Our Take
Healthcare organizations mistake AI spend for innovation when they should treat it as infrastructure tied to measurable enterprise value like length-of-stay reduction or coding cycle improvement.
Why it matters
Hospital C-suite decision-making on AI is shifting beyond CFOs to CIOs and clinical leaders, which matters because clinical workflow design and data quality determine whether AI fails or delivers ROI. This timing coincides with post-EHR deployment lessons that executives are finally applying.
Do this week
CIO: Audit your AI purchasing criteria before June 2026 to ensure clinical workflow designers and informaticists sit on selection committees, not just technology buyers.
CIOs become the primary AI decision-maker in hospitals
A survey cited by Dr. Deepti Pandita, CMIO and chief AI officer at UC Irvine, confirms that CIOs are expanding their strategic role in AI purchasing. By 2026, 45% of CIOs will be the primary decision-makers on AI investments (per the survey referenced in the source).
This shift reflects a broader recognition that AI spending requires alignment across technology, operations, and clinical strategy. Executives now evaluate AI's impact on workforce integration, clinical risk, accountability, and enterprise alignment, not just cost or feature parity.
But the headline masks a harder problem: most organizations deploying AI in healthcare are still getting the infrastructure wrong.
Broken workflows and data pipelines doom AI before it ships
Pandita's core warning is direct: "If you let the AI systems dictate what follows in terms of workflow or implementation, they are very likely to fail." This lesson came hard from EHR deployments, where hospitals bought systems first and redesigned workflows second.
The mistake repeats with AI. Organizations treat AI expense as innovation rather than infrastructure. They deploy point solutions without first auditing clinical workflows or data quality. The result: biased outputs, low adoption, and squandered ROI.
Having a practicing physician and informaticist embedded in AI system selection is not optional, Pandita argues. They catch workflow friction and data gaps that procurement alone misses. When AI systems are allowed to dictate downstream workflow, the whole deployment fails.
Pandita advises a different framing: tie AI explicitly to enterprise value. Measure success in reduced length of stay, improved revenue cycle coding, or faster appeals processing. "If organizations start looking at AI as infrastructure and not as point solutions, ROI will follow," she said (per the interview).
What hospital leaders should do now
Revenue cycle leaders looking at AI for automation and payer management should implement AI mid-cycle in the revenue process, not at endpoints, for the highest return. That requires mapping your current workflow first, identifying where data breaks, and then selecting tools that fit your process, not vice versa.Trust and patient outcomes depend on clinical design. AI that schedules telehealth during working hours because it knows a patient's schedule builds trust. But that only works if the data pipeline feeding the AI is clean, the workflow assumptions are validated with clinicians, and the system is treated as infrastructure that enables clinical staff, not as a replacement for judgment.
The convergence of CIO decision-making with clinical leadership creates an opening. Use it to embed workflow and informaticist review in your AI selection criteria before the next round of RFPs.