The Real Problem In Patient Recruitment… Is Knowing How And When To Use AI

By Vanessa Watson, COO, Sequoia Clinical Research

Ask anyone who has run a clinical trial about the main obstacle in meeting an enrollment timeline, and their answer is rarely the protocol. It's recruitment. Patient enrollment remains the single most common cause of trial delay. More than 80% of clinical trials fail to meet their original recruitment timelines, and low patient recruitment impacts trials across virtually all medical and therapeutic areas.¹ Most sites already have large patient databases. The problem is getting to the bottom of the large funnel quickly. A patient who's interested today can be unreachable tomorrow.

Most sites are fighting the same seven battles:

• Enrollment contracted number based on inaccurate data
• Searchable-but-disorganized data
• Slow first-contact times
• No visibility into where patients drop off
• Overloaded recruiters
• Static and inflexible resourcing  
• Single-study patient evaluations, when   patients could qualify for more

Tufts CSDD has estimated that the average cost of a single day of delay in a clinical trial program is several hundred thousand to over $1 million dollars, depending on the therapeutic area. A single day of delay is worth approximately $800,000 in lost prescription drug or biologic sales.²  These are real, expensive problems, and AI can close some of the gaps but not all of them.

Where AI Genuinely Helps

AI helps find patients faster. Acting as a tool that scans EMRs and other clinical data to identify potential patients, AI is faster than clinic staff combing through charts. The result isthe ability that sites can more accurately predict enrollment and reach more participants. AI can rank patients by likely eligibility, response history, location, and availability, giving patient recruiters a smarter starting point instead of a random list. It can also check a single patient against every open protocol at the site at once, so someone who doesn’t qualify for one study can be matched to another one instead of being turned away. AI can also decrease the burden of documentation bydrafting screening summaries and pre-populating eligibility checklists so recruiters can spend more time talking to patients and less time digging through medical records. Because these systems generate data as they operate, sites gain earlier insight into what is driving enrollment instead of in their lessons-learned meeting after a study has finished enrolling.

Each of these is a genuine improvement. None of them, on their own, solves recruitment.

The Part AI Doesn't Touch

AI is an efficiency layer, not a substitute for the trust a human recruiter provides.

Building trust in clinical research is a specific, learned skill, and it looks nothing like pattern matching. It's noticing hesitation in a patient's voice. It's adjusting an explanation of how the study medication works to match someone's health literacy, then pausing to see what aspects make them apprehensive. From my experience, it’s sitting across an exam room from a  a patient, listening to them, answering their questions and recognizing when they need the permission to loop in a spouse or adult child before they can say yes and knowing not to rush that decision. Unguided AI outreach can also create distrust: Some patients worry about how their health information is being used, and that HIPAA concern can show up as higher no-show rates. An experienced recruiter, on the other hand, can professionally outline the potential risks and benefits while explaining the regulatory, HIPAA protections, and medical oversight.

That trust extends through the full course of a study. Retention depends on the same relationship that drove enrollment. When my study coordinators give participants feedback on their health status, such as lab results and ECGs, or provide dietary counseling, and  answer their questions, participants stay engaged because they can see the site paying attention. Asking and answering patients’ questions takes empathy and judgment. In other words, a human being on the other end of the line. Trust is not just what gets a patient to say yes, it is what keeps them coming back for visits.

Informed consent is the clearest illustration of this, because speed is explicitly the wrong metric, by regulatory design. Consent exists to confirm understanding and willingness to proceed given this understanding. It is meant to be the first of many conversations and is not a checkbox.

A peer-reviewed literature review published in the International Journal of Medical Informatics in October 2025 found that, alongside technical hurdles such as data interoperability and regulatory uncertainty, limited stakeholder trust and algorithmic bias are major barriers to AI adoption in clinical trials.³ Those last two are human and ethical problems, not technical ones. This aligns with what clinical research coordinators see on the clinic floor: A patient who feels processed rather than cared for is a patient who either says no or says yes without truly understanding what they agreed to.

Guardrails And Regulatory Alignment

Any site adopting AI recruitment tools should build in a few guardrails.

Privacy comes first. Sites need clear data-handling policies and vendor agreements that limit how patient data can be used, stored, and retained, rather than leaving those terms to a vendor's default settings.

Human review should be a hard rule, not optional. No AI-surfaced match and no AI-drafted communication should be communicated to a patient without review by trained staff. AI narrows the list; the people decide.

Transparency matters. Sites should disclose when and how AI is being used in recruitment outreach or data record matching not buried on page 42 of a consent form.

