Generic AI assistants answer from one system. AQ reads one connected record across every module, so its answers, flags and drafts cover the whole trial, and a named person decides.
Revised procedures, protocol amendments and storage excursions affect related activities. AQ identifies those connections and assembles the information needed to plan a coordinated response.
Each draws on records held in several modules. A tool scoped to one of them produces one part of the answer.
Outstanding requirements and emerging issues appear alongside source records, assigned reviewers and current status, providing the context needed to assess priorities and coordinate follow-up.
BREEZE-2 · 5 sites · Items needing attention4 openIllustrativeIndividually each was closed locally. Together they suggest the procedure rather than the site is the cause.
The milestone cannot be marked complete, and the study is scheduled to move to Conduct this month.
Eleven of the seventeen hold delegated tasks that reference it.
The consignments are held pending assessment.
Illustrative, fictional data.
Why the first item matters · deviations by site, last 8 weeksIllustrativeNo single site’s record contains the pattern. It appears only when the deviations, the corrective actions and the procedure are read together.
AQ Platform uses AI to maintain inspection readiness all the time, instead of in the fortnight before an inspection. The site file, delegation log, training record, pharmacy ledger and quality system sit on one record, so Connected Intelligence can answer the question an inspector actually asks: was every task done by a trained, delegated person on the current protocol?
Document intelligence · BREEZE-2IllustrativeAQ reads a document on arrival and proposes its classification and the expected-document entry it satisfies, with the version and signature requirements for that entry. The site file owner confirms or corrects the placement.
Readiness monitoring · BREEZE-2IllustrativeAQ compares the record against the requirements for a defined event — site activation, a monitoring visit or an inspection — and lists what is outstanding for that event.
Workflow assistance · BREEZE-2IllustrativeAQ drafts a pre-visit summary of outstanding items, and replies to monitoring queries built from the record’s own history. The responsible person edits and sends.
Quality signal detection · BREEZE-2IllustrativeAQ flags patterns across protocol deviations, corrective actions, training records and approvals for a quality lead to assess.
Amendment impact · BREEZE-2IllustrativeAQ produces an impact summary against an approved amendment: training requirements that change, visits affected, documents needing a new version, stock allocated under the previous protocol.
Readiness forecasting · BREEZE-2IllustrativeAQ projects each site’s position against a scheduled event, from outstanding item counts, their due dates and historical completion rates at that site.
Inspection evidence · BREEZE-2IllustrativeAQ retrieves and assembles the records behind a defined request from the site file, delegation log, training records and quality system, for review before anything is shared.
Portfolio questions · BREEZE-2IllustrativeAQ Assist answers across studies and sites inside the asking user’s permissions, for example which sites hold outstanding training against a given procedure.
Drafted corrective actions · BREEZE-2IllustrativeWhere a pattern has been assessed as requiring action, AQ drafts the corrective action from the deviation records and the procedure they reference. The quality lead supplies the investigation and approves.
Where a check runs on a configured rule rather than a model, AQ describes it as automated. Illustrative, fictional data.
Each module has its own AI features. All of them read the same connected study record, so a finding in one module reaches the others that depend on it.
Files each upload and keeps the site file inspection-ready.
Builds the TMF as the study runs and tracks TMF completeness.
Sees recruitment and milestones before they slip.
Watches storage and stock across every study it serves.
Follows each sample from collection to result.
Keeps procedures and training in step.
Turns deviations into tracked, effective actions.
Knows who may do each task, and whether they still can.
AQ eSource holds the site’s source forms on the same platform, each released with a recorded reason.
Evaluate the information each feature uses, the task its output supports, the authority required for approval and the evidence retained for subsequent examination.
Reads the document it is given
Nine connected modules on one study record
Investigator delegation governs activities at site; sponsor and organisational activities follow the organisation’s own roles. A proposal is routed on whichever applies.
