Relationship Review: Beyond the Direct Match.

Everyone’s talking about Jev. I decided to build with it on the iPhone Duo. The naming opportunity was difficult to ignore.

The serious experiment is whether a focused model can make evidence review clearer for legal, trust and corporate-services teams.

A client returning no direct screening match doesn’t resolve questions about the surrounding relationships, payment arrangements or ultimate principal. This prototype helps a reviewer organise that picture.

Users create a scenario, identify who they represent, add related entities and connect relationships to supporting evidence. Documented and unverified links are distinguished, and reviewers can include or exclude individual sources.

One live TypeSafe API call asks Jev three deliberately narrow questions:

Is a substitute beneficiary or settlement intermediary involved?

Is there an unresolved ownership or ultimate-principal gap?

Is there an unexplained mismatch between the invoice and payment beneficiary?

Jev returns structured answers and confidence estimates. Ordinary application code combines those assessments with the supplied relationship paths to suggest specialist review, a request for more information or standard human review.

Built in SwiftUI and demonstrated in the iPhone Duo simulator, the app adapts to the unfolded display with a relationship map, evidence readers and a separate assessment view.

The graph comes entirely from supplied relationships. The prototype uses fictional cases, performs no live sanctions screening and makes no legal determination.

The next test is whether practitioners can reach a better-supported review more quickly, not simply whether the demonstration looks convincing.

Create and copy scenarios, including the client represented, entities, evidence and relationships.

Explore a directed relationship map with source-linked connections.

Distinguish documented relationships from unverified allegations.

Include or exclude evidence and reassess the remaining information.

Run three focused Jev assessments in a single live API request.

Preserve “insufficient evidence” as an explicit outcome.

Inspect the review route, model confidence and application rule.

Export a review record containing the scenario, selected evidence, request and response.

Use an adaptive native interface for folded and unfolded screens.

I’d welcome feedback from sanctions, compliance and legal practitioners on the questions, evidence presentation and escalation rules.

I’m particularly interested in testing ambiguous fictional scenarios and ordinary transactions that should not trigger unnecessary escalation. Does the workflow expose useful gaps, miss important context or create more work than it saves?

Feedback on the practical value of reviewing this material on a phone would also be useful.