Bias auditing is not optional. Because these tools learn from historical data, sites should periodically check whether list narrowing or outreach patterns are inappropriately excluding specific patient groups.

Vendor accountability closes the loop. Before adopting a tool, sites should ask vendors to explain its limitations and error rates, not just its headline accuracy number, which tells you almost nothing about how it fails. 

Patients need a clear human escalation path. Patients should always have an easy way to reach a real person if an AI-driven interaction doesn’t answer their question.

This guardrail framework also sits inside a shifting regulatory backdrop. On December 15, 2025, the FDA finalized guidance, pushing sponsors to justify restrictive eligibility criteria rather than defaulting to them, across both demographic and non-demographic factors.⁴ Separately, under Sections 3601 and 3602 of the Food and Drug Omnibus Reform Act of 2022 (FDORA), sponsors of Phase 3 drug studies and certain device studies will be required to submit Diversity Action Plans with enrollment goals broken out by race, ethnicity, sex and age group. FDA’s implementing guidance, issued in draft in June 2024, has not been finalized, and the requirement applies only to studies that begin enrolling 180 days after it is.⁵ The December guidance asks sponsors to revisit exclusions that are often applied without strong clinical justification, such as those affecting older adults, people with organ dysfunction, and children. An AI matching tool trained on historically narrow trial populations could just as easily reproduce the representativeness problem regulators are now asking sponsors to fix. For sites, that makes AI governance part of regulatory alignment not just operations. The guardrails above are how a site demonstrates it: Bias audits that check whether AI-narrowed lists reflect the populations a sponsor needs to enroll, human review of eligibility logic before it screens anyone out, and enrollment data that can be compared against a sponsor’s demographic goals.

Speed + Trust: A Dual-Engine Model

What does a recruitment process look like when speed and trust both get the attention they deserve? It starts with giving each engine its own job. AI runs the top of the funnel, where volume and speed matter most. People run everything that depends on judgment and relationship. In practice, the hand-offs look like this:

• Identification and matching. AI scans records and flags likely matches across all open protocols. A coordinator confirms each match before anyone is contacted.
• First contact. AI prioritizes the list and drafts outreach. A person makes the call or approves the message, so the first contact happens in hours rather than days without losing the human voice.
• Prescreening documentation. AI pre-populates eligibility checklists and screening summaries. Staff verify them rather than build them from scratch.
• Informed consent. Fully human-led. This is the step where taking time is the point.
• Retention. Human-led, with AI-generated data helping coordinators spot participants who may be at risk of dropping out so they can reach out early.

Measured this way, success isn’t just a shorter time to first contact. It is enrollment that holds: participants who understood what they agreed to and keep coming back for visits.

AI tools can find a qualified patient faster, but they can't earn that patient's trust. That part is still, and will remain, a human job.

References:

1. Mira Desai, “Recruitment and Retention of Participants in Clinical Studies: Critical Issues and Challenges,” Perspectives in Clinical Research, May 6, 2020. https://pmc.ncbi.nlm.nih.gov/articles/PMC7342339/
2. Zachary Smith, Joseph DiMasi and Kenneth Getz, “Quantifying the Value of a Day of Delay in Drug Development,” Tufts Center for the Study of Drug Development white paper, August 2024. https://csdd.tufts.edu/sites/default/files/2025-02/Aug2024%20Day%20of%20Delay%20White%20Paper%20Final.pdf
3. David Olawade et al., “Artificial Intelligence in Clinical Trials: A Comprehensive Review of Opportunities, Challenges, and Future Directions,” International Journal of Medical Informatics, October 2025. https://www.sciencedirect.com/science/article/pii/S1386505625003582\
4. U.S. Food and Drug Administration, “Enhancing Participation in Clinical Trials — Eligibility Criteria, Enrollment Practices, and Trial Designs,” guidance for industry, Dec. 15, 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/enhancing-participation-clinical-trials-eligibility-criteria-enrollment-practices-and-trial-designs
5. U.S. Food and Drug Administration, “Diversity Action Plans to Improve Enrollment of Participants from Underrepresented Populations in Clinical Studies,” draft guidance for industry, June 26, 2024. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/diversity-action-plans-improve-enrollment-participants-underrepresented-populations-clinical-studies

About The Author:

As COO of Sequoia Clinical Trials, Vanessa Watson brings more than 25 years of clinical research experience spanning operations, process development, recruitment, compliance, project management, and sponsor and CRO collaboration. Vanessa leads operational strategy across the Sequoia network, supporting consistent processes, efficient study execution, quality, and continued growth.