Case study · Loan operations, CRM integration, and agent operations

Caul Group / Sympli Mortgage — Operations Platform & Integration Architecture

Connected loan operations, execution visibility, and engagement workflows without duplicating the system that governs the loan.

Integration Architect · Technical Product Owner · Operations Automation Lead

  • LiveLive loan integration
  • Structured demoAgent Hub pilot-gated
  • Sanitized reconstructionSanitized public case

01 · At a glance

What this system enabled

Business need
Connected loan operations, execution visibility, and engagement workflows without duplicating the system that governs the loan.
Nelcys’s role
Integration Architect · Technical Product Owner · Operations Automation Lead
What was delivered
Loan operations, CRM integration, and agent operations
Core systems
ARIVE · n8n · Sympli Ops Hub · GoHighLevel

Maturity and public status

  • LiveLive loan integration
  • Structured demoAgent Hub pilot-gated
  • Sanitized reconstructionSanitized public case

≈55%

reduction in automation execution cost

An engagement result from consolidating redundant executions while preserving necessary operational controls.

Inspect the supporting decisions ↓

02 · Business need

What the business needed

ARIVE governed the loan and borrower record, while operations and engagement teams needed clearer execution visibility without creating a second loan system.

The operating model had to connect approved information through the ARIVE Direct API, orchestrated through n8n, while keeping source ownership and borrower-safety rules explicit.

The live loan integration and the related pilot-gated Agent Hub are separate maturity claims.

03 · What needed to change

The operating risks to resolve

Connected loan operations, execution visibility, and engagement workflows without duplicating the system that governs the loan.

  • Loan truth needed to remain governed in ARIVE.

  • Duplicate records or uncertain matches could create borrower-facing risk.

  • Operational teams needed visible milestones, exceptions, and next actions.

  • Every movement needed traceable evidence before downstream action.

04 · What I delivered

My role in the system

I owned the business outcome, operating logic, system boundaries, acceptance criteria, delivery direction, validation, and final acceptance. AI-assisted development and specialist collaboration supported implementation where the context required them.

  • Defined the operating model, source-of-truth boundaries, and approved data movement.
  • Set stable-key, idempotent synchronization, reconciliation, and fail-safe escalation requirements.
  • Directed the n8n orchestration model for approved ARIVE Direct API flows.
  • Specified visibility, exception, and evidence requirements for Sympli Ops Hub.
  • Kept the live loan integration separate from the pilot-gated Agent Hub.

05 · How it works

How the system works

ARIVE connects through its Direct API, orchestrated through n8n. Sympli Ops Hub makes approved operating context visible, and GoHighLevel receives only approved engagement context. Stable identifiers, idempotent synchronization, reconciliation, and fail-safe escalation protect every boundary.

Selected system

ARIVE

Owns the governed loan and borrower record.

Boundary: Loan truth remains in ARIVE.

Systems in context

ARIVE
Authoritative loan and borrower state
n8n
Approved Direct API orchestration and exception handling
Sympli Ops Hub
Execution visibility, milestones, tasks, and exceptions
GoHighLevel
Approved CRM and engagement context
Supabase
Governed operational data environment
Vercel
Application deployment environment
GitHub
Source control, review, documentation, and handoff

06 · How I think

The decisions behind the system

Tools change. The durable work is deciding what owns truth, what may move, what must stop, and what evidence is enough.

Keep ARIVE authoritativeARIVE owns canonical loan and borrower state.
Why
A second operational truth would introduce conflicting statuses, reconciliation failures, and unsafe borrower-facing behavior.
Consequence
The custom platform visualizes, orchestrates, and enriches work; it does not become another loan-origination system.
Orchestrate approved Direct API movementUse n8n to coordinate approved ARIVE Direct API flows, validation, retries, and exception paths.
Why
The integration layer should move information without competing with the governed source.
Use stable keys and idempotent updatesRecords are matched through stable identifiers and updated through idempotent logic.
Why
Names, email addresses, and other mutable fields are not sufficient as the only reconciliation key.
Guardrail
Running the same sync again must update the same logical record rather than create another one.
Keep the operating layer distinctSympli Ops Hub owns visibility, handoffs, and exceptions while loan truth stays in ARIVE.
Why
Clear boundaries prevent duplicate truth and make ownership easier to inspect.
Fail safely around borrowersUncertain matches or validation failures escalate internally instead of contacting a borrower.
Why
Operational speed is not worth an incorrect or duplicated customer interaction.

