Methodology
The 132 AI Curiosity Indexes, and the engagement engine that acts on them.
Every claim VertisPro makes about finding patients faster, ranking sites more accurately, or catching a referral gap earlier traces back to the same underlying system: continuous AI analysis of how patients and caregivers actually research a condition and its treatments.
We track multi-modal data: online searches, social media interactions, scientific publications, patient forums, and patient-generated content, across hundreds of distinct curiosity indexes, in real time, across multiple regions.
Patient research behavior changes shape as someone moves through their diagnostic and treatment journey. We track it at each stage:
| Stage | What it signals |
|---|---|
| Symptom Curiosity | A patient is experiencing symptoms or has been prescribed a lab test |
| Condition Curiosity | A patient has just received an initial diagnosis |
| Treatment Curiosity | A patient is actively evaluating treatment options |
| Trial Curiosity | A patient is seeking cutting-edge or alternative treatments, including trials |
That progression is what makes the data actionable rather than descriptive: it doesn’t just say patients with a condition exist somewhere, it says where they are right now in the decision that determines whether they’ll ever reach a trial.
One signal does not mean one patient. A curiosity signal represents patient research activity in a given region and stage, not a headcount. We build regional demand quantification, site rankings, and referral-node maps on top of aggregate patterns across hundreds of indexes, not on treating any single signal as a confirmed individual patient. We say this plainly because it’s the honest description of what the data is, and because it’s also why the intelligence holds up: it’s built on convergent signal across multiple independent behaviors, not a single noisy proxy.
Claims data shows where patients were treated 6 to 18 months ago. Site databases show how a site performed 1 to 3 years ago. Both describe the past. Curiosity Indexes describe what patients and caregivers are researching in the last 30 to 90 days, which is the window that actually predicts whether a patient reaches your trial or is lost to standard-of-care first.
Curiosity Indexes answer where and when to look. A separate engagement engine answers what happens the moment a patient actually responds, and it’s a distinct part of the platform worth explaining on its own.
Every message a patient or caregiver sends back, on WhatsApp or any other channel we run, is read and understood the instant it arrives, in the context of everything that patient has told us before. That’s what makes always-on triage and a real response-time target possible: no message waits in a shared queue for a coordinator to become free.
A patient’s engagement is treated as one continuous relationship for as long as they’re in the funnel, not a series of disconnected chats that reset whenever they go quiet for a few days. The system keeps a running understanding of where that patient actually is, what’s been asked, what’s been answered, and how their intent has shifted, so a message that arrives three weeks in is read against the full history of that conversation, not evaluated cold. This is what makes trust-building over a genuinely long-term conversation practical at the scale of tens of thousands of patients, rather than something that only works one relationship at a time.
Every patient message is processed for identity protection before any part of it is analyzed: names, phone numbers, and emails are cryptographically tokenized first, and the analysis step never works from a real identifier. Documents and images a patient shares, lab results, referral letters, ID photos, are read and stripped of identifying detail locally before anything is sent for further analysis, and voice messages are transcribed the same way, on infrastructure we control. This is the concrete mechanism behind the anonymization-before-sharing and data-minimization commitments described elsewhere in our data protection principles, not a separate policy layered on top of them.
The engine infers a patient’s intent and where they are in the recruitment journey, but a human can see and override every inference at any time, and both the system’s read and the coordinator’s correction stay on record. AI handles the volume and the speed; people stay responsible for judgment calls.
