The feasibility questionnaire lands in your inbox on a Tuesday. By Friday, the sponsor has already “identified” your site using an AI platform, ranked your PI’s publication history, mapped your metastatic breast cancer patient volume from claims data, and slotted you into a projected enrollment curve. Congratulations — you have been selected. Now the real work begins, and none of that algorithm knew about your pharmacy’s two-week IP training lead time, your IRB’s current 47-day average review cycle, or the fact that your lead coordinator is covering two other active oncology protocols.

That is the operational gap sitting underneath the announcement that ASCO and Ryght AI are partnering to identify research sites for ASCO’s CDK4/6 Inhibitor Dosing Knowledge (CDK) Study — an active trial investigating optimal dosing strategies for CDK4/6 inhibitors in older patients with hormone receptor-positive, HER2-negative metastatic breast cancer, with enrollment expected to run through February 2027. The collaboration is genuinely interesting, and the underlying problem it is trying to solve is real. But the part of the pipeline that AI cannot yet touch is the part that actually determines whether your first patient is in at day 90 or day 167.

What the Algorithm Gets Right — and Where It Stops

Let’s be honest about what AI-driven site selection actually solves. Traditional feasibility campaigns in oncology are painfully slow. A 2018 AACI benchmarking survey of 61 cancer center members found a median trial activation time of 167 days — nearly six months from selection to first-patient-in — against an NCI target of 90 days. Ryght AI’s own case study data claims a leading global CRO completed an oncology site selection campaign in 26 days using their platform, exceeding its site recruitment goal by 330%. Those are not trivial numbers. Compressing the identification phase from months to weeks does matter, particularly for a study targeting older patients with metastatic disease where the enrollment window for eligible individuals is genuinely narrow.

But site identification and site activation are not the same operation. Identification is the algorithm’s domain. Activation is yours. The moment a site appears on the selected list, the clock shifts from data science to contract negotiation, IRB submission, SIV scheduling, IP ordering, and coordinator onboarding — none of which an AI model can execute on your behalf. Across our network at CliniBiz, the sites where startup most consistently blows past 90 days are rarely the ones that took longest to identify. They are the ones where the feasibility data looked clean but the operational infrastructure had hidden constraints the questionnaire never surfaced.

That is the structural problem with any algorithm operating on retrospective data. Claims-based patient volume tells you how many HR+/HER2- metastatic breast cancer patients received treatment at a site last year. It does not tell you how many of those patients are currently eligible for a CDK4/6 dosing study, how many are already enrolled on competing protocols, or whether the oncology coordinator who managed your last breast cancer study left six months ago. ICH E6(R3), Section 5.2 on investigator selection, explicitly requires sponsors to evaluate each site’s “resources and facilities” — the guideline has always acknowledged that historical data alone is insufficient. AI platforms are becoming better at ingesting real-time signals, but the verification step still requires a human being to pick up the phone.

This is also where budget assumptions break down fastest. Sponsors using AI-driven selection often build enrollment curves assuming the sites the algorithm identified will activate at the same rate they performed on prior studies. They will not, consistently, because the site that enrolled five patients per month on a first-line therapy study two years ago is now carrying a heavier protocol burden, has less coordinator capacity, and faces a more competitive enrollment landscape. When those curves miss, the remediation cost — rescue site activation, expedited contract amendments, emergency feasibility assessments — routinely exceeds what a slower, more operationally grounded selection process would have cost upfront.

The Activation Gap No Platform Closes

ASCO’s CDK Study is an instructive case to watch precisely because ASCO is not a traditional sponsor. The organization launched its first clinical trial, the TAPUR study, in 2015 — a basket study designed to address gaps in understanding targeted agents across tumor types. The operational learning from TAPUR, including how to manage a multi-site oncology study without a full CRO infrastructure, will matter here. Adding AI-powered site selection does not eliminate the need for that institutional memory — it makes the period between selection and activation even more operationally exposed, because the speed of identification can create a false confidence that the hard work is done.

The AI-based clinical trials solutions market was valued at $1.9 billion in 2023 and is projected to reach $7.8 billion by 2030 at a 22.1% CAGR. That growth rate guarantees that more sponsors will arrive at site selection meetings with an algorithm-generated list and an optimistic timeline. The sites on that list need to understand what they are agreeing to before the SIV is scheduled.

Here is what the activation gap looks like in practice. An AI platform selects 20 sites for an oncology study. Fourteen receive draft clinical trial agreements within the first three weeks — fast, genuinely fast. But six of those fourteen have budget templates that do not account for the CDK Study’s specific screening procedures: bone density assessments, echocardiograms at baseline, and geriatric assessment tools appropriate for an older metastatic population. The per-patient budget negotiation on those line items adds four to six weeks before signatures. Three sites have local IRBs that require full board review for a dose-optimization study in a vulnerable population — add another four to eight weeks. Two sites have pharmacies that need updated IP handling SOPs before the study drug can be received. By the time all 20 sites are activated, the “26-day selection” story has been replaced by a 140-day activation reality, and the enrollment curve the sponsor presented to their steering committee is already wrong.

What Operators Need to Do Monday Morning

For sites receiving AI-generated selection notifications: treat the outreach as a feasibility trigger, not a commitment. Before you respond with interest, run your own internal capacity audit against the specific protocol requirements — coordinator hours per patient, pharmacy lead time for IP handling, your IRB’s current review cycle for the indication, and your current active protocol count per PI. The CDK Study’s February 2027 enrollment deadline sounds distant until you account for a 90-to-120-day activation window and a competitive oncology enrollment environment. If your IRB is running 47 days and your contract team has a backlog, that deadline math changes fast.

For sponsors deploying AI site selection on this study and others: build a mandatory post-selection operational verification step into your activation workflow before the contract clock starts. A 30-minute pre-contract call with the site coordinator and pharmacy lead, specifically covering IP handling readiness and IRB submission pathway, will surface the constraints the algorithm could not see. Sites I work with that have implemented this as a standard pre-SIV checkpoint report an average of three fewer contract amendment cycles per study — not because the protocol changes, but because the operational constraints get negotiated once, correctly, before signatures rather than twice after the fact at higher cost to both parties.

The AI market will keep growing, the platforms will keep improving, and ASCO’s CDK Study may prove exactly what Ryght AI’s partnership intends to demonstrate. But the 167-day activation median is not a data problem. It is an operational handoff problem — and no algorithm hands off to your IRB, your pharmacy, or your coordinator’s calendar. Watch what ASCO’s activation timeline actually looks like when it publishes enrollment updates in late 2026. That number will tell us far more than the selection story.

References

  1. FierceBiotech — “ASCO, Ryght AI team up for breast cancer trial site selection”
  2. FirstWord Pharma — ASCO CDK4/6 Inhibitor Dosing Knowledge (CDK) Study enrollment details
  3. Ryght AI — CRO oncology site selection campaign case study: 26-day completion, 330% goal exceeded
  4. AACI — 2018 benchmarking survey: median trial activation time 167 days; NCI 90-day target
  5. ASCO — TAPUR Study: ASCO’s first clinical trial, announced June 2, 2015
  6. Grand View Research — AI-based Clinical Trials Solution Provider Market: $1.9B in 2023, projected $7.8B by 2030 at 22.1% CAGR
+ posts