If you're deciding whether to bring AI into your admissions process, the most important question may not be which system to use. It may be whether you can explain how that system works. And once it's in place, would you be comfortable showing an applicant the evaluation it produced?
Admissions professionals hear a lot of vendor pitches, each promising streamlined review, better outcomes, and integrated data for better decisions. Some of those promises are real. But after 35 years in admissions, including nearly a decade at Dell Medical School, I've learned that the tool matters less than whether you can stand behind what it does.
One Size Does Not Fit All
No two medical schools admit students the same way. Each institution's process reflects its mission, its community, and the physicians it hopes to train. Accreditors expect this, and the courts have long recognized that admissions decisions are individualized and grounded in institutional mission.
So why would we assume an AI tool can step in for faculty review straight out of the box? If you can't shape what it values, you've adopted someone else's admissions philosophy without noticing.
It's Not Enough to Say "We Use AI"
You need to be able to explain how you use it. Before adopting any tool, ask five questions:
1. Can I configure it to reflect our mission? If you can't see or adjust what it looks for, it isn't your process anymore.
2. Is the evaluation repeatable? The same application should produce the same result every time it's run.
3. Can I see the evidence behind the evaluation? Every rating should trace back to something specific in the application.
4. What record does the tool keep, and who owns it? That record is part of your admissions file, whether you planned for it or not.
5. Could I defend it under scrutiny? When an accreditor or a subpoena asks what the institution considered, "the software decided" is not an answer. A person must remain accountable for every decision, and the tool should be tested regularly to make sure it treats applicant groups fairly.
The Question That Covers Them All
All five questions point to one simple test:
Would I feel comfortable showing an applicant their AI evaluation?
If the answer is yes, it's because you understand what the tool considered, you trust that it's consistent, you can trace its reasoning, and it reflects your mission. If the answer is no, that hesitation is worth paying attention to.
Large language models are a transformative technology, and they will change our profession. The question is no longer whether admissions offices will use AI. It's whether we understand it well enough to use it responsibly.
No two schools admit the same way. Why would one AI tool fit them all?