Practical AI · Built in Melbourne
Automation that connects your stack
Multi-step workflows across CRM, inbox, calendar, and database. We build the wiring — and the fallback path for when an upstream API doesn't behave.
Why most automations break in the second quarter
Connecting two SaaS tools with an off-the-shelf integration usually works. The fragility shows up at month three, when a CRM custom field changes shape, an API rate-limits at the wrong hour, or an edge case nobody planned for sends the workflow into a quiet retry loop. Three failure modes keep showing up when we audit existing automations.
Generic connectors that snap on schema drift.
When the Salesforce admin renames a field, the off-the-shelf workflow breaks silently. Nobody notices until a customer flags the missing record three weeks later.
No retry logic for upstream API failures.
Every API has a bad day. Workflows that don't budget for transient failures lose data — or worse, double-create records when somebody manually re-runs the trigger.
Approval steps stuck in nobody's queue.
Workflows that route to a person without a clear queue or deadline turn into invisible bottlenecks. The work happens — eventually, after someone chases it down in Slack.
How we build it
Reliable automation is mostly error handling. We start with the happy path — n8n, Make, or Zapier for the well-trodden integrations — and pair it with a custom layer for the boundary cases generic connectors don't cover. Workflows get split into discrete steps with idempotency keys, so a re-run never double-creates a record. Every external call has a retry policy with exponential backoff and a dead-letter queue for the cases that can't recover automatically. Approval steps land in a real queue with notifications and a deadline, not a nameless email inbox. Where a third-party connector doesn't exist or doesn't fit, we drop in a Python service or a FastAPI endpoint that talks the API directly. The orchestration runs on n8n self-hosted or Vercel infrastructure depending on data residency. Every workflow run logs the trigger, the steps, the outcomes, and the timing — so when something does go wrong, the post-mortem is a query, not a guess.
Tools we lean on: n8n · Make · Zapier · Python · FastAPI · custom APIs
Pipeline shape · Business Process Automation
- 01Trigger
- 02Fetch
- 03Map
- 04Validate
- 05Sync
What the automation layer does, end to end
Six capabilities every workflow we ship includes by default.
Multi-step orchestration.
Triggers, conditional branches, parallel paths, and joins — modelled as discrete nodes with idempotency keys, not a single linear chain.
System integrations.
CRMs (Salesforce, HubSpot), accounting (Xero, MYOB), helpdesk (Zendesk, Intercom), Microsoft 365, Google Workspace — and a custom adapter where the off-the-shelf one falls short.
Retry logic and dead-letter queues.
Every external call has exponential-backoff retry. Failures that exhaust the retry budget land in a dead-letter queue with full context for manual recovery.
Human approval steps.
Approvals route to a real queue — Slack, Teams, or a lightweight web UI — with deadline reminders. The workflow never stalls invisibly.
Audit trail per run.
Every workflow run logs trigger, steps, outcomes, and timing. Exportable for finance audits or compliance reviews. Retention configurable per workflow.
Conditional routing on data shape.
Workflows branch on field values, customer segment, deal size, or any predicate the data carries — not just on event type.
Packages start at three sizes
Most clients land on Scale. We re-quote against the systems involved and the workflow complexity after the audit.
Automate
From $2,000 AUD
Single workflow connecting two systems. Basic retry. Email notification on failure.
- One workflow
- Up to 2 systems
- Basic retry on transient failures
- Email failure notifications
Scale
From $5,000 AUD
Multi-workflow suite across 3-5 systems with conditional routing, human approval steps, and a monitoring dashboard.
- 3-5 workflows
- Up to 5 systems
- Approval queue with notifications
- Dead-letter queue for failures
- Run dashboard with audit trail
Transform
From $10,000 AUD
Workflow programme — multiple suites, custom adapters, advanced monitoring, ongoing reliability tuning.
- Unlimited workflows
- Custom API adapters
- Advanced monitoring + alerting
- Quarterly reliability reviews
- Ongoing workflow maintenance
Real-world scenario · 2025
5 hours of payroll work cut to 30 minutes
A Melbourne logistics company was running fortnightly payroll across four systems — a rostering tool, time-tracking software, a payroll platform, and an accounting package. Every cycle, two staff spent five hours each reconciling the data by hand. Misalignments were common; corrections were slow; superannuation calculations occasionally needed a third pass.
We built a five-stage automation: trigger on the close of the time-tracking period, fetch from the rostering and time systems, map fields to the payroll schema (with conditional branches for casual versus full-time and after-hours rates), validate against business rules, and sync to the accounting package. Approval steps for anomalies — outlier hours, unmatched rosters — routed to a Slack queue with a four-hour deadline. n8n handled the orchestration; a Python service covered the field-mapping logic the off-the-shelf connectors couldn't express.
The fortnightly reconciliation collapsed from ten person-hours to thirty minutes. The dollar saving — labour cost plus audit-correction overhead — landed at about $40,000 AUD a year. The two staff still own the process, but they're auditing exceptions and tuning the rules instead of typing.
Read the full case studyReconciliation time per cycle
Systems integrated end-to-end
Annual saving
Questions clients ask before they book the call
Why not just use Zapier and skip the agency?
Use Zapier where it fits — it covers most of the well-trodden integrations cleanly. Where you need us is the boundary cases: schema drift handling, idempotency on retry, approval queues with deadlines, and custom adapters when no off-the-shelf connector exists. We often build with Zapier or Make where they fit, and drop a custom service alongside for the parts they don't cover. Cheaper than rebuilding the whole stack from scratch.
What happens when an upstream API breaks?
Every external call has retry with exponential backoff plus a dead-letter queue for failures that exhaust the retry budget. Items in the dead-letter queue carry full context — the trigger, the input data, the failed step, the upstream error — so a human can replay them once the upstream recovers. Workflows are idempotent by default, so a manual replay never double-creates a record. We also alert on dead-letter accumulation rather than waiting for a customer to flag missing data.
Can you integrate with our system?
Almost certainly. We've integrated with Salesforce, HubSpot, Xero, MYOB, Microsoft 365, Google Workspace, Cliniko, custom-built databases, and a long tail of internal systems. The integration approach is always either a vetted connector for the well-supported APIs, or a dedicated adapter where the off-the-shelf connector doesn't fit. The discovery call covers this in the first 15 minutes.
Where does my data live? Australia, or overseas?
Your choice. We can run the orchestration on n8n self-hosted in your own AWS or Azure tenancy in your region. The default cloud option runs on Vercel infrastructure in Sydney. For workflows that touch sensitive data — financial, health, customer PII — we recommend the self-hosted route and document the data flow so your privacy officer signs off cleanly.
How long does a typical project take?
Business Process Automation projects on the Scale tier ship in 4-6 weeks from kickoff. Automate-tier projects ship in 2-3 weeks. Transform-tier projects (workflow suites, custom adapters, ongoing monitoring) typically run 8-12 weeks. The first two weeks are always discovery — mapping the existing process, identifying the failure modes, and agreeing on the boundary between automated and human-approved steps.
Free 30-min audit · No prep required
See which processes pay back fastest.
Book a free 30-minute audit. We'll walk through your repeat workflows, name the highest-value candidates, and tell you whether automation is the right fit — or not.