Artificial intelligence (AI) was a hot topic on the agenda at the American Society of Clinical Oncology (ASCO) meeting 2026, with its benefits in unstructured data continuing to bear fruit.
Experts highlighted a rise in the number of AI-related abstracts in 2026 at the ASCO meeting, which was held in Chicago, Illinois, from 29 May to 2 June, with the technology being utilised at all different stages of drug development, from identifying key targets to analysing clinical data.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
According to GlobalData, parent company of Clinical Trials Arena, venture financing deals involving AI have experienced more than a 400% increase between 2014 and 2024, showing how industry confidence is growing in the technology.
While the technology has existed in oncology for over a decade already, Dr Vivian Yin, practising ophthalmologist specialising in orbital oncology and medical director for Intelligent, said that there has been an increased willingness in recent years to trust and utilise AI in clinical settings.
As a result, this higher trust level is enabling wider clinical and research adoption of AI tools. One of the main areas of challenge in oncology is the interpretation and structuring of real-world data (RWD) and real-world evidence (RWE), in which AI is now showing benefit.
RWD analysis supported by AI
AI’s speed positions it as a force multiplier for clinical trial teams, especially when evaluating RWD, as the volume of data is much larger than in traditional trials. When not utilising technology, Yin said that human teams are overwhelmed by the number of relevant trials while working with limited resources.
AI will be most beneficial in unstructured data, Dr Jessica Paulus, vice president of real-world research at Ontada, believes. She highlights that often, critical oncology information, such as biomarkers, treatment patterns, and outcomes, is stored in progress notes and PDFs from external facilities. Historically, this required manual chart abstraction by clinicians, which is time‑consuming, expensive, and still imperfect.
AI, however, can be used to extract structured concepts at scale from unstructured records, enabling faster, broader real‑world analyses.
The major challenge with this is ensuring the data is validated and is “research‑grade”.
“Just like we do with manual abstraction, we must validate that it’s coming up with good enough research‑grade information. We must subject it to validation procedures to make sure that we’re getting the right information,” Paulus explained. “Even human abstraction is not perfect, and it’s important to keep that in mind. When it comes to human extrapolation, not only is it expensive and time-consuming, it’s also imperfect.”
AI could therefore reduce the time burden and allow more time for oversight of the extracted data to ensure accuracy.
Yin agrees, adding: “Clinicians just don’t have the time, and AI will allow us to streamline that. I don’t think it’s going to completely replace the need for clinical trial assistance or research coordinators. We still need them a lot, but I think what it does is it really helps cut back, the low-hanging fruit, as the more high-volume work will be replaced by AI, but you still need the human brain or the human interface between that AI and unstructured data scrubbing.”
Scribe software use on the rise
One way that AI is being used to collect RWD is through ambient listening software and scribe software.
Dr Debra Patt, medical oncologist in haematology at Texas Oncology-Austin Central, explained that this software is already being used, with her site using DeepScribe ambient AI software. Ambient listening in oncology clinics will transcribe and structure the live patient–oncologist conversation.
Patt said this allows her and her colleagues to focus fully on patients in the room, which can bring patients closer to their doctors, rather than replace the relationship.
She does, however, express concerns about the potential for hallucinations and errors, and notes that clinicians should not use AI to replace clinical judgment and the doctor-patient relationship must be preserved.
Patt also highlighted that it is vital that clinicians ensure a patient properly understands how and what data is being used by AI. HIPAA and privacy protections should be properly explained to put patients at ease, and she has found that once these boundaries are explained, most patients are put at ease.
Paulus said she is particularly excited about how the technology can capture richer and less pre-filtered information than traditional clinician notes and eliminate potential opinion and bias. She explained that one of her colleagues automatically applies their assumptions in clinical notes.
Paulus said: “He said, ‘when I write up my notes after a patient encounter, I’m essentially already processing that information and applying my own filters and assumptions, and of course, you know, I’m already analysing the data in some sense’. There may be more richness that could be beneficial to capture without that element of pre‑processing from the oncologist.”
She also believes this kind of use will ensure a higher dimensionality and quality of RWD and could vault oncology into a new era of RWE generation.
Paulus does caution this excitement with implementation concerns, however, primarily from a patient perspective. This includes data protection and privacy, with Paulus highlighting that oncology discussions are highly sensitive and that some patients may not wish for their sessions to be recorded. On top of this, while patients may accept recording sometimes, they may reject it at others.
Patients and providers may also change how they speak if they know conversations are recorded and processed, with possible negative consequences.
Yin touts AI scribe tools as one of the “hottest” applications of AI in clinical workflows, but notes that there are many competing products with no clear superior solution at this stage. She states that there does need to be a clear accuracy benchmark that AI scribe outputs must reach in order for them to be confidently utilised by oncologists, given the amount of information the patient will discuss in appointments.
Oncology patients also utilising AI
As AI becomes more public-facing through large language model (LLM) agents such as ChatGPT and Claude, patients are also utilising AI; however, this can be dangerous, warns Patt.
Patients may not know the right questions to ask an LLM, and they may not interpret the responses properly, highlighting the continued need for clinician involvement.
Patt shared an example from the clinic where a breast cancer patient used ChatGPT to interpret a molecular test, which stated she had a low risk of needing certain treatments. Patt stated that missing data from the request means that the treatment course is different.
“I said to her, ‘when you have a one millimetre margin with your cancer,’ which she didn’t put in the model, ‘your risk of recurrence without radiation is 31% and that that would really influence my choice much more so than your molecular characteristics’, so we spoke about that a bit more so she could make a more informed choice,” Patt explained.
AI is really becoming a more day-to-day tool in the clinic by physicians, helping to bolster richer RWD, but it is still in relatively early stages, with several hurdles that must be overcome. While there is more trust in the software, it is not completely trusted yet, but this could be a good thing, as it retains a need for oversight. When trust is complete, errors may occur without being noticed, and this could be more dangerous for patients.
