Responsible AI Use Ushers in a New Era of Oncology Practice and Patient Care

Douglas Flora, MD, LSSBB

Artificial intelligence (AI) platforms, such as generative pre-trained transformers (GPTs), as well as next-generation sequencing (NGS) and other genetic testing methods, have greatly increased the extent to which cancer care can be personalized and patient-focused, although the clinical use of these tools must acknowledge their current limitations, according to Douglas Flora, MD, FACCC, LSSBB.

“For me, in 2025, [the role of AI in oncology is] pattern recognition and addressing some of the rote, mechanical [tasks] that eat up our days,” Flora said in an interview with OncLive®. “I’m looking forward to future years, where [AI] might help us with making better decisions as well.”

In the interview, Flora explained how tools such as Tempus' NGS-based tests and xM and xM NeXT Personal Dx MRD assays have enabled more intelligent treatment plans, praised AI for its pattern recognition capabilities, and emphasized the role of AI in reducing oncologists’ data overload. He also stressed the importance of ensuring that AI use disseminates accurate data and urged the oncology field to keep pace with new technological developments and approach AI with a balanced level of optimism.

Flora is the executive medical director of Oncology Services and the Robert and Dell Ann Sathe Endowed Chair in Oncology at St. Elizabeth Healthcare in Edgewood, Kentucky. He is also the editor-in-chief of the peer-reviewed research journal AI in Precision Oncology.

OncLive: How have the increasing roles of molecular markers and genetic testing evolved to individualize the treatment of patients with cancer?

Flora: We’ve seen a revolution over the past approximately 5 years [in] cancer medicine, as treatments have truly become personalized. Each patient now has the opportunity to undergo NGS and identify the Achilles heel of their cancers. We can approach cancer with more care, more intelligently designed treatment plans, and—hopefully—fewer toxicities.

How has AI been integrated into the oncology field so far, and where might this technology continue to find a foothold based on the trajectory of its capabilities?

It’s been remarkable how quickly some of these [AI] tools have been adopted. Today, AI is a wonderful pattern recognizer, and the big data [AI works with] can help oncologists regarding our ability to find signals in the noise. For example, the ability to find MRD is arriving on our doorsteps, and [MRD status could] guide neoadjuvant or adjuvant therapies; AI sorts out the signal from the noise there.

A lot of us are using tools to help screen CT scans for incidental nodules and detect findings we couldn’t see otherwise. [Additionally], diagnostic mammography is a nice place to apply this pattern recognition. We’re also seeing a lot of improvements in pathology as these tools augment the abilities of the reading pathologists.

How quickly have you seen AI technology evolve from its early days?

With the arrival of GPTs, we saw an evolution in digital health care. Now, with [most people] having access to one of [several] active GPT models, [AI is] entering every possible realm of medicine, ranging from screening for patients who might be eligible for clinical trials by scraping the data from the Epic [Clinical Trial Management System], to treatment selection and everything in between. For health care providers, the most exciting [developments in 2025] will probably be [the increasing number of] us having access to the tools that will record our visits with patients and document the important topics that were discussed in that room. [That will] hopefully save us time and burnout by [preventing] us from staring at the computer all day and letting us look at our patients once again.

Given AI’s potential to guide treatment planning, how might these tools help community oncologists sort through the wide scope of available information?

Oncologists today are buried in data that are siloed, hard to [access], and hard to digest. There [were approximately] 144,000 articles indexed to ‘oncology’ on PubMed [in 2024]. There’s no way that even the most discerning, diligent oncologists can keep up with that deluge.

AI might help us learn faster and smarter by giving us access to all the important information we need at a moment’s notice, whether that be the complete and thorough patient chart in front of us, or all the most up-to-date data from the most recent congresses. It can help us digest articles and form executive summaries so we can file those away electronically or print them. [AI decreases] the time to read and understand an article from approximately 30 minutes with a highlighter to maybe 5 or 10 minutes, and [we see] the salient points and the statistical analysis for strengths and weaknesses [of the investigational treatments] all at once.

What does the oncology community still need to understand about applying AI tools to real-world clinical practice?

Oncologists have to be sober and responsible about [AI] before we let it start guiding treatment decisions. A lot of us are getting involved early on, trying to understand the nomenclature and how these models work. The reason I started AI in Precision Oncology was as a way to show validated data in a peer-reviewed setting, so oncologists can trust what they read rather than just listening to what vendors are telling them.

For me, [AI is] going to be instrumental in taking better care of patients. It’s going to be fundamental in giving us hours back in our days. I’m excited about the promise, but I want doctors to understand the tools, the bias that can happen, and the importance of the training set that goes into [using] those tools so they can be sober in their application of these technologies.

How do you explain the potential role and benefit of AI tools to patients?

I explain that AI is a tool [for oncologists], just like an Excel spreadsheet is a tool for a chief financial officer. We need to be savvy. We need to be well-trained to use the tool and understand its promise and peril.

A lot of my patients are coming into the clinic with GPT-generated lists of questions for me, or they’ve already studied and read summaries of their disorders generated by GPT models before I even meet them. Patients are also learning [about these tools] in their daily personal or professional lives. My hope is that we can all start to train each other as [these tools evolve]. The role for me as a physician is to make sure the data are accurate, that there are no hallucinations, that we’re interpreting [findings] with care and diligence, and that we use whatever tools are at our disposal to make the care of the patient the center of the discussion.

What is your main message about how novel genetic tests and AI models are being integrated into the oncology field and where the future of these tools may be headed?

It’s our responsibility to keep pace with the new knowledge that’s available to us, just as it was when we had to relearn genetics or relearn immunology as the science of oncology evolved. [AI] is just that next frontier. If we approach it with optimism and skepticism at the same time our patients are going to benefit. I’m leaning optimistic because I’ve studied [AI] a lot for the past 5 or 6 years, and I see the power of these tools. I just want to make sure we approach them responsibly.