See the discovery and investigation run.
Reveal.IQ discovers unknown patterns and performs the investigation experts largely conduct by hand today. This is what it produces, what it runs on, and what you can watch it do on a real study.
See it work
What lands on the reviewer's desk.
For example, safety review looks at one domain at a time. Liver chemistry against its own thresholds, adverse events against their own coding, dosing against protocol. A pattern that lives across all three, weak in every one, is not visible in any of them.
Reveal.IQ does not hand a reviewer a chart and a hunch. It hands them a finding, the evidence behind it, and the questions that are still open.
Each family is weak on its own. Consistent together, and building toward the outcome, they form the precursor pattern the engine detects.
- Does the pattern hold across the whole trajectory, not a single reading?Supported
- Is it explained by a single-site measurement artifact?Weakened
- Could a concomitant medication account for it?Unresolved
- Gate - enough governed evidence to reach a conclusion?Passed
What comes back
- The pattern, stated plainly, and the participants carrying it.
- Every timeline assembled behind it, each fact traceable to the record it came from.
- Whether the pattern is distinctive to this group or common in the study population.
- The competing explanations, including the ones the evidence does not settle, left in front of the reviewer rather than closed by the system.
- What the analysis could not establish, stated as plainly as what it could.
Run it again on the same data and you get the same result. Not a similar one.
In practice
Two investigations. One framework.
Both start with a pattern nobody was anticipating. Both end with a reviewer who arrives at the decision point already equipped. Neither conclusion was reached by a model improvising.
Clinical discovery: a hepatic safety signal
Discovery surfaces an unexpected cross-domain pattern: treatment timing, a liver lab trend, and site behavior, together. No rule was looking for that combination.
The hepatic playbook evaluates the treatment effect, a concomitant medication interaction, site conduct, and an operational artifact. The signal persists after site adjustment. The temporal sequence is clinically plausible.
Clinical escalation recommended, with the full evidence package. The reviewer arrives at the decision point already equipped.
Operational discovery: a dropout anomaly
Discovery surfaces statistically elevated dropout clustered at Visit 8 across three sites. No single site crossed a KRI threshold. Invisible to conventional monitoring.
The operational playbook confirms the Visit 8 concentration, rules out a protocol amendment, and identifies a scheduling workflow change at the affected sites. No safety signal association.
Operational follow-up recommended, classified as a likely artifact, with visit distributions and peer comparisons attached.
How it works
Rules detect. Discovery uncovers. Playbooks investigate. AI explains. You decide.
Detect
Data quality rules, KRIs, QTLs and medical alerts, applied consistently across every study on one harmonized data layer. Detection does not have to be ours; the rules already running in your systems feed the same investigation, so nothing has to be replaced to get started.
Discover
Patterns no rule was written to catch, surfaced across domains rather than one at a time. What reaches a person has already been tested for whether it deserves to be there.
Investigate
Every detected and discovered signal is investigated, not only the ones someone had time for. What comes back is a conclusion with the evidence behind it.
Explain
A study insights report and a chat interface to interrogate the finding, both strictly bounded to the evidence in the package.
Why this is different
Not another dashboard. Not a chat interface. A governed investigation system.
Doing nothing
Patterns accumulate undetected. You pay for them later, in amendments, inspection findings and surprises.
Caught early
Unknown patterns surface before they become protocol changes or inspection issues.
Validation rules and KRI platforms
Detect issues and surface alerts, then stop. Interpretation is left entirely to the reviewer.
Investigates, and concludes
Governed playbooks produce a conclusion with an evidence package, not a flag.
Generative and chat-first AI
Flexible conversation with no audit trail, no reproducibility, and a standing risk of confident invention.
AI explains, never invents
AI explains governed results, never generates them. Every finding traces to the evidence behind it.
Manual expert review
Inconsistent, time-consuming, and impossible to scale across hundreds of sites and thousands of signals.
Expert reasoning, encoded
Playbooks encode expert investigative reasoning and apply it consistently across studies, sites and signals.
Reveal.IQ does not automate clinical judgment. It automates the disciplined investigation that precedes it.
No black box
Three pillars.
Governed
Playbooks are inherited from the library rather than written from scratch. Your experts adapt them, or author new ones with an AI assistant, and every playbook is approved, versioned and permissioned before it runs. The method becomes a reusable asset that stops walking out the door when a senior physician retires.
Traceable
Every recommendation is source-linked back to the underlying data. Inspection-ready by construction: what was found, what evidence was used, what stayed uncertain, and what action followed.
Human-decided
Reveal.IQ assembles and tests. AI explains. The accountable expert confirms and acts. The system never adjudicates.
What a playbook is, and the library it lives in
The investigation method a senior reviewer already carries in their head, written down once.
A playbook contains your experts' judgment; what it guarantees is that the same judgment is applied every time, by everyone, with the reasoning on the record afterwards. Reveal.IQ assembles the evidence, builds the patient timeline, and tests each explanation, showing what was found, what supports it, what was ruled out, and what is still uncertain.
The library ships with the platform and is extended by your own experts. It reaches across renal, hematologic and cardiac safety, adverse event and concomitant medication adjudication, operational and data quality review, and playbooks that assemble a participant's story for the reviewer. Each one is approved by the experts who own that review today, and the library compounds with every study.
One method behind every finding, whether the trigger came from a business rule or from discovery.
The data foundation
Governed data in. Defensible decisions out.
An SDTM-based canonical layer
Playbooks run on a standards-based canonical layer, so investigation logic works identically across studies and sponsors. Bring your own SDTM, or we map it.
Raw-to-target lineage
AI-assisted mapping preserves source history. Every finding traces back to the file, the site and the entry date it came from. An event dated January 1 but not entered until January 19 is visible as exactly that.
Your study, not a reference study
The protocol and the annotated CRF are ingested, so findings are evaluated in the context of your study rather than against generic reference data.
Built to be adopted
Reveal.IQ is designed to fit real study environments without requiring a transformation program first.
When an inspector asks how you decided this, the answer is in the system.