07 · Agent Hub product chapter

Preserve what works. Gate what is not decided.

  • Structured demoDecision-ready / pilot-gated

The current structured demo supports meeting and training check-in, feedback, coaching requests, leadership visibility, meeting controls, and CSV export. A production pilot still requires approved identity, persistence, roster, messaging, and policy decisions.

Selected product environment

TanStack Start · React · TypeScript · Vite
Pilot-gated Agent Hub environment for meeting and training check-in, per-agent duplicate checks by meeting, feedback, coaching requests, leadership visibility, meeting controls, and CSV export.
Preserve verified demo behavior

Working routes, brand assets, check-in flows, meeting controls, coaching requests, leadership visibility, and CSV export remain visible in the structured demo.

Prevent duplicate check-in by person and meeting

The demo checks the agent and meeting identifier together so one person does not create duplicate check-in records for the same meeting.

Keep demo capability distinct from pilot readiness

Training check-in, feedback, and coaching requests are verified demo functions. Production authentication, backend persistence, roster synchronization, and messaging are not presented as delivered.

Keep provisional policy visibly provisional

Attendance, training, role, SMS, retention, and privacy rules remain labeled as assumptions until the responsible stakeholders approve them. Demo behavior is not represented as production readiness.

Current demo and pilot boundaries

Verified demo behavior and remaining pilot gates
ActorVerified demo behaviorCurrent statePilot boundary
AgentMeeting and training check-in, feedback, coaching requestLocal demo stateReal identity and persistent records remain gated
LeaderDashboard, meeting controls, and CSV exportClient-side demo controlsProduction permissions remain gated
DataPer-agent duplicate check by meetingLocalStorage-backed demoBackend, roster, and retention decisions remain gated
PolicyAssumptions remain visibly provisionalDecision support onlyAttendance, messaging, retention, and privacy require approval

08 · How it was delivered

From model to working system

  • Connected ARIVE’s Direct API through n8n under explicit source-of-truth rules.
  • Defined stable identifiers, controlled synchronization, reconciliation, retries, and exception paths.
  • Built execution visibility around milestones, tasks, exceptions, and next actions.
  • Gated engagement enrichment on approved operating context.
  • Documented the application environment through Supabase, Vercel, and GitHub without publishing private implementation details.

Guardrails

Where the system must stop

  • ARIVE remains the canonical loan system.
  • n8n moves approved information without becoming a source of truth.
  • Uncertain identity or state escalates internally.
  • Repeat syncs update a logical record instead of duplicating it.
  • Live integration and pilot-gated Agent Hub maturity remain separate.

09 · What changed

What changed

  • Canonical truth remained in the governed loan system.
  • Borrower-facing uncertainty failed safely and escalated internally.
  • Duplicate-risk controls were built into synchronization.
  • Reduced automation execution cost by approximately 55% while preserving necessary controls.

10 · Public evidence

What the public record can show

Evidence is described conservatively. Any visual proof added to this page must be redacted, captioned, and checked before publication.

  • Diagram

    How the systems work together

    Shows ownership, approved movement, controls, and evidence across ARIVE, n8n, Sympli Ops Hub, and GoHighLevel.

  • Narrative record

    Synchronization control narrative

    Explains stable identifiers, idempotent synchronization, reconciliation, and fail-safe escalation without exposing client records.

  • Narrative record

    Maturity boundary

    Separates the live loan integration from the pilot-gated Agent Hub and the sanitized public case.

11 · What this proves

This case demonstrates integration architecture, technical product ownership, source-of-truth design, automation governance, and public-safe evidence for a live operational system.

Discuss a similar system