Accounting Firm · n8n Workflow Automation
Accounting — Client Document Collection Tracker
Portfolio DemonstrationEach fictional client is checked against a service-specific required-document matrix; missing documents and days-until-deadline combine into a High/Medium/Low risk score that drives reminder cadence and, only for High risk, accountant escalation - all within a real Split-In-Batches loop that aggregates into a dashboard once complete.

Problem
Chasing clients for missing tax and financial documents ahead of deadlines is time-consuming and easy to lose track of.
System Design
Checklist and deadline loop: a Schedule Trigger feeds a Split-In-Batches loop where each fictional client runs through a document-checklist and deadline-risk scorecard before looping back.
Workflow Stages
- 01Client Loop
- 02Document Checklist & Risk Scoring
- 03Reminders, Escalation & Completion
- 04Filing Readiness Dashboard
Tools Used
- n8n
- Code (JavaScript)
- If
- Loop Over Items
- Schedule Trigger
- Set
- Switch
Safety & Human Review
- Workflow name begins with DEMO | and is inactive.
- No credentials are attached to any node.
- No enabled external send node exists anywhere in this workflow.
- All data is fictional (example.com emails, invented names/addresses).
- Uses only built-in n8n node types.
Implementation Evidence
Real screenshots of the working system — collapsed by default so this page stays readable.
What's Working
7 structural checks passed, including: Workflow JSON parses (local + server-returned copy); active is false; Name begins with DEMO |; No credentials attached to any node. Structural and logic validation completed. Live n8n execution pending manual safe test.
Limitations & Confidentiality
Classification and scoring logic is deterministic/rule-based rather than a live AI model, by design (provider-independent, fully inspectable, no credential required). No real system of record, messaging channel, or calendar is connected - every downstream action is simulated or drafted, never sent. Live n8n execution has not been triggered as part of this build.
Portfolio demonstration built for a fictional business with fictional data. It is not connected to any real client, system, or customer.
What I'd Improve Next
Replace deterministic classification with a real AI model behind the same confidence/human-review pattern where one exists. Connect the mock/simulated steps to real systems of record once reviewed and approved. Wire a real (approved) communication channel behind the draft-message steps.
Have a similar workflow to automate?
I'll tell you honestly whether automation is the right fix.