Built in response to the trust gap exposed by AI-assisted approvals, it makes every compliance decision transparent, reviewable, and defensible across employees, finance operations, and compliance teams.
| Area | Details |
|---|---|
| Role | Lead Product Designer (IC), end-to-end |
| Duration | 12 weeks |
| Type | Concept project, tested prototype |
| Users | Compliance reviewers, Finance Ops, employees submitting expenses |
| Platforms | Web for compliance operations, mobile for employees |
| Validation | 4 moderated prototype sessions, 4 SME reviews, 3 major iterations |
Boundary1 improved compliance speed using AI-assisted decisioning, but introduced a critical gap:
Decisions were fast — but not explainable or defensible.
This created two breakdowns:
For reviewers
Could not trace why a decision was made
Could not justify overrides during audits
Lacked evidence-linked reasoning
For employees
No clarity on how data was used
No visibility into verification logic
Low trust in compliance outcomes
Turn compliance from a pass/fail workflow into a reviewable decision system.
I designed 4 core system capabilities:
1. Explainability Layer
Every decision links to:
rule triggered
supporting evidence
validation outcome
2. Privacy & Consent Layer
explicit consent per verification
no passive tracking
minimum-necessary data usage
3. Auditability Layer
full decision history
override tracking
structured audit logs
4. Recovery Layer
structured exception flows
no dead-end failures
guided correction paths
making AI outputs explainable in context
making rule evaluation visible to reviewers
making human overrides traceable for audit
preventing workflow dead-ends via recovery flows
| Boundary1 | Boundary2 |
|---|---|
| Fast approvals | Explainable decisions |
| Hidden AI reasoning | Rule + evidence visibility |
| Limited audit support | Full audit traceability |
| Passive tracking | Consent-based verification |
Explainability over Automation
Increased review complexity
Enables defensible compliance decisions
Consent over Continuous Tracking
Reduced passive data collection
Desktop Ops Workspace over Lightweight Flow
Added complexity in UI surface
Research showed that compliance work extended beyond approvals.
Reviewers needed a workspace for:
I moved compliance work from a lightweight approval flow into a desktop-first operational workspace.
The biggest shift in Boundary2 was making decisions inspectable instead of opaque.
Reviewers did not need more raw information. They needed the minimum defensible explanation:
rule -> evidence -> status -> next action
So I designed the review experience to show not only the outcome, but the logic behind it.
Boundary2 makes compliance-related data usage visible where trust matters most: verification, consent, and status.
Instead of relying on persistent tracking, the system uses explicit, point-in-time consent and minimum-necessary precision for location verification.
Boundary2 was designed to stay usable when compliance workflows break down.
Instead of treating failures as dead ends, I created structured recovery flows for:
The principle behind these flows was simple:
Boundary2 supports recovery across the issues most likely to interrupt expense review.
Flags missing VAT details at review time, links the failure to the relevant compliance rule, and routes the case into a correction flow instead of allowing ambiguity to continue downstream.
Turns a compliance blocker into a clear correction path
Surfaces payout failure clearly, explains the issue, and preserves next-step actions without breaking the workflow.
Keeps payout issues actionable instead of opaque
Routes low-confidence extraction into a guided correction flow instead of a silent failure, keeping the issue visible and recoverable.
Turns document failure into a clear next step
Flags possible duplicate submissions early and helps users resolve the conflict before the expense moves further into review.
Reduces duplicate reimbursement risk upstream
Requests access only when compliance verification is required, with explicit purpose and minimum-necessary precision.
Supports privacy-first, audit-aware verification
Lets employees continue expense capture in low-connectivity conditions, preserving submission progress until sync is restored.
Prevents submission loss at the point of capture
I tested Boundary2 through:
4 moderated sessions with employees, finance ops, and compliance leadership
4 SME reviews with compliance and privacy stakeholders.
What validation confirmed
trust increases when decisions are tied to rule + evidence
reviewers prefer minimal defensible explanation over full data exposure
structured recovery reduces workflow breakdowns
consent visibility improves employee confidence
What changed through iteration
Boundary2 transforms compliance from an opaque approval system into a transparent decision infrastructure that:
I learned that in compliance systems, transparency is not a feature — it is the product itself.
Trust is created when decisions are: