Beyond data migration: Preparing health care organizations for AI at scale

Health care organizations find themselves at a pivotal juncture. Across enterprises, there is a concentrated effort under way to bring together fragmented data from different sources into a unified system to facilitate real-time decision-making and foster artificial intelligence (AI) adoption at scale. However, companies often find themselves left with modern platforms that underperform or yield insights their clinicians don’t trust.

Treating data migration as the foundation for AI-led transformation, rather than a technical process, can help unlock value from data. Effective data migration goes beyond simply transferring data from existing systems to a new environment. Legacy data systems often contain inconsistencies, duplication and incomplete records that, when transferred without standardization, result in a modern platform plagued by legacy issues and complications. Further, clinical, administrative and technical teams typically operate in silos with different expectations about what the new system must achieve. Without a shared understanding of outcomes, decisions about how to map data fields or reconcile discrepancies tend to be made based on technical factors rather than use-case-based considerations.

Why data migrations fail

Health care data are highly contextual: A lab value, for instance, is meaningful only when tied to units, reference ranges, timing and patient history. When migrations strip away or distort this context, the data may remain technically intact but be clinically unusable.

Unclear scope and data clarity

The scope of the data migration exercise cannot only be defined in technical terms; clinical and operational factors are equally critical.

Ambiguity in scope impacts how trade-offs are made. For instance, if the goal is to support specific workflows such as care coordination or predictive modeling, the data required for those workflows must be identified and prioritized.

Ignoring clinical workflows

The disconnect between technical migration and clinical workflows can significantly affect the effectiveness of data migration. Data may be transferred accurately on paper, but if they do not appear where clinicians expect them and in a format that they are used to, care delivery will be disrupted. Clinicians are saddled with slower workflows and longer processes rather than a streamlined, modern system, leading to frustration and an erosion of trust in the system. In health care, where accuracy and speed are critical, this is not just a minor inconvenience but a fundamental failure.

Related coverage: The new cost engine: How AI and automation are reshaping health care economics

Building data trust

As health care organizations move toward AI adoption at scale, the importance of data trust cannot be overstated. AI systems are only as reliable as the data on which they are built.

Governance plays a central role in establishing this trust. Effective governance ensures that decisions about data are consistent, transparent and in line with organizational goals.

This must then be supported with robust metadata and data provenance. Metadata describes aspects such as what the data represent, how they were generated and how they should be interpreted. If this is missing, the data become ambiguous, especially when aggregated from multiple sources. Provenance adds another layer of trust by addressing critical aspects like where the data come from, whether they have been modified, and who has access to them. With accountability and traceability being at the heart of data adoption in health care, these aspects are nonnegotiable toward building trust among users.

This framework gives users the confidence to use the information for critical decision-making, which impacts patient outcomes and operational processes.

Designing around care, not code

A defining characteristic of successful data migrations is the orientation toward outcomes. This means identifying clinical and operational use cases, the decisions that will need to be supported and the workflows that will need to be enabled once the migration is completed. Working within this framework will ensure that the data will do what they are meant to.

For instance, streamlining revenue cycle management by identifying possible causes of claims denial and flagging these before submission is often an early AI use case. The AI system can analyze the data to understand precisely why a particular payer is denying claims and then help proactively fortify claims before submission to reduce denials and speed up the approval process.

Validation plays a critical role in this process. After the data have been transferred, it is essential to verify that they behave as expected within the clinical and operational workflows and are in a usable format. Given the inherent complexity of health care environments, simplifying data systems can improve reliability and boost user adoption.

Doing this effectively requires close collaboration between technical and clinical teams. Technical teams, clinicians and administrators must work together to define priorities, resolve ambiguities and ensure that data retain their meaning throughout the process. Once migration is seen as a shared responsibility, different teams start working together to identify potential issues early and ensure that the system supports real-world use cases.

When are health care data ready for AI?

The transition from data readiness to AI readiness isn’t always smooth. There is a misconception that once data are centralized and accessible, they are ready for use. The reality is that the data need to meet several other parameters to serve as a reliable base for AI implementation.

To prepare for AI adoption at scale, health care organizations must rethink their approach to data migration. Rather than a one-off technical step, it should be seen as the foundation for a broader transformation process. This shift requires a change in mindset. Although every migration process will come with its own set of challenges, following standardized best practices will help streamline the process.

Health care leaders who understand that data will never really be fully ready will focus on early use cases that maximize what the available data can be used for. Identifying areas where AI has demonstrated value, often in processes like clinical decision support or operational efficiency, will help deliver immediate, tangible results. In turn, this can help create trust and build momentum for widespread adoption within the organization.

A new mindset for data and AI

The path to AI adoption in health care does not begin with algorithms. It begins with data. The scope of health care data is expanding rapidly. In addition to structured clinical records, organizations must now manage unstructured data such as images, documents and video, as well as data generated by patients. This means that traditional approaches that rely on rigid structures will no longer be relevant. Organizations will need to adopt flexible environments that can accommodate this diversity in data types and design systems that are robust and adaptable.

Migration must be seen as a strategic enabler of care transformation. It must be approached with the same rigor and attention as any clinical initiative. Success cannot be defined by the volume of data migrated, but rather by how effectively technology supports care delivery and fosters trust through strong governance practices.

Ultimately, the organizations that succeed will not be those that move data the fastest, but those that can derive the most value from them.

John Squeo, MBA, FACHE, CHCIO, CDH-E, is a seasoned health care technology executive with over 27 years of experience spanning health systems, interoperability and cloud technologies. As a senior vice president at CitiusTech, he leads business development, account management, sales and partner channels for the provider and health care services market.

Tyler Smith serves as the CEO of Health Data Movers (a CitiusTech company). Tyler co-founded Health Data Movers and played a central role in building the company’s client services and finance and administration functions before becoming CEO in 2021.