In clinical trials, AI-powered quality risk detection means reading deviations, corrective actions, training records and approvals together to find the patterns that come before an inspection finding. It is the operational side of risk-based quality management (RBQM) under ICH E6(R3). AQ Platform does this across the connected study record, because those signals rarely sit in one system.
Quality signal · CAPA-0231 · closed 30 SepIllustrativeQuality signal · Dispense IMP · J. Okafor · Riverside TrustIllustrativeQuality signal · Last participant in · BREEZE-2IllustrativeQuality signal · Consent version not current at visit · 3 of 5 sitesIllustrativeAI identifies documentation gaps by comparing what a study record holds against what its protocol, milestones and delegations require: the expected-document list for the trial master file (eTMF) and the investigator site file (eISF), current versions and signatures, and the training and delegation behind each task. AQ reports missing documents and gaps in TMF completeness as they arise, while there is time to fix them.
What the file should hold at this point in the study.
Whether the filed version is current and signed.
Whether required training is assigned and complete.
Whether the person is delegated and the delegation is current.
One gap, read against two records · BREEZE-2 · Riverside TrustIllustrativeThe expired CV is a missing item in eISF 02.03, sitting in a queue of documents to chase.
Consent is being taken this week under a delegation whose supporting requirement has lapsed.
AQ Assist answers from the whole study record across every module, cites the records behind each answer, and walks your team through AQ itself.
Ask about BREEZE-2Answers come from every module and cite the records used.Reading 3 records…Eight items are outstanding against the visit, two of them overdue. The three below gate the visit; the remaining five are listed in the readiness view.
Reading 3 records…BREEZE-2 has recruited 112 of 160 participants across four recruiting sites. Screen failure is 19% overall; Northgate is highest at 31%.
Reading 3 records…Open CAPA and choose New deviation. AQ fills in the study and site from where you are working and suggests a category from your description. Three steps:
Reading 3 records…Protocol v4.0 sets a window of ±3 days for visits 2 to 6 and ±7 days for follow-up visits. Amendment 3 narrowed visit 4 from ±7 to ±3 days.
Reading 3 records…Four staff hold a live delegation to dispense investigational product across the five sites. Training is current for three. One has SOP-014 version 3 outstanding.
Illustrative, fictional data. Interface shown as in AQ Platform.
AI suggestions follow the applicable review and approval process, with supporting sources, reviewer details and decisions retained, including rejected suggestions.
Audit trail · entry 48213 · proposal not acceptedIllustrativeThe AQ Platform and the study record it holds are hosted in Microsoft Azure UK regions. AQ meets the NHS Data Security and Protection Toolkit standard (Standards Met, 2026).
AQ reads protocol deviations, corrective actions, training records and approvals together, and flags patterns for a quality lead to assess, supplying the records behind each flag. This supports risk management under ICH E6(R3).
Document classification, learning assignment on a version change, expected-document checks, and the assembly of records behind a request. Each produces an output a person confirms. Configured rules such as milestone gates continue to operate as rules.
The proposal and the decision are both retained, including where the proposal was not accepted: the model version, its confidence and rationale, the reviewer and their recorded reason.
No. Role permissions, delegation and blinding apply to AQ Assist as they apply to the person asking.
Each feature is scoped to a defined use and measured against known-correct examples before release.
Yes, where the system is validated for its intended use, a named person reviews each output before it becomes a record, and the audit trail shows what the model proposed and who confirmed it. Our guide sets out what the MHRA has published and what to check before relying on an AI feature. Read the AI in GxP guide
Explore document review, amendment assessment, quality oversight and evidence preparation in a walkthrough tailored to the organisation’s study requirements.
Pick a module, your organisation type, or both — we'll match the guides and email them to you.
Try one filter at a time, or clear to see all guides.
✓ Written for first-in-human & Phase 1 sites
✓ Inspection-ready checklists & templates
✓ Aligned to MHRA, FDA & EU Annex 11