Dr. Taha Kass-Hout, Global Chief Science and Technology Officer, GE HealthCare.gettyOne of the most consequential applications of AI is in cancer care. Roughly one in five people will develop cancer in their lifetime, according to the World Health Organization, with about 20 million new cases diagnosed globally each year. We’re already seeing the impact of AI, where the improvements in molecular understanding are translating into more precise approaches to care.​ For example, two years ago, the Nobel Prize was given to a group of researchers who showed that machine learning models can infer the shape of protein structures, which is crucial to understanding their function. That work is essential for cancer therapy. However, when it comes to cancer care, there’s another pressing bottleneck we can solve by harnessing the power of AI: improving workflow efficiencies that help reduce the time from the first diagnosis to treatment. A Workflow Problem In many cancer centers, the path from diagnosis to treatment unfolds as a series of disconnected steps that are only loosely coordinated across systems. A standard radiation oncology workflow typically involves a series of interconnected steps from patient registration and imaging to contouring (outlining the tumor and nearby organs on the scan so the radiation plan can be built), prescription (the clinician’s treatment order for dose, target and number of sessions), treatment planning and scheduling—with the steps playing out across separate systems. This fragmentation causes delays driven by the repeated data entry and the additional need for manual confirmation at every step. This is where AI can be useful. Because the technology is inherently good at recognizing patterns, it can standardize repetitive steps and surface exceptions earlier. AI-supported orchestration can help capture patient information once, route imaging and treatment data automatically, structure prescriptions and flag missing or inconsistent information. We’re already seeing the impact of AI in removing manual friction from the steps that surround clinical decision-making. In a peer-reviewed evaluation of AI auto-segmentation across breast, head and neck, lung and prostate radiotherapy cases, AI tools reduced contouring time by roughly 14 to 93 minutes, depending on the cancer site. Meanwhile, a 2026 multicenter study found that automated radiotherapy planning could generate directly deliverable plans within five minutes, with more than 80% meeting clinical criteria. In our work tracking these gaps ​in these workflows at GE HealthCare, we've found that the objective is to condense these workflows from weeks to days. The next evolution is to scale what’s working into a reliable operating model. That means working on enabling a future where we define which data must be captured at each point, which tasks can be triggered automatically and which exceptions require human review. Enabling Precision Care The workflow problem becomes more complicated in radiopharmaceutical therapy, where the workflow challenges described above begin to compound. Radiopharmaceutical therapy combines imaging and treatment by delivering targeted radiation directly to cancer cells identified on scans. That makes the pathway more complex because clinicians are not only planning treatment, but they are also measuring how much radiation reaches the tumor and nearby organs through a process known as dosimetry. In radiopharmaceutical therapy, a wrong dose or isotope, a rescheduled appointment, a missed lab review, a delayed pharmacy coordination, a missed dose order, an incomplete documentation or an unavailable infusion space can create safety and compliance risks. Variability adds another workflow challenge. When different imaging methods, reconstruction approaches or analysis techniques are used, final dose estimates can vary substantially, with uncertainty of 10% to 15%. AI has the potential to help bring predictability to this type of more complicated workflow by standardizing the path from simulation to treatment. That includes standardizing imaging protocols, contouring inputs, prescription fields, dose constraints, plan-quality checks and approval thresholds. AI can compare each step against those expectations, flagging missing structures, inconsistent parameters, unusual dose distributions, missed labs, scheduling conflicts or documentation gaps for human review. Implementing AI In Your Oncology Workflow For healthcare leaders, the move to AI-enabled workflows should start with the question: Where does work slow down today? The first step is to map the workflow and identify where information is lost, repeated or delayed. Leaders can then define which data should be captured in structured form, which steps can be triggered automatically and where human review is required. Interoperability is often the hardest part. Organizations need secure data exchange, confidence that information is authentic and consistent, and governed use over time, including versioning, exception handling, escalation paths and audit trails. Governance also has to continue after deployment. Models and workflows can drift as protocols, patient populations and staffing patterns change, so leaders should monitor performance and know when a workflow should pause, escalate or roll back. The goal is a more predictable care pathway, where fewer handoffs are manual and fewer steps are repeated. Clinical teams can then spend more time on judgment and making patient-specific decisions. ​​​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?