Electronic Health Reporter

By Graham McMillan, CTO, Redgate Software.

AI is shaping the future of how healthcare organizations manage data, whether they’re ready or not. According to new research, 41% of healthcare organizations are already using AI for database management purposes, with a further 40% considering integrating it soon.

Many practices are already finding value in leveraging AI for their operations, with top applications including data quality assurance, automating database management, and data modeling.

While AI has the ability to generate massive upside for efficiency, it can also wreak havoc across existing data estates if they’re not properly prepared for adoption and integration. When piloting a new AI initiative, it’s imperative that there’s a solid foundation for the model to work on top of. An unstable base could topple down in an instant, unraveling years of work.

Where DBAs should look first

Database administrators (DBAs) must take stock of the key issues with their estate and address them before AI is added into the system. The keys to successful AI adoption can be easily broken down into three key categories: people, process, and data.

DBAs need to first ask if their team is ready to adopt AI. If the humans overseeing it aren’t prepared, then your initiative could fail before takeoff. When timelines are compressed to meet ROI projections set by stakeholders. That means training people with the skills to use AI and the freedom to deploy what they learn in the workflows they are familiar with. Top-down AI usage mandates are not going to help.

Next, DBAs must have a strong grasp on how value flows throughout the organization. Understanding key bottlenecks, which processes are load-bearing, and how to achieve measurable operational outcomes is essential to AI success. Without clarity, AI can be implemented in the wrong places, cascading chaos. It’s easy to point it at a problem that generates no value, or have it contribute to meaningless metrics rather than real outcomes. And fixing the current ones will not be sufficient. Once the first bottleneck is resolved, new ones will emerge that need to be addressed.

Finally, the most critical problem is the data itself. Healthcare databases can be enormous. Estates and their management processes are often passed down from managers from past years or decades. These legacy processes can lead to platforms that are a jumbled mess of software that doesn’t work together, programs that can’t communicate with each other, fragmented estates, undocumented schema changes, and split ownership. AI doesn’t magically clean up these problems, it simply acts as if there’s nothing wrong. If you don’t create a good foundation for AI to operate, then it will churn out confident answers based on broken information, providing solutions that generate no value

Building real foundations

Database governance should be the top priority for any DBA who’s looking to deploy AI. Right now, nearly 40% of all healthcare practices operate across 4 or more database platforms. The best way to address problems of database fragmentation and software sprawl is to pull everything together under a single umbrella, offering a unified view. Without full visibility, issues quickly turn into costly downtime, impacting revenue and customer satisfaction.

Addressing problems with the database’s structure is only half the battle. Once DBAs have cleaned up issues from the past, they must prepare for the future. The most crucial step is to create clear management processes so teams are aligned. Fragmentation occurs when there’s no standardized process for deploying changes or creating pathways. DBAs need to set clear guidelines for deploying updates and tracking schema developments. When engineers have no guiding principles, they create sprawl which could decimate AI processes down the line.

It might be a pain in the short term, but DBAs who dedicate the time to clean their data estate will realize exponential value down the line.

Looking forward

AI is set to revolutionize the way healthcare data is managed. It has the potential to quickly anonymize massive datasets, streamline database management, design schema, and much more.

However, most practices aren’t prepared to realize AI’s true value, and many will suffer due to poor implementation. DBAs need to be cognizant of the foundations that AI needs to thrive, audit their team’s ability to work with AI, identify the bottlenecks within their organization, recognize which internal processes are load bearing, understand how to generate measurable outcomes, and scrutinize the data itself.

AI can only thrive within clear governed processes and solid support. Don’t fall into the trap of thinking it will automatically fix everything.