While AI has ruled our headlines lately, the most consequential transformations in AI have been taking shape in healthcare. Experiencing a personal health crisis reshapes one's perspective on what truly matters. Facing a rare cancer diagnosis would shift anyone’s focus, for me, political headlines became much less alarming than navigating cancer.

My diagnosis was appendiceal adenocarcinoma, a rare cancer of the appendix. According to the National Cancer Institute, appendix cancers are often discovered after surgery for suspected appendicitis or during investigations of another condition. The anatomy makes my particular finding significant: a conventional colonoscope can show the opening of the appendix but cannot examine its interior. There is currently no established screening test specifically for appendix cancer.

There are about 10,000 things I would rather discuss publicly than my colonoscopy. But the reason I am here to write this begins with that examination in June 2026. Dr. Tamas Otrok found a tiny lesion at the opening of my appendix. He could not see its full borders, so he took biopsies. That decision began the investigation that led to a cancer diagnosis.

I consider myself incredibly fortunate. I am also clear about what that good fortune required: an experienced physician who recognized a concerning finding, advanced surgical tools, and people who cared for me through one of the most challenging periods of my life.

The visible abnormality was a clue. Establishing what it meant required biopsy review, surgery, and examination of the removed tissue. Seeing something, interpreting it, and deciding what to do next were separate steps, each requiring human expertise.

The invasive tumor measured just five millimeters on surgical pathology. It had already reached the muscle layer of the appendix. Early discovery gave us an opportunity, but addressing the cancer still required substantial treatment: removal of my appendix and part of the cecum, the beginning of the large intestine, followed one week later by a right colectomy and removal of 19 regional lymph nodes. The colectomy removed the right portion of my colon.

I chose to move forward with the aggressive approach recommended by my team. My body was not in agreement with that approach. Recovery brought postoperative bleeding, blood transfusions, and four days in intensive care.

Finding the cancer early did not make treatment easy. Fear, endless questions, pain, and exhaustion were all part of my recovery.

The procedures were robot-assisted, with my surgeon controlling the instruments. That is a powerful example of advanced equipment extending human capability. The FDA describes these surgical systems as tools that operate under direct human control. My operations illustrate the value of surgeon-controlled technology; they were not autonomous AI surgery.

In the ICU, however, there was a different set of needs: urgent decisions, close observation, and care informed by years of experience. I needed people who could explain the next steps, listen to concerns, and include my husband when I could not carry the full weight of the conversation.

That experience has helped shape the standards I now bring to Human Oversight Group. When we introduce AI into consequential work, we need to plan carefully for how people will work with it. Someone needs the expertise to recognize an exception and the authority to act. A review step has value only when the reviewer can exercise human judgment.

A useful deployment should show how the system behaves in its intended setting, how uncertain cases reach the appropriate expert, and how the organization learns when outcomes differ from expectations. As capabilities improve, those responsibilities should be revisited with fresh evidence.

I expect AI to do increasingly remarkable things. My enthusiasm includes a commitment to getting its use right today. Evidence that a tool can assist with a particular task does not establish that it is ready to take over an entire chain of consequential decisions. That requires testing in the intended setting and a clear understanding of where the technology can fail.

I am here because an experienced physician recognized a small finding worth pursuing, technology helped my surgeons act, and a team of humans cared for me through the hardest part. That is the future I want to help build: ambitious about what technology can do, precise about the evidence, and deliberate about the people responsible for the outcome.

Human oversight in healthcare AI

AI's ability to identify patterns and process large amounts of information offers opportunities to improve clinical and administrative work. The question is how to turn those capabilities into reliable help for the people delivering and receiving care.

My experience with a rare condition illustrates why that distinction matters. An abnormality at the opening of my appendix prompted further investigation. It took tissue evaluation and clinical expertise to establish what it meant and decide how to respond. Detection, diagnosis, and treatment were connected, but each required its own evidence and judgment.

The same attention to context belongs in AI deployment. A system's performance on one task does not settle whether it will help a particular practice, serve its patients well, or fit safely into its workflow. Those questions need to be answered before the practice relies on it.

For me, responsible adoption rests on three commitments:

  • Technical capability with clinical context. AI can help identify patterns, while qualified clinicians assess what the findings mean for an individual patient and whether they warrant action.
  • Accountability and oversight. Clinicians need the knowledge, protected time, and clear authority to question, override, or correct automated outputs.
  • Preserving the human connection. Efficiency gains should create more room for meaningful clinician-patient interactions, shared decision-making, and trust.

During my recovery, those human interactions were essential. I needed people who understood what was happening, could respond as my condition changed, and were willing to stay in the conversation. As technology advances, I believe responsibility for that care must remain with people.

Before AI enters the workflow

The business case for medical AI is understandable. Reducing administrative work can help practices function and give clinicians relief. But evaluating a product also means examining the work it is supposed to improve.

How does the process work today? Where does it break down? What must be configured for this specialty and these patients? Adding automation before answering those questions can make a flawed process move faster.

Training needs the same attention. How much protected time will staff have to learn the tool? How will they demonstrate that they can recognize an error, correct it, and escalate a concern? Completing a training session does not necessarily mean a team is prepared.

In the American Medical Association's 2026 physician survey, 85% of respondents wanted to be consulted about or responsible for AI adoption decisions, and 92% wanted more AI education and training. The people expected to use these systems are asking to help shape their use.

Oversight also needs an owner with time and authority to act. A clinician assigned to review AI output needs space in the schedule to do it. Staff need a clear way to report problems. Leaders need evidence from a limited rollout, including correction work, missed information, patient experience, and safety concerns, before expanding use.

I would measure success partly by what happens to the time saved. Does it give a clinician more room to listen and think? Or does every recovered minute become another appointment, leaving the promised human review squeezed into an already full day?

If AI saves a clinician time, patients should get some of that time back.

Helping leaders examine AI adoption before it changes the processes people depend on is central to the work I am building at Human Oversight Group. That means asking practical questions about vendor claims, staff readiness, workflow, and accountability, with qualified clinical and technical specialists contributing where their expertise is needed.

I want organizations to use AI confidently where the evidence supports it. I also believe that human accountability must endure as the technology evolves. Listening to a frightened patient, understanding what matters to them, and taking responsibility for the next decision are obligations of care we should protect.

I am grateful to Dr. Tamas Otrok of Gastroenterology of the Rockies, Warren Strutt, MD, FASCRS, of the Colon and Rectal Clinic of Colorado, and my medical oncologist and hematologist, Gregory Britt, MD, of Rocky Mountain Cancer Centers. The nurses and care teams who helped me through this chapter have my enduring gratitude. My husband, Brian, carried more than any short article can explain. This story would not be complete without noting what powerful medicine love can be.

My cancer was small. The effort required to treat it and get me home was not.

As I build Human Oversight Group, I want the benefits of remarkable technology to reach more people. I want those people to have what mattered so much to me: someone with the expertise to look closer and the responsibility to stay involved.