What if the biggest threat to AI implementation in healthcare isn’t regulation or funding, but impatience?
According to IBM’s recent CEO survey, only 25% of AI initiatives have delivered expected ROI over the last few years, and only 16% have scaled enterprise-wide. Healthcare isn’t immune to this AI disappointment. A Bessemer Venture Partners survey found that only 30% of AI pilots in healthcare reach production, held back by factors like data readiness.
With mounting pressures to innovate, it can be tempting to place big bets on AI while skipping the essential—and decidedly less sexy—work of building reliable data infrastructure and governance first. But AI’s impact hinges on having this foundation in place.
Data Governance: Enabler, Not Obstacle
Data governance has a bad reputation. The phrase suggests bureaucratic committees, endless approval processes, and rules that slow down innovation. It’s seen as the department of “no.” But this perception couldn’t be more inaccurate.
The role of modern data governance is making data more accessible, reliable, and useful across the organization. Think a conductor orchestrating a symphony rather than a traffic cop.
Mark Ramsey, former Chief Data Officer at GlaxoSmithKline, puts it this way: “Effective data governance is less about control and more about enabling the flow of information to the right people and systems.”
Consider what happens without proper data governance: a physician can’t access a patient’s complete medical history because it’s scattered across incompatible systems. Quality improvement teams can’t identify patterns because they don’t trust the data they’re looking at. And AI systems trained on this fragmented, unreliable data will inevitably produce flawed predictions and recommendations that clinicians can’t trust.
Effective data governance solves these problems by creating infrastructure that makes data work for everyone. It establishes clear pathways for data to flow where it’s needed, when it’s needed, in the format that’s most useful. It’s the difference between a hospital where clinicians waste time hunting down information and one where these insights surface automatically at the point of care.
The reality check: Are you ready for AI?
Before making significant investments in AI, healthcare organizations need to ask themselves a critical question: Do we have the proper data infrastructure and governance in place to make AI actually work?
Without these, even the most sophisticated AI becomes an expensive experiment that can’t deliver results at scale. It’s like building a skyscraper without a foundation—the structure will inevitably fail regardless of how advanced the engineering is.
Successful AI implementation depends on several foundational elements:
- Clean, High-Quality Data: AI systems are only as good as the data they’re trained on. Poor quality data—incomplete records, inconsistencies, duplicates, or errors—leads to unreliable AI outputs. In healthcare, this could mean misdiagnoses, incorrect treatment recommendations, or failed predictions. Further, for AI to work across an entire health system (not just in pilot projects), data must be standardized and consistently formatted. Different departments, locations, or systems often store the same information in different ways. Data governance establishes standards for how data is collected, stored, and formatted, making it possible for AI systems to scale beyond individual use cases. Clean, accurate, standardized data is non-negotiable for AI models to identify meaningful patterns and make reliable predictions.
- Transparency-driven trust. AI systems must demonstrate compliance with these stringent healthcare data regulations, which requires robust data governance frameworks that track data usage, ensure proper consent, and maintain audit trails. Healthcare providers also need to understand how AI systems make decisions, especially for clinical applications. This requires knowing exactly what data the models are using, where it came from, and how reliable it is. Strong data governance provides this transparency through data lineage tracking and quality metrics.
- Data Integration: Healthcare data exists in silos—EHRs, lab systems, imaging systems, billing platforms, wearables, etc. Without proper data integration and interoperability, AI systems can only see fragments of the patient story. Bringing data together from all of these disparate sources enables the kinds of insights that make AI valuable (Think about receiving text reminders for your next vaccine; this requires seamless integration of multiple data sources).
The Path Forward: Patience as a Strategic Advantage
It’s true: the opportunities for AI-driven transformation across the healthcare spectrum are vast. Leaders are focused on technologies that deliver meaningful results for both providers and patients: improving outcomes, expanding access, and easing the strain on overburdened systems.
But the healthcare organizations that will ultimately pull ahead in the AI race are those methodically building from the ground up. While competitors chase headlines with AI pilots that fail to scale, forward-thinking healthcare leaders are investing in the unglamorous work that makes this transformation possible.
The choice facing healthcare leaders today is stark: continue the expensive cycle of AI experimentation that leads nowhere, or step back and build the infrastructure that turns AI’s promised capabilities into practice.
About Houdini Abtahi
Houdini Abtahi has 15+ years of experience in healthcare consulting. His clients have spanned across payors, providers, pharmaceutical, and life science companies. As Resultant’s private sector healthcare lead, Houdini oversees solution delivery practitioners and project delivery teams while driving business development. He’s most passionate about improving the patient experience while helping companies reach their innovation goals.
