AI can summarize evidence, but it can't replace physician expertise

A senior physician recently told me something that many in medicine now believe:

"AI will replace all such platforms."

At first glance, this sounds reasonable. AI tools are improving rapidly. Platforms like OpenEvidence can already summarize papers, answer clinical questions and surface guidelines in seconds.

But this argument fundamentally misunderstands what platforms like Open Medicine actually are.

Open Medicine is not "ChatGPT for doctors." It belongs to an entirely new class of applications that were designed to leverage AI, not compete with it.

This distinction makes all the difference.

AI can summarize papers

AI is extraordinarily good at finding information.

Search the literature. Summarize a trial. Pull NCCN recommendations. Compare regimens. Surface evidence faster than any human ever could.

This is incredibly valuable.

But medicine isn't just about retrieving information. Real clinical practice lives in the grey zones:

• What do you do when the patient doesn't fit the trial criteria?

• What matters when two options are both "reasonable?"

• How do experienced physicians sequence therapies in the real world?

• What tradeoffs actually influence decisions at the bedside?

• What do experts do when evidence is incomplete, conflicting or rapidly evolving?
  

These are not search problems. They are judgment problems.

And judgment comes from clinicians. Not from AI.

AI can't replace lived expertise

The most valuable knowledge in medicine doesn't exist in papers. It exists in:

• Pattern recognition

• Tacit clinical reasoning

• Accumulated real-world experience

• Treatment sequencing instincts

• Practical heuristics

• Multidisciplinary nuance

• Edge cases

• Failures

• Intuition developed over thousands of patient consults
  

This is why tumor boards still exist. Why fellows seek mentorship. Why community oncologists call subspecialists. Why physicians follow experts on X.

Medicine has always depended on networks of human expertise. Open Medicine represents the next generation of these networks.

Living Algorithms are dynamic, not static

Traditional guidelines are static PDFs. AI search tools help navigate them faster.

But Living Algorithms are fundamentally different. They are:

• Dynamic

• Physician-authored

• Continuously evolving

• Experience-driven

• Interactive

• Opinionated when appropriate

• Transparent about uncertainty

• Designed around real clinical workflows
  

Importantly, they make expert reasoning visible. This is the key difference.

A physician using a Living Algorithm is not merely seeing "the answer." They are seeing how an expert thinks through a problem.

That is educational. That is scalable mentorship. That is clinical cognition made accessible.

And AI will make this more valuable, not less.

The future isn't AI vs. physicians

The future is physicians amplified by AI.

The strongest medical platforms of the next decade will combine:

• AI-powered retrieval

• Real-world physician expertise

• Structured reasoning

• Continuously updated workflows

• Transparent clinical judgment
  

In other words: AI handles information abundance, while physicians provide meaning, prioritization, context and wisdom.

Even leading AI companies increasingly recognize this distinction. Many are shifting toward expert-guided systems, specialty-specific reasoning models, and clinician-centered workflows rather than pure autonomous automation.

AI will replace legacy tools

AI will replace many tasks and tools that are now obsolete:

• Inefficient search

• Manual literature review

• Basic summarization

• Repetitive administrative work (prior auth)

• Low-level information retrieval
  

But AI does not eliminate the need for trusted human expertise. If anything, the explosion of AI-generated information increases the importance of:

• Trusted experts

• Transparent reasoning

• Curated clinical judgment

• Physician communities

• Explainability
  

When everyone can generate answers instantly, credibility becomes more important.

What AI can't replace

At a fundamental level, AI cannot replace:

• Mentorship

• Trust

• Lived clinical experience

• Judgment under uncertainty

• Nuanced tradeoffs

• Human credibility

• Real-world expertise networks
  

And it can't replace physicians teaching other physicians.

Open Medicine doesn't compete with AI. It's what medicine looks like after AI.