From Pilots to ROI: A Model for Healthcare

The healthcare industry is poised to be profoundly reshaped by artificial intelligence. Automation, predictive analytics, ambient listening, diagnostics, notation. Even today’s relatively nascent AI technologies have moved some 85% or more of healthcare providers and payers to explore AI to reduce costs, deliver quality care, and address the sector’s perennial staffing shortages. Over the past year, C\R Strategy Partners has seen this exploration and adoption up close.

In early 2025, we partnered with the Health Plan Alliance to deliver a year-long AI Impact Accelerator, working with the organization’s membership of regional health plans to fast-track their AI strategies, establish foundational governance, and begin to build their AI centers of excellence. Across twelve months, these C-suite executives learned from the vanguard of AI experts, including Charlene Li, Gartner’s Laura Craft, and McKinsey’s Carlos Pardo Martin. They built their governance frameworks with the expertise and guidance of attorneys at Epstein Becker Green and Sheppard Mullin who shared their deep experience in regulatory and compliance matters. With Carissa Rollins, former CIO at UnitedHealthcare and Illumina, they prioritized use cases and mapped a path from proof of concept to production systems delivering game-changing return on investment, not only in real costs but in human capital more effectively deployed in service to their members. They grilled leading vendors on their AI roadmaps and partnership strategies. And they learned to leverage Repplit and other vibe coding tools to rapidly build prototypes that tested ideas ahead of significant investments.

Which is to say, they did a lot in one short year to begin to reap the benefits of AI in organizations that - like virtually every health payor and provider - is under existential pressure to become more agile, more responsive, and much, much more cost efficient.

We build programs like this one to - as the name implies - accelerate our clients’ learning and adoption of transformational technologies. We sometimes underestimate just how accelerated our own learning and adoption will be. That was certainly the case here. Our deep dive into AI and healthcare reinforced our commitment to both the technology and the market. We worked with our team to build our AI proficiency and leverage that proficiency in our client delivery models.

We expanded our learning circle to include experts and leaders in both the payer and provider side of the healthcare industry. And we listened intently to the challenges, limitations, and opportunities our accelerator members brought to every meeting.

The program proved to be an extraordinary vantage point to better understand how healthcare organizations are integrating AI technologies, what they are getting right, and why. We came away with three insights:

• AI is everywhere, but ROI is still elusive.

• Healthcare organizations need an “operating system” to scale

• A Transformation Readiness Model helps place the right bets

The Pilot Trap: Activity Outpaces ROI

We’ve seen it first hand: Healthcare leaders aren’t short on AI pilots. 

Most executive teams can point to dozens of active experiments: call center automation, utilization management optimization, clinical summarization, prior authorization support, document intelligence, and internal productivity tools. The energy is real, and the intent is serious.

Yet the financial and operational outcomes of these trial balloons are often disappointing. AI seems everywhere, but ROI is still elusive, leaving many leaders to conclude that pilots were “interesting but not transformative” or that something failed in the execution.

That diagnosis is usually wrong.

In healthcare, the most common failure mode is not failure. It is stalling. And the stall happens in a predictable place: between pilots and enterprise scale.

Pilots are easy to justify because they feel contained. They can be scoped narrowly, funded quickly, and showcased compellingly. They allow leaders to learn and appear responsive in a fast-moving market. In regulated environments, pilots also feel safer than enterprise commitments.

But pilot-heavy strategies create structural consequences. When the number of experiments grows faster than the organization’s capacity to integrate, govern, operationalize, and measure them, the organization enters what many teams experience as “pilot purgatory.” Results remain local. Lessons remain fragmented.

Costs accumulate quietly. Leaders feel busy, while the enterprise remains unconvinced.

This is not a leadership failure. It is a systems mismatch: pilots are optimized for learning; ROI has to be optimized for repeatability.

Across health systems and health plans, three constraints explain most of the gap between activity and ROI:

Workflow integration. Adoption is not a communications problem; it is a workflow design problem. AI that adds steps, increases cognitive load, or creates ambiguity about responsibility will be ignored, even if it performs well technically. Clinicians and operators do not resist AI as a concept. They resist friction and unclear accountability.

Governance. Healthcare governance is not an accessory; it is the chassis that enables speed safely. When privacy, safety, auditability, model risk, and escalation processes are unclear or introduced late, teams encounter delays, rework, and organizational conflict. The organization learns to treat compliance as a blocker, and innovation teams learn to route around it. Both dynamics reduce trust and slow delivery.

Measurement and benefit capture. Many pilots produce real improvements, but those

improvements are rarely captured as enterprise value. “Time saved” does not automatically become ROI. Unless a system exists to translate time saved into capacity created, cost removed, revenue enabled, or risk reduced—using consistent baselines, instrumentation, and reporting—impact remains anecdotal. Anecdotes do not scale.

When these three capabilities are weak, leaders can spend a year building promising pilots and still produce little defensible ROI. Not because leaders lacked commitment, but because the system wasn’t designed to turn their effort into enterprise value.

The immediate value of pilots

The most important outcome of early AI cohorts is often not a set of scaled deployments. It is clarity. First movers test reality. They reveal friction in workflows. They uncover governance gaps. They discover where data access breaks down.

They illuminate where measurement systems are absent. In other words, they de-risk the next wave and provide a roadmap for what must change.

This is why the right executive question is not, “Why didn’t our leaders deliver ROI?” It is, “What capabilities were missing that made ROI structurally unlikely?”

Organizations that escape pilot purgatory do not do it by running more experiments. They do it by installing a repeatable operating discipline that makes pilots convertible into products, and products convertible into enterprise value.

