Product discovery
Defined the target problem, user workflow, evidence rules and smallest coherent demonstration.
PORTFOLIO CASE STUDY
Verisy: designing an evidence-controlled AI workflow.
A portfolio product-discovery and prototyping project created to explore how organizations can draft complex responses without losing the evidence or human accountability behind them.
ROLEProduct Owner and AI-Assisted Builder
DOMAINHealthcare technology and security assurance
DELIVERABLEInteractive zero-cost portfolio prototype
STATUSFunctional guided demonstration
THE PROBLEM
Questionnaires turn expert knowledge into a repetitive bottleneck.
Customer security reviews repeatedly ask organizations to prove the same controls in different formats. Policies, previous answers and supporting records may disagree or become outdated. Under deadline pressure, teams risk copying language that is incomplete, stale or broader than their evidence permits.
Verisy explores a controlled alternative: make the source visible, preserve qualifications and require a person to decide what the organization may claim.
MY CONTRIBUTION
From an ambitious idea to a testable, honest product demonstration.
Business analysis
Translated questionnaire work into requirements, outcomes, review states and acceptance criteria.
Workflow design
Separated model confidence, evidence sufficiency and human approval so one signal cannot disguise another.
Competitive research
Reviewed established questionnaire, RFP and GRC products and revised the product thesis when overlap became clear.
Iterative delivery
Directed repeated interface improvements across evidence views, decision controls, industry terminology and export behavior.
Risk control
Constrained the portfolio version to fictional, precomputed results with no customer files, patient data or paid AI connection.
WORKFLOW ARCHITECTURE
The evidence remains attached from intake through export.
KEY PRODUCT DECISION
Confidence is not evidence.
A system may be highly confident that it found the most relevant paragraph while that paragraph still fails to support the requested claim. Verisy therefore treats these as separate signals.
94% confidenceHow strongly the source matches the question
QualifiedWhether the available evidence supports the complete claim
ApprovedThe human decision governing release
WHAT RESEARCH CHANGED
The original feature was valuable—but it was not unique.
What I discovered
Established platforms already offer AI drafting, citations, confidence indicators, content libraries and human review.
How I responded
I stopped treating evidence-backed drafting as the differentiator and reframed the opportunity around controlling unsupported business claims.
Why that matters
Product judgment includes challenging the original premise, narrowing scope and changing direction when evidence contradicts an assumption.
CURRENT LIMITATIONS
A functional workflow prototype—not a production AI service.
Precomputed fictional analysisKeeps the demonstration reliable, truthful and free to operate.
No customer document processingThe public portfolio workflow does not receive sensitive business or patient data.
No compliance claimsThe prototype does not certify an organization or replace expert review.
Production path identifiedA future version would require secure ingestion, retrieval evaluation, access controls, monitoring and carefully governed model use.
EXPLORE THE PROJECT
See the product decisions in action.
Run the guided SampleCo scenario, inspect the evidence and make the final review decisions yourself.
Open the working prototype ↗