Moving Healthcare AI Beyond the Pilot: Building the Foundation for Real ROI
Artificial intelligence can save thousands of hours across an organization, but creating efficiency is only the beginning. The bigger question is what an organization does with that newly created capacity—and whether its AI investments ultimately produce measurable business and clinical value.
Foxtrot Services co-founders and co-CEOs Nicole Sanders and Christine Williams joined Megan Antonelli on Digital Health Talks from Health Impact Live for a wide-ranging conversation about moving healthcare AI beyond experimentation and into secure, scalable production environments.
Modernize and Innovate at the Same Time
Healthcare organizations face a difficult balancing act. They must maintain complex, highly regulated systems while taking advantage of rapidly evolving artificial intelligence and data technologies.
According to Sanders, organizations do not necessarily need to wait for every legacy platform to be replaced before they begin creating value with modern technology. New data architectures, applications and AI capabilities can operate alongside existing systems while longer-term modernization continues.
The goal is not simply to choose between replacing legacy technology and adopting AI. Organizations can develop a strategy that allows both efforts to move forward together.
AI Starts With a Trusted Data Foundation
One of the strongest themes throughout the conversation is that successful enterprise AI begins well before an AI model is deployed.
Reliable AI depends on secure and governed data architecture, including clearly defined access controls, trustworthy data ingestion, data lineage and appropriate permissions. In complex healthcare environments, these foundations make it possible to connect information that may historically have existed across disconnected systems while maintaining control over who—or what—can access sensitive information.
This becomes increasingly important as organizations introduce both human users and AI agents into workflows built around enterprise data.
Saving Time Is Not the Same as Creating ROI
AI projects are frequently measured by the number of hours they save. Sanders and Williams argue that the metric alone doesn't tell the entire story.
If an AI-enabled workflow saves an employee two hours but those two hours are simply filled with additional administrative work, the organization has improved a task without necessarily improving the broader outcome.
Leadership therefore has an important role after automation succeeds. Organizations must determine how newly created capacity will be redirected toward higher-value work, better decisions or improved outcomes.
That means AI transformation still requires many of the same disciplines as other major technology initiatives: change management, adoption, training, clearly defined metrics and executive ownership.
Managing the Real Cost of Enterprise AI
The discussion also addresses an increasingly important issue for organizations deploying AI at scale: compute cost.
Higher infrastructure costs can be justified when they are directly connected to workflows producing meaningful business value. The problem occurs when compute consumption grows without a corresponding outcome.
Foxtrot's engineers approach this as an architecture problem. Data structures, pipelines, ontologies and model selection can all affect the cost and performance of an AI system.
Not every task requires the most advanced reasoning model available. Selecting the right model and infrastructure for each workload can help organizations control cost without sacrificing overall system performance.
Why Too Many AI Pilots Can Become Another Form of Technical Debt
Healthcare organizations are experimenting aggressively with artificial intelligence, but experimentation without coordination introduces its own risks.
Different departments may independently build applications around the same data, assets or business concepts. Without common architecture, naming conventions and governance, organizations can unintentionally duplicate infrastructure, increase compute consumption and create another generation of technical debt.
The technology may be new, but the underlying enterprise challenge is familiar: systems work best when organizations think holistically rather than solving individual problems in isolation.
Governance Is What Makes Speed Possible
Governance may not be the most visible part of an AI implementation, but Sanders and Williams describe it as one of the first things organizations should establish.
Before focusing on dashboards, applications and new AI functionality, organizations need agreement around ownership, permissions, controls and the environment in which experimentation will occur.
Establishing those rules early creates a safer environment for innovation and can ultimately allow teams to move faster as new applications, models and use cases are introduced.
A Different Approach to Leadership
The conversation also explores the unusual leadership structure behind Foxtrot Services.
Sanders and Williams founded the company as co-CEOs despite being advised that one person should take the CEO role. Instead of dividing the company into separate areas of responsibility, they describe genuinely sharing the job.
They credit their shared values, mutual trust and willingness to challenge one another with improving decision-making while also providing an example of collaborative leadership for the rest of the organization.
What's Next for AI in Healthcare?
Despite the challenges surrounding modernization, governance and adoption, Sanders and Williams see enormous opportunity ahead.
Modern data platforms are making it possible for clinicians, researchers and other healthcare stakeholders to connect and analyze information that was previously difficult or impossible to bring together. AI can then help those teams identify patterns, ask better questions and analyze information across increasingly large and complex datasets.
As the technology catches up with the complexity of healthcare's biggest problems, the organizations that establish the right foundations today will be better positioned to turn experimentation into meaningful results.
Watch the full conversation above to hear Nicole Sanders and Christine Williams discuss healthcare AI, data governance, enterprise modernization, compute costs, leadership and the path from AI pilots to production.
Learn more about Foxtrot Services at foxtrot-services.com.
