AI drafts the study.A qualified human signs it.
Forms, edit checks and queries drafted from your protocol in minutes, not weeks. Nothing reaches a site until a data manager has reviewed it and put their name to it — on one continuous data model, with one audit trail underneath.
- SUBJ-00482Adverse Events · SeverityOpen
- SUBJ-00139Vital Signs · Blood pressureAnswered
- SUBJ-00251Con Meds · Start dateOpen
- SUBJ-00044Demographics · DOBClosed
Three steps. One of them is a person.
The draft is automatic. The checking is automatic. The signature is not, and that is the point — nothing reaches a site, a monitor or a regulator without a qualified human putting their name to it.
Build an eCRF without writing code.
Study build runs four to eight weeks of hand-coding on most platforms. Here you drag field types, conditional logic and edit checks into place — no code, no vendor ticket.
- · Conditional logic without programming
- · Edit checks at the point of entry
- · Amendments deploy without a rebuild
Catch it. Rank it. Draft the query.
Edit checks fire as data lands, not in a monitoring pass weeks later. The discrepancy arrives with subject, site and visit attached, and the site query is drafted and logged to the audit trail.
It then waits. Nothing leaves the platform until a data manager reads it and decides it should.
- 1Detect
- 2Prioritise
- 3Draft
Consent date precedes first study procedure
Verify that consent is signed before any visit or drug exposure.
To: Site 04 — Study Coordinator
Re: SUBJ-00214 · Informed consent date
The consent date on the Informed Consent form is blank, but Visit 1 procedures are recorded for 12 Mar. Please confirm the date consent was signed and attach the signed ICF page.
The part that stays human.
Every draft the platform produces is a proposal until someone qualified accepts it. The reviewer, the timestamp, the value before and the value after are written to the same audit trail as the data itself.
- · Attribution and reason-for-change on every edit
- · Electronic signatures under 21 CFR Part 11
- · One immutable trail, queryable at inspection
Three costs recur on every study.
Every one of them looks like an operations problem. Every one of them is your data model.
Vendor fragmentation across BA-BE pipelines.
Volunteer recruitment in one system, randomization in another, dispensing in a third, eCRFs in a fourth. You reconcile the lot in spreadsheets.
Paper workflows at remote sites.
Connectivity drops away from major centres and your coordinators fall back to printed CRFs. Every page is transcribed again by hand.
Contracts priced in a currency the trial is not.
You quote in one currency and pay in another. The gap comes out of your margin, and it is widest on high-volume, thin-margin work.
One platform. One schema. One audit trail.
Every module reads and writes the same records. There is no reconciliation between modules, because there is no boundary between them.
one schema · one subject record · one immutable audit trail
EDC is live today. The remaining modules are in build on the same model.
External integration will always be a requirement. What changes is that it happens once, at the boundary — not repeatedly between internal modules.
See every supported integrationSee every moduleCompliance sits in the schema, not on top of it.
Audit trails and electronic signatures are properties of the data model, so no module can be configured out of them.
“Most systems we looked at were built for a trial that does not look like ours. Trialion starts from Schedule Y, and from what a coordinator at a remote site can realistically do on the day. The audit trail is the part that holds up when an inspector starts asking.”
The questions we get asked before a build.
Answered as we would answer them on a call, including the ones where the answer is not yet.
What environments do I get, and can test data reach production?
Today a study moves draft → UAT → live, and configuration is promoted between them. A separate development environment and an independent QC stage, with test subject data isolated from production, is the next piece of architecture we are building. If you are planning a build, ask us where it stands on the day.
Show me a mid-study amendment with subjects already enrolled.
Amendments version the study configuration and deploy without a rebuild. What we will not claim yet is a complete change-control workflow across separate environments, with a documented impact on already-enrolled subjects. Bring the scenario to the call and we will walk it honestly rather than demo around it.
How is my study validated, separately from your platform?
They are two different things and we keep them separate. Platform validation is ours: the system is architected to GAMP 5 and the formal exercise runs with an external auditor. Study validation is proving that your CRFs, edit checks, derivations and visit structure behave to your specification — today that happens in UAT, and a dedicated QC stage with retained evidence is in build.
Can we configure roles beyond PM, DM, CRA, CRC and PI?
Yes. Roles and their permissions are configurable rather than fixed, at organisation, study and site level, because no two CROs use the same titles. Field- and form-level restriction is the next level of granularity and is not there yet.
What happens when a required field genuinely cannot be filled?
Today it raises a query, and that is wrong — a deliberate blank is not a discrepancy. Standardised missing-data reasons (not done, not applicable, unknown, not available) recorded with a comment on the audit trail are a change we are making, because statistics needs to tell an omission from a fact.
Do we have to use the AI?
No. A study can be built entirely by hand, and nothing the model drafts takes effect until a person accepts it. The AI writes a proposal; a qualified reviewer decides whether it becomes the study.
What is retained about an AI-drafted form?
The generated output, every human change made to it afterwards, and who approved it are on the audit trail. Retaining the model version and the prompt alongside them is a change in progress, so an inspector can see what was proposed as well as what was accepted.
Do you support blinded studies?
Not today. Trialion is built for open-label work, including bioequivalence and Phase I–IV designs that randomise without blinding. Blinding touches randomisation, drug supply, unblinded pharmacy roles and emergency unblinding at once, and we would rather say no now than ship something that leaks a treatment assignment.