery
Finally, all of these pieces are tied together with a collaborative infrastructure on the human, physical and system fronts. The implementation matters. All of the above pieces can exist, but without the right collaborative design and delivery implementation, it will all fall apart.
Recent research by IBM Cognitive systems and inc.digital into the DNA of successful AI implementations showed that 50%+ of all AI projects were measured as failures by their owners. This is not because of the hardware, the software, the intellectual capital or even the type of AI projects being done (from machine to deep learning, visual AI or natural language AI). It is simply because teams could not find the right measured pathway to collaborative design, delivery and value through the enterprise’s ecosystem. AI projects that sit in data science departments only or under a data scientist’s desk have a 75% chance of falling into the half of AI projects that fail even in the minds of those managing the projects. The power of AI to perform most frequently occurs on high and varied works loads (doing all four types of AI) and not just on one or two AI disciplines. There is an inherent multiplier, defined by how well you focus on collaborative design and delivery, that meant 15% of the enterprises that did this the right way generated some 47% of the total ROI (25 metrics) achieved by all respondent organizations in the research.
First things first, it is important to recognize that a company does not have to be an AI-only company to be an AI company. Value from AI innovation is being realized by the operating companies putting that technology to use. In today’s economy, many companies are finding that they must develop, build, train and otherwise innovate in AI technologies to compete in their industry. If it is not already the new norm in your industry, it won’t be long; change is coming. It is inevitable with the gains that are made possible through use of these technologies. Let’s look at John Deere, for example. John Deere is not an AI-centric tech company. One does not typically think of them next to NVIDIA, IBM or Tencent. They are, however, leading the way in developing and implementing AI technologies in the equipment they manufacture. In 2017, John Deere acquired Blue River to bring advanced computer vision technology in house to enhance their machinery.
It comes as no surprise that not every company implementing AI technologies has strength in each of the five pillars mentioned above. IP plays an important role by opening the doors to collaboration between companies. It provides a mechanism for trade of otherwise nebulous or intangible assets. Data can be shared, talent can be seconded and developed technology can be licensed or sold. Intellectual property rights in the form of patents, copyrights and confidentiality agreements provide the basis for protection and lay a path with appropriate guardrails for collaboration. With this mechanism for sharing available, companies should be looking inward and outward to determine how to approach AI implementations holistically.
As companies look to develop AI strategies within their organizations, understanding these five pillars is increasingly important. The key is to recognize strengths and weaknesses, capitalizing on the strengths and filling in the gaps on the weaknesses. Artificial intelligence technologies introduce infinitely smarter capabilities when compared to humans in many ways. Technically speaking, this is huge, but there are challenges to successful implementation. The technology can be rigid in its application and gains can drop off quickly as needs arise to understand non-rules-based behaviors. Flexibility and adaptiveness are provided by in-market human interactions. Real-world feedback, creative application of the technology and interpretation of user experience is needed more and more to bolster the potential of this technology. Most of the best AI comes from this tug-of-war. Again, recognizing capabilities or deficiencies become important to competitiveness. Some organizations and even cultural practices are less well suited to applying technical innovation creatively in the market. Approaches that focus on compliance and volume behaviors versus tensions and contradictions will struggle in a global market even where the underlying technology may be groundbreaking.
These ecosystem characteristics set things up nicely for a collaboration, cross-licensing and idea-sharing – leveraging strengths and fortifying weaknesses. In addition, it’s the implementation, application and continued fine-tuning that makes a difference. This is the old commodity versus engineering argument from the industrial revolution. The U.S. has more open access to markets, wider industrial base and a stronger focus on digital transformation. Thus, it’s the puzzle piece and components, the integration and interoperability through systems and ecosystem partners and applying it to a real-world problem where the value is created. The value created still requires a tremendous amount of thought into the customer experience or process workflow focusing on customer moments and reducing friction by the better, faster, cheaper principles. Above-mentioned is enabling an organization to harness the power of AI, but also note AIaaS is being embedded in mini-work processes en masse horizontally throughout the enterprise as well, from calendaring to chatbots to machine-learning models.
About the authors
Michael Gale is the chief executive officer of inc.digital, board member, co-author of The Wall Street Journal bestseller The Digital Helix, AI influencer, and podcast host about the world in 2030.
Thomas Marlow is the chief technology officer at Black Hills IP. As a patent attorney and technologist, he is a published author, recognized speaker and expert in the field of intellectual property and technology.
Manjeet Rege, PhD, is the director of the Center for Applied Artificial Intelligence and an associate professor of Graduate Programs in Software and Data Science at the University of St. Thomas. Rege is an author, mentor, thought leader, and a frequent public speaker on big data, machine learning, and AI technologies.
Dan Yarmoluk is an adjunct faculty at the University of St. Thomas, Graduate Programs in Software Engineering, Data Science and member of the Center for Applied Artificial Intelligence. He is a recognized author, entrepreneur and IoT, data science and automation expert.