Healthcare investments in AI are burgeoning. As per a Forrester study from earlier this year, US healthcare providers are expected to increase their technology budgets to $69 billion this year, with 36 percent of this flowing into AI-enabled analytics and other software to build the intelligent healthcare organization. Yet, the predominant risk with healthcare AI initiatives is adoption speed, often without a thoughtful strategy to back it.
With the ambition to build an integrated intelligent healthcare organization, many healthcare organizations are venturing into a series of AI pilots or point solutions. These bolt-ons to existing workflows would not work unless the foundational questions around data ownership, process design, legacy system debt and organizational change management are resolved. Healthcare leaders must start the modernization with a razor-sharp focus on these areas to scale their AI investments for long-term success.
Own the shift from data ownership to data lifecycle management
Technology strategy is now inseparable from geographic reality. Previously, conversations centered around “Who owns the patient health records and clinical data?” As part of their strategizing, leaders looked at the location of their database servers and downstream access issues.
Today, questions revolve around where the extract-transform-load (ETL) pipeline stages the data for AI algorithms, where the AI models execute real-time inference, as well as the geographic routing paths of the various API calls along the way. This is the data residency equation. The dashboards may reside in a compliant region, but if the prompts or API calls send protected health information (PHI) to LLMs residing overseas, this could be a residency violation.
Data residency requirements also shape which deployment model makes sense: cloud, on-premises or some hybrid of the two. Infosys, a global leader in AI-first business consulting and technology services, brings over 150 pretrained healthcare AI models and 20,000-plus healthcare-focused specialists to that decision, helping organizations build cloud foundations suited to their specific regulatory footprint.
One way to resolve the data residency challenge is a hybrid multicloud model. Keep the most restricted clinical data on private cloud infrastructure or within local sovereign regions and route less sensitive data and workflows to public cloud resources. A complementary option is a non-ETL, or data virtualization, architecture that queries clinical repositories in place instead of copying them, cutting the risk of PHI getting cached or replicated outside its home jurisdiction.
CIOs can also consider Domain-Specific Language Models (DSLMs) that are customized for particular medical data within a geographic area and deliver real-time, local AI inference. Such models often deliver superior results with contextual accuracy.
Redesign workflows with Human-in-the-Loop touchpoints
Past digital investments in healthcare, whether in disparate areas such as virtual care services or CX technologies, have typically underperformed. The problem has been more with the legacy systems and manual (often broken) workflows than with the technology being deployed. The additional challenge with such deployments has been operational friction.
Before deploying advanced software like agentic AI, healthcare leaders must target highly manual, document-heavy administrative workflows that can be overhauled. Investments must focus on redesigning vital workflows like utilization management (UM) and prior authorization (PA) that are often siloed, costly and slow. Strategists must work on end-to-end workflow design with built-in human-in-the-loop (HITL) touchpoints to make them AI-ready.
In a workflow automation effort, Infosys helped a mutual benefit and health plan operating a large network, replace a legacy case management system to ensure mandate compliance, automating member correspondence and consolidating fragmented data into a single view for care management teams. As a result, transaction processing time fell from 70 hours to 90 minutes, and patient satisfaction improved by 75 percent.
Tackle legacy debt and modernize urgently
As momentum builds to scale up with predictive models and AI agents within healthcare ecosystems, leaders must confront the decades of legacy technical debt: the architectural “sediment” that often consumes a major part of IT budgets to run the business rather than grow and transform it.
While overall healthcare IT budgets are expanding, the net value gets heavily diluted, as CIOs must funnel a substantial amount of their budgets into patch management, custom system integrations or compliance maintenance for legacy tools.
Outdated data management practices, old compromises in data infrastructure or data source documentation, and sub-par data storage practices are significant obstacles for CIOs: it is predicted that organizations who delay data debt remediation could face 50 percent higher AI failure rates by 2027, not to mention rising costs.
There are bright spots in the industry. For example, core EHR (electronic health record) modernization has progressed significantly in the US; however, legacy issues such as custom local databases and legacy billing software may impede true interoperability.
Prioritize organizational change management
Healthcare leaders must maintain strategic focus on organizational change management and workforce readiness. Rushing AI tools into production without adequate workflow readiness can trigger a “trust tax.”
Major clinical organizations are raising questions about the usage of AI, particularly in clinical workflows. Concerns range from dilution of professional autonomy to ambiguous liability risks and inherent algorithmic prejudices. They are issuing AI playbooks and building guardrail frameworks that healthcare leaders are well-served to operate by.
Moving to a trust-based governance model that prioritizes transparency and clinician safety, can go a long way towards building the foundations for AI deployments. The frameworks must mandate rigorous validation with built-in HITL reviews that preserve clinical integrity.
Healthcare is a high-stakes, essential sector that is transitioning to mature, enterprise-wide AI adoption. To ensure this shift truly benefits healthcare consumers and employees, as well as the business, healthcare leaders must modernize their digital foundations first.
