Prepared for Louise Chen and the Share with Oscar technical team
October 2026

How your automation works, step by step

This walkthrough shows each workflow in the three phases. For each step, you see the n8n node, the platform it talks to, the API call and the output.

Your systems
n8n foundation
Privacy check
Claude
Dashboards, answers, alerts

The architecture

Your systems stay as they are. n8n runs on a server that you own. Claude only gets data after the privacy check.

Your systems

Stay in place. We connect to them.

Booking platform databaseRead-only user, read replica
Booking platform APICustomers, bookings, listings
Payment providerStripe shown as example
Google WorkspaceEmail delivery

Spark automation foundation

n8n on your own cloud server

n8n workflowsTriggers, logic, API calls
Approved reportsPre-built queries only
Privacy checkBlocks personal data
MCP serverSecure link to the Claude app

AI layer

Gets totals and trends only

Claude APIAnalyse, sort, draft
Claude appAsk questions on desktop or phone

Business tools

Where your team sees the results

Live dashboardTool to confirm at discovery
HubSpotService tickets and sales CRM
ApolloLead search and sequences
Slack or emailAlerts and approvals

HubSpot and Apollo are used from Phase 2 and Phase 3. Phase 1 uses the database, n8n, Claude, the dashboard and email.

Workflow walkthrough

Select a phase and a workflow. Use Next step, or the arrow keys, to walk through it.

Spark foundation (n8n) Your systems Third-party platform AI (Claude)

Platforms in this workflow

Final output

How customer data stays on your servers

Two layers of protection stop personal data before it reaches Claude.

Your database

Holds all customer data. Nothing changes here.

Layer 1: reporting views

Your developer creates views without names, emails, phones, addresses or bank details.

Approved reports only

n8n runs fixed queries on the views. Claude cannot write its own queries.

Layer 2: privacy check

Each result passes an allowlist and a pattern check. On a fail, the run stops.

Claude

Gets totals and trends. Commercial terms: no training on your data.

Read only in Phase 1

Phase 1 reads data. It never changes your platform.

Human approval for actions

Any change on your platform waits for a yes from your team, and n8n logs it.

Short log retention

n8n deletes run logs after a short period. Workflows with personal data do not save run data.

What we need from your technical team

These items let us start each phase on time. Phase 1 items have the most effect on the 4-week target.

Phase 1

  • Read-only database user, and a read replica if possible
  • Reporting views, or access for us to propose them
  • List of tables and fields, with personal fields marked
  • Cloud account for the n8n server (Australian region)
  • Payment provider name and read-only API access

Phase 2

  • Booking platform API documentation, with write endpoints and rate limits
  • HubSpot private app token with ticket and contact access
  • Webhook access from the payment provider
  • List of actions to automate, and actions that need approval

Phase 3

  • Apollo API key and HubSpot Sales Hub access
  • Ideal customer profile and target segments
  • Decision on the sequence tool: Apollo or HubSpot
  • Sending domain set up with SPF, DKIM and DMARC
To confirm at discovery
  1. Booking platform API paths. The paths in this walkthrough are placeholders.
  2. Payment provider. Stripe is shown as an example.
  3. Dashboard tool, for example Looker Studio or Metabase.
  4. Alerts by email or Slack.