That discipline begins with a small set of enterprise bets, each with a clear owner and a measurable value hypothesis. It includes early governance gates that make approvals predictable instead of political. It includes a production pathway with monitoring and support. And it includes a measurement and benefit capture function that makes value visible and defensible.

The AI era will not reward organizations with the most pilots. It will reward organizations that can reliably convert pilots into scaled outcomes without increasing clinical or reputational risk. That is not a story about heroic leaders. It is a story about enterprise operating capability.

Creating a Playbook for Scaled Value

If you want healthcare AI ROI, you need an operating system, not a project plan.

An operating system is not a platform purchase. It is the set of enterprise capabilities that make AI outcomes repeatable across initiatives and scalable across sites. It answers questions that most pilots avoid:

Who owns the solution after the pilot ends? How will it be monitored? What happens when it fails? How do we ensure adoption in real workflows? How do we prove value, and who captures it?

Healthcare AI becomes enterprise value only when these questions have repeatable answers.

A practical healthcare AI operating system has four pillars.

1) Focus: Portfolio discipline. Most organizations lose ROI by funding too many bets. Leaders should reduce AI efforts to a small number of enterprise priorities and insist on explicit kill criteria. The goal is not to slow innovation. The goal is to prevent dilution and create the depth of adoption required for measurable impact.

2) Trust: Governance spine. Governance must be designed to accelerate delivery, not delay it. That requires predictable gates for privacy, safety, auditability, monitoring, and incident response. When teams know what “good” looks like, compliance becomes a path, not a fight. Trust is not built through messaging; it is built through design.

3) Delivery: Product operating model. AI must be treated like a product or service, not a one-off project. Projects end; products persist. A product has an owner, a backlog, a release cadence, a support model, and monitoring. When pilot sponsorship replaces product ownership, pilots die after the demo.

4) Value: Measurement and benefit capture. If value isn’t captured, it doesn’t exist at the enterprise level. That means baselines before deployment, instrumentation of usage and adoption, outcome measurement linked to workflow change, and a translation model that turns outcomes into capacity, cost reduction, revenue, or risk reduction. Many organizations have “impact.” Few have benefit capture.

The CEO–CIO handshake and the 90-day sprint

This operating system succeeds only when CEOs and CIOs align on a shared truth: AI ROI is a business capability. The CIO cannot deliver it alone, and the CEO cannot demand it without funding the supporting systems.

The CEO role is to set enterprise focus, protect capacity, and define what value the organization is pursuing. The CIO role is to build the pathways—governance, production, monitoring, measurement—through which that value becomes repeatable.

When CEOs and CIOs are aligned, AI becomes a managed portfolio. When they are not, AI becomes a collection of disconnected initiatives.

Organizations do not need perfect data to create momentum. They need disciplined moves that reduce sprawl and increase repeatability. Here’s a straightforward approach that moves pilots from experiments to production decisions in 90 days:

• Choose two to three enterprise bets tied to strategic outcomes.

• Assign true product ownership, not just sponsors.

• Define early governance gates and make cycle time visible.

• Establish minimum viable measurement: baseline, adoption instrumentation, outcome metric, and value translation.

• Implement a portfolio cadence that makes “commit, scale, or kill” decisions routine.

This approach is politically safer than it sounds. Kill criteria protect leaders from being judged on pilots that were never structurally capable of scaling. It shifts performance from heroics to system design.

The organizations that win in healthcare AI will not be those who pilot the most. They will be those who install the operating system that turns pioneering work into scaled, measurable value.

A Transformation Readiness Model for Healthcare AI

A transformation readiness model provides that early visibility. It translates “AI readiness” from a vague concept into a diagnostic system that guides what to build next.

A practical readiness model has two layers.

Layer 1: A five-stage pipeline.  AI transformation follows a predictable sequence: Prioritize the right bets, Prove workflow and risk feasibility, Produce a monitored production deployment, Propagate through playbooks and replication, and

Perform through benefit capture and value reporting.

Most organizations do not stall in the first two stages. They stall in Produce, Propagate, and Perform, where operationalization and measurement live.

Layer 2: Ten capabilities that unlock the pipeline.  The pipeline is not driven by effort; it is driven by capability thresholds. Leaders and organizations must have ten capabilities: portfolio discipline, value framing, workflow/adoption leadership, governance/model risk leadership, internal leverage for data and platform constraints, product ownership mindset, production pathway competence, scaling systems, measurement and benefit capture discipline, and coalition building across clinical, operational, IT, compliance, and finance stakeholders.

The purpose of these capabilities is not to judge individuals. It is to reveal what the system must provide for success, which requires three overarching capabilities:

• Leader readiness (what individuals can drive)

• Team coverage (whether the pilot team collectively has all required capabilities)

• Enterprise readiness (whether governance, production pathways, and measurement systems exist)

Leader development can improve leader readiness. Enterprise ROI requires enterprise readiness.

The power of a readiness model is what it produces: actionable artifacts.

At the cohort level, it produces a heatmap that shows where a cohort is strong and where scale will likely stall. It produces a stage readiness view that makes the missing middle visible. And it produces a program intervention list that tells accelerator designers what scaffolding to provide.

At the individual level, it produces action plans that are psychologically safe and operationally specific: what a leader’s strengths enable, where they are likely to stall, what to do in the next 30/60/90 days, and who they should partner with to cover missing capabilities.

This is how readiness becomes a lever, not a label.

The bottom line

Healthcare AI is entering a phase where experimentation will no longer differentiate organizations. Repeatability will. Organizations that can diagnose readiness early, build the missing middle deliberately, and capture value consistently will turn AI from a series of pilots into an enterprise advantage.

A readiness model doesn’t replace leadership. It makes leadership effective by giving it a system to operate within.