Company
Enrolment is the bottleneck in bringing a drug to people.
Around 80% of trials miss their enrolment timelines. Meanwhile eligible patients sit in records nobody searched, because criteria are dense prose and charts are unstructured narrative. Accrual exists to close that gap.
Why this problem, and why now.
Trial matching has been attempted for two decades and mostly failed for the same reason: the systems could search structured fields, and eligibility criteria are not written against structured fields. They are written against what a clinician wrote in a note.
Reading that narrative reliably has only recently become possible, and language models are the reason. What that unlocks is not one product but two: the same AI read that finds a patient for a site also answers, for the first time with evidence, the question sponsors have been guessing at forever — will this site actually enrol?
Diversity requirements arrived at the same moment, pushing sponsors towards sites that have never had research infrastructure. Those two facts together are why this is worth building now rather than five years ago.
What we hold to
Five things we are not willing to trade.
The patient is the point
Every number on this site is ultimately a person who could have been in a trial and was not. We keep the argument commercial because that is what moves sponsors, but the reason the work is worth doing is that eligible patients are currently going unfound.
Show the evidence, always
A match without the chart line that produced it is a guess wearing a confidence score. Every determination the model makes is traceable to the text it came from, because a coordinator has to be able to check it in seconds. An AI that cannot be checked has no business near a patient record.
Flag ambiguity, do not resolve it
When a record cannot settle a criterion, the honest output is 'needs review', not a probability. Overclaiming here produces screen failures, which is the exact waste we exist to remove.
The boundary is the business
Sponsors never see patient data. If that boundary ever softens, sites stop participating and there is no product. Treating it as a constraint rather than a feature is the mistake we most want to avoid.
Go where the infrastructure is not
The large teaching trusts are already served. The structural opportunity — and the harder engineering problem — is the district generals and community trusts with real patients and no research staff.
Careers
We are small, and hiring ahead of the roles.
There is no open-roles board yet, and posting invented vacancies would waste your time. If you have worked in clinical research operations, on clinical NLP, or on health-system integration, write to us with what you have built and we will make the conversation happen.
We are especially interested in people who have sat on the site side — research nurses, study coordinators, CRAs — because the product fails in ways only they can see coming.
Where we are
Accrual is pre-launch. We have no customers to name, no funding to announce, and no benchmark numbers of our own — so you will not find any invented on this site.
What we do have is a clear read on the problem and a product built around a boundary that makes sites willing to participate. If that is worth a conversation, request access or write to us directly.
Get in touch
Sites, sponsors, and people who want to build this.
All three start in the same place.