For research sites
Every day, patients who meet the criteria of an open protocol pass through your clinics unnoticed — because finding them means reading thousands of charts against pages of prose, and nobody has those hours. Accrual's AI has those hours.
A coordinator working through a protocol manually can review a few charts an hour, and only the charts they thought to open. Accrual's models read the whole population against every open protocol and hand back a ranked list with the evidence cited. You review a shortlist instead of a filing cabinet.
Matches arrive inside the workflow the treating clinician already uses, while the patient is still in front of them. That is the only moment the conversation reliably happens.
Because the model resolves every criterion against a specific line in the chart before the patient is approached — and flags what it cannot settle rather than guessing — fewer patients reach consent and then fail screening, which is where the wasted effort actually sits.
Sites that can evidence their population get put on lists they were never on before. That matters most if you are not a major academic centre.
What it costs the site
Nothing
Sites are not the customer. Sponsors and CROs pay — per qualified enrolled patient, or per feasibility study — and that is what funds the platform.
We would rather be direct about the arrangement than pretend it does not exist: your population is what makes the sponsor-side product possible. In exchange you get the matching engine, at no cost, with the data controls above written into the agreement rather than promised on a website.
Data governance
These are the questions that come up in every information governance review, from the R&D office through to the Caldicott Guardian. None of the answers are hedged, because a hedge here means the review stalls.
You do. The site is the controller under the UK GDPR. Accrual is a processor acting on your documented instructions under an executed Article 28 data processing agreement, and for no other purpose. Where you operate a Caldicott Guardian function, the purpose is agreed through it before any record is read.
Processing runs within the governance boundary agreed in your data agreement. Nothing is copied out of that boundary for any purpose you have not authorised in writing.
Aggregate, anonymised population evidence — counts, distributions and forecasts. Never records, never identifiers, never individual-level data. Small counts are suppressed rather than reported.
No. Records are processed to answer your questions and are never used to train or fine-tune models, never pooled across sites, and never shared with a model provider for improvement. Zero-retention terms are required of any provider in the path, and this is in the agreement rather than only the policy.
Yes, and it is enforced in processing rather than left to policy. Mental health notes, sexual health records and gender identity information carry heightened confidentiality expectations and are commonly restricted at the site level.
No. The connection is read-only. The clinical record remains yours and unaltered, and the AI takes no action on a patient — it produces a shortlist for your staff.
Their own care team, always. Accrual does not contact patients and provides no mechanism for a sponsor to.
The full posture is on the security page, and the patient-data boundary is set out in the patient data notice.
District generals and community trusts
Trial volume has always concentrated in a handful of academic centres — the large teaching trusts with dedicated research staff to do the finding. Diversity requirements are now forcing sponsors to look elsewhere, and elsewhere is you.
No staff
The AI does the notes review, so continuous screening needs no dedicated research nurse
No project
Read-only connection, no data warehouse build, no new system of record
Be found
Population evidence puts your site on sponsor shortlists it would not otherwise reach
For sites
Tell us which system you run and roughly how many patients you see. We will tell you honestly whether we can help.