Is your data ready for AI? Part 2
In our last post, we explored the importance of building a resilient data architecture to prepare your data for enterprise AI. The next step in achieving scale involves creating well-defined governance, including data management, data stewardship and change management practices.
If you have poor-quality data, applying AI will only help you make bad decisions faster. Implementing foundational capabilities will help you put governance processes in place to identify bad data, capture it and make any changes as needed. Structuring data through validated business rules and putting appropriate access and controls around it will help this process go smoothly.
Here are additional data management principles to follow:
- Store once. By using concepts of "one fact in one place" and a "single source of truth", enterprises can ensure data integrity. They can also lower storage and operational costs while reducing data redundancy, proxy rules and inconsistent access.
- Build trust in the data. Setting quality control measures at the source ensures data quality by reducing the potential for conflicting business results.
- Use a common vocabulary. Companies should define consistent data standards and models, then use them uniformly across the enterprise. Companies must have the ability to capture all metadata initially, then keep it accurate and up to date, managing and governing it as a corporate asset.
- Allow data science access. Data scientists can discover hidden data relationships and anomalies when allowing a provisional analytics sandbox environment for conducting test-and-learn activities rapidly.
- Speed up data use. Deploy architecture that allows for minimizing the time between data, insight and action.
- Secure data. Companies should implement security holistically, in accordance with external regulatory policy and client standards. We do not explore security further in this article series, but you can learn more from the IBM Institute for Business Value.
IBM helped one global insurer that was aiming to parse through various sources of data to better verify coverage and policy information. After properly establishing the data architecture and applying an AI solution, we created a single, unified view of their data.
The client can now structure unstructured source data in real time to extract relevant information, eliminating the manual search and analysis that was previously required. Furthermore, once we developed a clean data architecture, we could layer on an AI-enabled, real-time fraud detection solution into the claims process.
Data is an asset, so it's important to define formal accountability for all data governance responsibilities. Data stewardship is about following your vision and determining how your enterprise will execute it by examining user groups within the organization. Companies should prioritize developing an organizational structure to determine:
- How the data is being used
- Who will be assigned ownership of the data
- Who can make decisions about the data
You can increase adoption by figuring out which business units are the best candidates to migrate from the legacy environment to the new data platform. From there, you can help these users build skills to get the most out of the new technologies. Enterprises need a mechanism in place for creating the proper organizational structure and metrics to track improvements. This requires a broader perspective that incorporates data architecture and design principles.
Executive support and change management
It's critical to ensure strong executive leadership support for these visionary transformational initiatives at the level of CEO, CFO and/or chief data officer. Leaders at many competitive organizations already know this. The IBM Enterprise AI report found that 86 percent of outperforming companies have enterprise-wide data governance in place, with the understanding that investing in your data architecture and governance will create company-wide growth.
Once executive leadership is on board, institute change management practices. People need training in how to work effectively in the new architecture and understand the importance of compliance.
To achieve successful transformation throughout the enterprise, the executive team must deliver a strategic vision stating the compelling reason for the change and outlining the desired future state. Enterprises also should have an actionable business case that depicts the value of change across each level of the organization, focusing on the "people" dimension of competitive advantage.
Executive support must remain visible and consistent throughout the entire project.
It can be done
A large healthcare provider set out to diversify its income through first-of-its-kind analytics. By reinventing its data platform with AI, the company can now process more than 35 trillion operations in less than 20 minutes, creating more than $100 million in operational gain annually. The platform is now being deployed as part of a commercial offering.
While it's possible to take a one-off approach to AI, it's not optimal. AI market leaders move beyond experimentation, investing in AI at scale as an enterprise activity instead. With the data governance practices above in place to ensure the right foundation, you can effectively integrate and scale AI technologies and other emerging tech to create exponential value for your business.