PSI and Appian Automate Site Selection: Faster Process, Same Data Question
Published November 2026 — On October 28, 2026, PSI CRO and Appian launched an AI-driven workflow for identifying and selecting trial sites. It is one of several such launches this year. For sponsors, the useful question is not how fast the process runs but what data it runs on.
A CRO Puts Site Selection on Rails
PSI CRO and the software company Appian announced a strategic partnership on October 28, 2026, along with their first joint product: an AI-powered process that automates site identification and selection across therapeutic areas, indications, and geographies. The companies describe it as live, with further uses to be evaluated.
PSI's stated goal is to further reduce or eliminate non-enrolling sites and enrollment delays. That goal is as old as the industry's data on the problem. This case study looks at what was announced, what the published evidence says, and what a sponsor should ask of any automated site selection process.
Key Findings
A Tufts CSDD analysis of nearly 16,000 sites in 151 Phase II and III trials, run from 2008 to 2010, found 11% of activated sites failed to enroll a single patient.
In the same analysis, 48% of all sites under-enrolled against their targets, and 53% of studies ran past planned timelines.
A 2025 review in Applied Clinical Trials reported that many trials continue to see 30% to 40% non-performing sites.
Neither company has published outcomes for the new process. The announcement describes what it does, not what it has achieved.
What Was Announced, and What Was Not
The announcement is specific about workflow and less specific about data.
What the Product Does
According to the companies, the process consolidates steps that were previously manual into one digital workflow, from project initiation through the stages of site selection. It combines Appian's AI and process tools with PSI's core systems and data warehouse.
What It Draws On
The only data source named is PSI's own. That is a reasonable foundation for a CRO, which knows how the sites it has used have performed. It also means the process is strongest on sites PSI already knows. No external data sources are named in the announcement.
Why Data Matters More Than Speed
The 2025 review found that inaccurate feasibility projections were the most common concern in the literature, raised in 9 of the 14 papers it examined, and noted that the inputs to site selection are snapshots that change during a trial. Automating a process does not improve the information fed into it.
What Prediction Can and Cannot Do
A 2024 study in PLoS ONE by Janssen researchers found that site questionnaires often overestimate recruitment, and that a machine-learning model outperformed a baseline built on historical medians. The authors also reported that the model's explanatory power remained low to modest. Better tools help, within limits.
What This Means for Sponsors
When a CRO presents a site list produced by an automated process, three questions are worth asking.
Ask whether the list reflects only the CRO's own history or also draws on independent sources, and how sites the CRO has never used are assessed.
Self-reported capabilities and historical counts are not the same as checked evidence. Ask which inputs have been verified against documents and how recently.
Ask for the measure: the share of selected sites that enroll to target. A tool that shortens selection but leaves that number unchanged has saved time, not solved the problem.
Conclusions
PSI and Appian are doing what much of the industry is doing in 2026: taking a slow, manual process and making it faster. That has real value. Site identification and selection has been measured in months for years.
Speed is the part of site selection that technology can most easily fix. The harder part is knowing, before selection, which sites will actually enroll. That depends on the quality and independence of the underlying data.
An Independent Check on Any Site List
SiteScore™ is not tied to any one CRO's history. Every site is scored on the same 0–100 scale across execution, operational depth, market access, financial discipline, and risk and compliance. A PreScore uses public data, and a Validated Score is checked against the site's own documents.
Sponsors can use it alongside a CRO's recommendations: to check a proposed list, to assess sites the CRO has not worked with, and to compare candidates side by side on verified evidence.
This industry case study is based on published announcements and industry research. Figures are reported as stated in the sources below and have not been independently verified by Clinitiative. Companies named are not Clinitiative clients unless stated.
- Appian and PSI CRO: Strategic partnership to advance AI-driven clinical trial operations (October 28, 2026)
- PSI CRO: Appian and PSI CRO launch AI-powered clinical trial site selection
- Applied Clinical Trials: Enrollment performance: weighing the facts (Getz, 2012)
- Applied Clinical Trials: Beyond prediction: an evidence-based framework for assessing site selection decisions (2025)
- PLoS ONE: Hulstaert et al., machine learning for site enrollment prediction (2024)
See How SiteScore™ Works
Standardized, document-verified scoring for clinical research sites, used by sponsors, sites, and investors.