Case study · Field-service / facilities contracts

Perth Reticulation Experts

Hundreds of contract work-orders a month, hand-keyed one by one. We automated the intake — and gave back a full quarter of a full-time admin's year.

Perth, WAOperations automationLive · ongoing

$23,400/yr: Admin time reclaimed (≈520 hrs)at $45/hr · conservative estimate
6 workflows: Running live since early 2026
24/7: Work-order triagewas manual, business hours only

Case studiesPerth Reticulation Experts

The situation

Where they started

Perth Reticulation Experts runs a major national facilities-management contract — which means a firehose of inbound email: new work orders (each with a PDF and a property to match), version updates, cancellations, claim confirmations, reminders. For a long time, one admin — call her the office manager, on about $45/hour — sorted every one by hand and keyed the real jobs into ServiceM8 one at a time. It worked, but it was slow, repetitive, and entirely dependent on one person being at her desk.

What was at stake

On a contract like this, admin capacity is the ceiling. Every hour spent copying a work order off a PDF and into the job system is an hour not spent on the work that actually earns — and a single mis-keyed or missed job is a job that runs late, which on a facilities contract is a reputation and renewal risk. Growth was quietly capped by how fast one person could type.

What we did

The approach

Phase 1 — automated the intake (live)

We built a set of six n8n workflows that now run the inbox end to end. An hourly router reads every incoming email and classifies it. Work requests are parsed by a cheap LLM that pulls the property ID straight off the PDF, matches it to the right site, and creates the ServiceM8 job — with the PDF attached and staff allocated. Cancellations are handled automatically; anything ambiguous is logged and flagged for a human. A daily digest and an error monitor sit on top. Humans only touch the judgment calls.

Built the way money-path systems should be

The office-automation glue runs in n8n; anything that touches jobs, stock or the ledger is built in typed, retried, reconciled code — not a fragile drag-and-drop flow. Every action is logged for audit. A few hundred emails have already been classified and enriched into a clean dataset, which is what makes the next phase possible.

Phase 2 — the full pipeline (scoped & costed)

The next round, already scoped and dollar-estimated, automates the rest of the pipeline: a single property register as the source of truth, a work-order ledger, a portal agent that files claims, a pricing engine that protects a revenue-critical review step, and an auto-drafting query responder. Conservative forward saving from the cyclical-portal step alone: about two weeks of admin work per round, three rounds a year.

The proof

The numbers

Metric	Before	After	Change
Work-order intake	Hand-keyed, business hours	Automated, hourly, 24/7
Admin time on data entry (Phase 1)	~1 quarter of an FTE/yr	Reclaimed	≈520 hrs = $23,400/yr · est.
Workflows in production	0	6 + error monitor
Cyclical portal entry (Phase 2, scoped)	~2 weeks/round × 3/yr	Automated	forward estimate

Phase 1 is live. Time-saved figures are conservative estimates derived from the documented manual workload, not an audited time-and-motion study; Phase 2 figures are scoped forward estimates.

The result

What it added up to

The contract's admin no longer scales with a person typing. Work orders land in the job system around the clock, the office manager's time shifts from data entry to the exceptions that actually need a human, and the business can absorb more contract volume without adding headcount. Conservatively, that's roughly **520 hours a year — about $23,400 at a $45/hour admin rate — given back**, every year, from the intake automation alone. It's the same principle behind everything Salty builds: the owner (and the team) get their time back, and the system keeps running when nobody's watching it.

What this means for you

Most trades and contractors have one or two people drowning in repetitive admin that a machine should be doing. We find the highest-volume, lowest-judgment task, automate it properly, and measure the hours it gives back — then move to the next one. This is the operating system we ran to take our own business to $4M with the owner on the other side of the Tasman.

Workflow automation (n8n)LLM document extractionServiceM8 integrationOps consulting

What's your team doing by hand that a machine should do?

Book a call. We'll find the task eating the most hours and tell you what it costs — and what automating it would give back.

Book a call More case studies