Artificial intelligence could contribute multiple trillions to the global economy in the coming decade, but even while poor data quality becomes increasingly detrimental to businesses’ bottom line, AI is not yet being widely incorporated. Heidi Maher of CGOC discusses the widespread need for AI and machine learning.
According to a recent study by PWC, artificial intelligence (AI) could contribute up to $15.7 trillion to the global economy in 2030, more than the current output of China and India combined. This contribution was calculated based on increased labor productivity and enhancements to products that will drive consumer demand. It’s no wonder that venture capital funding of AI companies soared 72 percent last year, hitting a record $9.3 billion.
For a company, the potential of AI and machine learning (ML) initiatives to improve operational decision-making, product design and revenue potential is practical and compelling. For example, AI could be used to analyze internal communications to assess employee morale, enabling the company to predict attrition rates and create targeted programs to raise morale. AI could also be used to identify and even predict changes in customer demographics or preferences, enabling the company to develop new customer experience or product strategies that keep pace with these changes.
However, a study by MIT Sloan Management Review found that out of 85 percent of executives who believe AI would help them establish or maintain a competitive advantage, only 20 percent are currently incorporating it into their businesses. So what is the holdup? The challenge of ensuring high-quality data.
How to Improve Data Quality
High-quality data is essential. Health care companies, for example, must be able to trust that data in their business systems and data collected from clinical trials has integrity if they hope to transform that information into intelligence that can drive current programs and potentially lead to breakthroughs that will revolutionize the way care is delivered. Poor data quality can end up doing more harm than good!
Most executives understand this and know that before jumping into the proverbial data lake of AI, they must take certain preparatory steps to ensure the organization will obtain the greatest possible benefits from the investment in people and new technology — while certifying the management of the data powering these programs satisfies internal and external governance requirements.
Success in AI comes down to having the right IT infrastructure, enough quality data to learn from and the cognitive talent to put it all together. And businesses must take care to build this foundation properly. Gartner estimates that poor data quality costs organizations an average of $15 million per year.
Unified governance is the foundation organizations need to define and implement the overall management of the security, integrity and effectiveness of data. Any company poised to implement ML or AI should follow a well-designed and sustainable enterprise governance model, such as the Information Governance Process Maturity Model (IGPMM), that can provide the required governance without limiting the overarching and competitive potential of these technologies.
Unified Governance Steps for Successful AI and ML Initiatives
Inventory your organizational data. For years, organizations have squirreled away data haphazardly. As a result, they may be unaware of large amounts of data stored within their enterprise. Start by focusing on a specific project, especially one that could bring faster gains, and expand into wider organizational coverage.
Eliminate redundant, obsolete and trivial (ROT) data so all valuable data can be identified and protected no matter where it resides — in the cloud, on premises, on mobile devices or in the hands of partners.
Catalog the remaining data using metadata tags to identify data types, usage, ownerships, data lineage and beyond. Because companies in certain industries share common needs, pre-built data models for these industries can expedite the cataloging process via industry models.
Integrate data from multiple sources by bringing structured and unstructured data together and allowing integration with open technologies.
Create a data flow that is automatically synchronized with the original to help ensure that the most recent data is available in data lakes, data warehouses, data marts and point-of-impact solutions.
Secure and protect strategic and sensitive information assets. Their data life cycle should be managed from creation to disposal.
As these steps reflect, a successful governance program obviously brings about additional business benefits regarding data privacy, security and compliance, as well as legal and other operational costs, but most important for AI and ML initiatives, it will instill confidence in the data being used. Data will be deemed to have integrity because it’s reliable and consistent — freeing data scientists to spend more time formulating and refining algorithms and models instead of parsing through data sources.
AI and ML undoubtedly offer exciting and limitless possibilities for organizations of tomorrow. However, just as you can’t have a Formula One race without high-performance cars and an FIA Grade 1 track, you can’t have a successful AI or ML initiative without a solid unified governance program.