ChaiTalks
← All prototypes

Prototype · 11 Nov 2026

VCTA — Virtual Clinical Trial Agent

An AI agent that stress-tests a clinical trial design using simulated participants, spotting red flags before the first real patient is enrolled.

Prototype, not a product. This is a personal learning experiment built with public or simulated data. It is not clinically validated, regulator-approved or intended for real-world decisions.

The question

Many clinical trials run into trouble that could have been predicted at the design stage: eligibility criteria that are too strict, visit schedules that are too demanding, or assumptions about recruitment that don't hold up.

What if you could rehearse a trial before running it?

What I built

VCTA is a personal prototype that takes a trial design and runs it against a population of simulated (virtual) participants. It then highlights:

  • Red flags to fix before the first patient: design choices likely to slow recruitment or cause drop-outs.
  • Checkpoints to keep the trial achievable: where the plan needs a safety margin.
  • Scenario comparisons: how results might change across different sites or assumptions.

What I learned

Agentic AI is most useful here as a thinking partner for the design team, not as a decision-maker. It surfaces questions earlier; people still have to answer them.

Limits

This is a learning prototype built with simulated data. It is not validated against real trial outcomes and should not be used for real study decisions.

Have thoughts on this approach? Share your view.