Every health/AI data company sitting on oncology notes is eyeing clinical trial matching right now. Most are about to spend 2 years solving the wrong problem.

The pitch sounds airtight. Patients miss trials because nobody can screen thousands of eligibility criteria against messy charts. AI can. Build the matcher, close the gap. True, as far as it goes. Which is about 20% of the way.

But, watch what actually happens after the match arrives. An oncologist gets a flag: "Your patient may qualify for TRIAL-XYZ." That's when the clock starts, and nothing about the matcher helps with what comes next.

Problem one: the portfolio itself. Trial portfolios turn over constantly. Arms close, cohorts open, amendments rewrite eligibility mid-stream. No clinician holds all of this in working memory, and no one is paid to try.

Problem two: there is no action path. Every protocol is its own machine. Washout periods, mandatory biopsies, screening windows, referral logistics, who to even call. The requirements for the patient and for the physician change trial by trial.

Mastering one protocol takes real time. In the community setting, that time is unpaid while physician productivity is tied to patient volume. The physician who slows down to work up a trial gets financially punished for doing the right thing.

So a perfect match can still die on the desk. The physician has a trial name and a matched patient, and still faces hours of discovery work before anyone can say a word to that patient.

If you're building in this space, clinical trial matching is table stakes. The product that wins converts a match into an executable next step. What to order, who to call, what to tell the patient, this week. The biggest barriers to trials are still wide open. They just aren't the demo ones.

Charles Jiang, MD, MPH, is an Assistant Professor in the Department of Internal Medicine at UT Southwestern Medical Center, specializing in genitourinary medical oncology.

Disclaimer: This commentary represents the views of the author and may not necessarily reflect the views of ASCO, Conexiant, or ASCO AI in Oncology.

Have you experienced similar challenges with clinical trial matching or has it worked well at your practice? Share your own AI experiences with the oncology community.