AI-Powered Study Delivery

PicnicAI, 2026 · Lead product designer

PicnicAI delivers clinical research with an AI-powered platform in an industry where manual data entry is still standard practice. Agent-led abstraction has reduced costs by over 60% and increased data quality by 20%.

Part of study delivery is finding evidence in the medical records patients share when they join a study, to answer the study’s questions. This is called abstraction. Pulling each answer out of a set of records is skilled manual work, done about 3,500 times per study.

This was the first end-to-end workflow we re-thought to be AI-powered, and I set the experience model for handing that work to agents. Instead of working task by task, people now monitor the pipeline stage by stage and step in only for the exceptions, with a focus on giving them enough visibility to trust what the agents are doing.

An agent’s uncertainty note explaining how it estimated a medication end date, beside the answer it proposed
Delivery overview with tasks by stage, quality and escalation rates, and the failed tasks waiting in each stage
A study overview and its enrollment dashboard, with cumulative enrollment against target and the participant funnel