Recently, McKinsey, Harvard Business Review, O’Reilly, and Deloitte have conducted multiple studies about barriers to the adoption of AI in the enterprise. The common thread among them has been that most businesses already use AI technology in at least one of their business processes. However, a vast potential for creating value through the application of AI largely remains untapped, which includes removing organizational silos at the data level and using AI to maximize the output value of core business processes such as sales, procurement, or human resources management. All surveys combined produce an impressive list of barriers to the adoption of AI, but the two most frequent barriers are: the lack of useful data infrastructure and the lack of talent to manage and analyze the data.
The challenge organizations face today is to make all business process data available in one centralized location where it can be accessed, cleaned, normalized, and analyzed by AI for decision making. Most organizations that have undertaken such projects have spent about a year with the preparation of data before it can be used with AI. Clearly, these are costly projects, which highlights the significant value these organizations expect from AI implementations. But this is only the beginning. The next challenge to successful deployment of AI is to create a continuous stream of data that the AI can use for online decision making–only then it is possible to apply the AI’s output to the management of its business process and maximize the AI’s value. Organizations typically face two obstacles. First, they need to put the technical infrastructure in place for access to all IT systems associated with the business process, which enables the continuous cleaning and normalization of data. Second, they need to overcome organizational silos that often exist between functions. It is therefore important that large AI projects have sponsorship from top-level management, sufficient funding, and a time budget of at least one year.
Most organizations also attribute the lack of talent as a major obstacle to the adoption of AI. This challenge isn’t about finding machine learning engineers or data scientists as much as it is about the lack of understanding inside the organization of how AI works and securing resources to hiring necessary personnel. Modern AI is based on algorithms that develop and optimize statistical models about available data and then make predictions about new data. This means that AI works with probabilities rather than with guaranteed correct output and depends to a high degree on the quality of data it is fed. For most organizations, this behavior is a long way from the deterministic processes they currently operate. It is understandably difficult to transition to a system that makes decisions with varying probabilities of correctness and that can be manipulated by feeding it biased or incorrect data. In addition, for the AI to be most effective it must be able to dynamically adjust each step of the business process. This often is a long way from current process operation procedures that are largely based on hierarchical approval workflows. It is important not only to calculate the potential cost savings or additional revenues during the planning phase of the AI project, but to also evaluate the cost of change in the organization where software is about to enter into the space of managerial decision making.
Most businesses that have successfully implemented AI projects achieve a significant to moderate value gain. With the value and competitive advantage that comes along with it, AI is clearly worth the effort to overcome these barriers of AI adoption in the enterprise.
About the Author
With over 17 years of product development experience, Christian Thun leads engineering at Agiloft, provider of AI-powered contract lifecycle management (CLM) software. Prior to joining Agiloft, Christian worked on complex projects at startups and international corporations in the global IT, cloud computing, and telecommunications markets. Christian specializes in driving strategy and bringing innovations to market. He holds an MSc in International Strategic Marketing from the University of Reading.
Sign up for the free insideBIGDATA newsletter.