Abhishek Kumar – Chief Information Officer at AccessHope.
The conversation around AI in healthcare is shifting, and not a moment too soon.
For years, we celebrated AI progress measured in faster analysis, cleaner summaries and increasingly sophisticated decision support. Those advances have been meaningful, but they’ve also created a false sense of momentum: the belief that better insights automatically translate into better care.
In reality, they don’t. Not yet, at least.
Across healthcare, and especially in oncology, AI now routinely surfaces evidence‑based recommendations during case reviews. The guidance is often clear, aligned with the latest research and directionally sound. But when teams revisit the case weeks later, a simple question often goes unanswered: Did anything actually change?
Consider a patient with metastatic lung cancer whose case review recommends molecular testing before initiating an expensive immunotherapy regimen. The recommendation may be evidence-based and clinically sound. Then, from what I’ve seen in my experience, stakeholders often still cannot determine weeks later whether the treatment changed or whether the patient ultimately remained on the original therapy. The recommendation existed, but the impact remains invisible.
Too often, the answer is unclear. The gap between what AI recommends and what happens in real‑world care is becoming the defining challenge for the next phase of healthcare intelligence.
Closing The Insight–Impact Gap
Most AI systems still operate as a one‑way layer. They generate guidance but rarely show whether it influenced decisions, altered treatment plans or changed the trajectory of care and patient outcomes.
This matters because the stakes are enormous. Cancer‑related costs in the U.S. are projected to reach $245 billion by 2030, a steep rise from $183 billion in 2015, according to the National Cancer Institute.
At the same time, according to research published by the American Society of Clinical Oncology, as many as 58% of clinicians had to deviate from evidence‑based guidelines, as a result of prior authorization.
Across oncology workflows, I routinely observe that when evidence-based guidance arrives too late or cannot be operationalized effectively, patients often progress further into high-cost, high-complexity care pathways before intervention occurs. For example, a delayed biomarker-driven therapy adjustment may not immediately appear significant clinically. Downstream, this can lead to avoidable ER visits, ineffective lines of therapy, additional toxicity management and unnecessary spend.
If AI cannot demonstrate whether it influenced a decision, it cannot demonstrate whether it influenced cost, quality or outcomes.
And that is where the illusion of progress becomes most visible. Healthcare has spent a decade measuring activity.
The next evolution of AI in healthcare is not generating more recommendations. It is creating closed-loop intelligence systems that can demonstrate whether recommendations influenced real-world care.
For decades, healthcare has struggled to connect clinical guidance with downstream outcomes. Recommendations often live in one workflow while treatment activity, utilization, pharmacy events and outcomes live somewhere else entirely.
Why Oncology Is An Early Proof Point
Oncology is becoming an early proving ground for how AI can meaningfully improve clinical decision‑making.
Cancer care is defined by a sequence of high‑stakes choices, therapy selection, sequencing, clinical trial consideration and ongoing adjustments, each one shaping what’s next. It is also where the stakes are highest for employers, health plans and patients.
Oncology is uniquely suited for AI because it combines high clinical complexity with significant economic impact. A single patient journey may involve pathology, imaging, genomic testing, treatment selection, prior authorization, supportive care and rapidly evolving clinical evidence—often distributed across fragmented systems.
Cancer organizations and experts have long emphasized that the first treatment decision is often the most consequential, influencing survival, toxicity and total cost of care. When early decisions improve, outcomes improve and downstream costs fall.
That dynamic makes oncology uniquely suited to reveal how clinical intelligence can change the trajectory of care.
Because oncology relies on vast, heterogeneous data that rarely lives in one place or one format, there is great potential for AI to translate the complex information, connecting evidence-based insights into better real-world decisions.
Oncology should not simply be viewed as a proving ground for AI innovation. Instead, it offers a model for what healthcare can become when intelligence systems support earlier, more informed and more consistent decisions and can demonstrate how those decisions affect outcomes, experience and cost over time.
Unlocking The Next Phase Of AI Value
To unlock its next phase of value, AI must evolve from simply generating insights to demonstrating impact. That means:
• Understand clinical intent, not just clinical language. Organizations need to understand the decision behind the recommendation, whether it was intended to change therapy, accelerate diagnostics, reduce toxicity or identify a clinical trial opportunity.
• Connect guidance to longitudinal data. Recommendations should not exist in isolation. Leaders need the ability to follow how care evolved across claims, pharmacy, treatment and utilization data after guidance was delivered.
• Measure why decisions changed or didn’t. Not every recommendation should be followed. Some may be delayed by access barriers, modified by physician judgment or superseded by changing patient circumstances. Understanding why is often as valuable as understanding what happened.
• Account for the reality of healthcare timelines. Clinical decisions, claims, prior authorizations and outcomes occur on different schedules. Effective intelligence systems must interpret change across months, not days.
Achieving this requires healthcare organizations to create a connected intelligence layer that links fragmented clinical, claims, pharmacy and operational data.
The first wave of AI focused on automation and efficiency. The next wave will focus on accountability and measurable impact.
The expectation is no longer that AI generates better recommendations. The expectation is that it demonstrates whether those recommendations changed decisions, altered care trajectories and improved outcomes.
Oncology offers a clear opportunity to prove that this future is possible.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
