Case study
One year inside a Sydney rail contractor.
From Discovery to a leaner, AI-supported business: new systems, new apps, and agents carrying the repetitive work across bidding, operations and people management.
- Client
- Rail safeworking and labour-hire contractor, Sydney
- Workforce
- Around 200 field and office staff
- Customers
- Sydney Trains, Transport for NSW and major rail contractors
- Engagement
- Ongoing, starting with Discovery
The starting point
A busy contractor running on a scheduling and fatigue-management system built in-house more than ten years ago, hosted on premises. Quotes, tenders, rostering, reporting and people matters were all handled by hand, by a small office team supporting a large field workforce.
Discovery mapped how the business actually ran: where the hours went, which processes repeated every week, and where errors and risk crept in. That gave the business a ranked list of where AI would save the most, and the evidence for a leaner structure that went on to deliver a seven-figure saving.
What changed
The results so far, across the operating model, systems and day-to-day work.
New systems, in people's pockets
The old on-premises system was replaced by a cloud-based staff management system, with two mobile apps built on top of it.
Cloud staff management system
The ten-year-old in-house scheduling and fatigue-management system was migrated to the cloud, so staff data is available anywhere and connects to the new agents.
Management app
Managers see every employee's details, location and work status on their phone, find available staff to cover last-minute cancellations, and track live projects and progress so resources can be moved where they're needed.
Staff app
Rolled out to around 200 employees in a short time frame. Staff receive and confirm their jobs on their phone, and their responses flow straight back into the management system.
Agents doing the repetitive work
Each agent works to rules built with the business. People review and approve, with the right delegation.
Contractor quotes
Every inbound labour request is priced against the right customer's negotiated rate book, with shift rules, penalties and minimums applied, and suspect rates flagged rather than silently used.
BiddingRFQs and tenders
Responses drafted from the tender documents: requirements extracted, schedules completed, questionnaires answered from one verified company profile.
BiddingTender review
Every headline number recomputed from the underlying rows. The agent caught a 730-hour error in a government client's own tender spreadsheet, and a claim in a draft response that would not have held up, before either was lodged.
BiddingDirector approval briefs
Contract value, margin by rate code and grade, the defects worth money, and the actions needed before lodging, on one page for the director to approve.
BiddingMargin and rate analysis
Rate cards tested cell by cell against real cost, exposing loss-making lines hidden inside healthy averages and the rostering changes that recover them.
CommercialContract administration
Long panel contracts mapped clause by clause, so rate reviews, extensions and notices are grounded in what the contract actually says.
CommercialManager support agents
Each manager has their own agent supporting their workflow, absorbing admin that previously needed extra headcount.
OperationsPolicies and performance
Policies and procedures monitored, and performance matters prepared and run to the company's own process, with every action approved by the relevant line manager.
PeopleWeekly workforce comms
The Friday update to the whole workforce, covering contract wins, new starters and safety instructions, built to the house format every week.
PeopleSafety and workforce statistics
Safety, headcount and hours figures recomputed from source records before they go into a tender, so every number can be defended.
ComplianceCompany knowledge base
Every decision, contact and lesson from a bid is logged, and win/loss lessons become house pricing rules. Knowledge stays with the business when people move on.
KnowledgeAdmin taken off staff
Across the business, repetitive admin has moved from people to agents, so staff spend their time on the work that needs their judgement.
EverywhereThe AI scheduler
Putting the right person on every job, automatically and within fatigue management rules, with a human approving every allocation.
-
Job comes in
A client request lands in the staff management system.
-
Skills assessment
The scheduler works out what the job actually needs: beginner, moderate or advanced skill level.
-
Find the right person
It finds available staff near the location, with the right skills, who are within fatigue management rules, and allocates fairly based on each person's previous allocations.
-
Offer to the staff app
The job is pushed to the person's phone to confirm.
-
Confirm or reallocate
Their answer flows back to the management system. Declined? The scheduler goes back to step 3 and finds the next best person.
-
Human approval, then client confirmation
A scheduler approves the allocation, and the client receives confirmation.
How it followed our five phases
The same stage-gated approach we offer every client.
| Discover | Mapped how the business really ran, measured where time and money went, and ranked the opportunities. |
|---|---|
| Design | A leaner process model and restructure, with AI built into the target processes. |
| Pilot | First agents proven on high-volume work: contractor quotes and RFQs. |
| Scale | Cloud migration, management and staff apps, manager agents, and agents across commercial, people and compliance work. |
| Sustain | A continuous-improvement backlog. The next wave is the AI scheduler. |
Want results like these in your business?
It starts with a discovery call.