Deputy runs the roster for thousands of Australian venues, from Sydney cafes to multi-site retail chains. It publishes shifts, tracks clock-ins and pushes timesheets to payroll. What it does not do well is answer the plain questions an owner asks on a Monday: who is drifting over their agreed hours, what did Saturday actually cost in wages, and which sites are carrying too much idle labour. Those answers usually live one export away, in a CSV nobody has time to open.
Claude closes that gap. You export the data you already have from Deputy, hand it to Claude, and ask the question in words. No new dashboard, no BI project, no waiting on a report that arrives a fortnight late.
The gap between a published roster and a straight answer
A published roster tells your team when to turn up. It does not tell you whether the plan is sound. Deputy stores the underlying numbers, but pulling insight out of them means filtering, pivoting and cross-checking against award rules by hand. Most small operators simply do not, so the roster runs on gut feel and last week's habit.
The questions that go unanswered tend to be the expensive ones:
Which staff are scheduled past their availability or their agreed maximum hours this fortnight?
What is the forecast wage cost of next week's roster, broken down by site and by day?
Where are we double-covered on quiet shifts and short-staffed on the busy ones?
How much of last month's wage bill went to penalty and overtime loadings versus base rates?
Are any casuals creeping toward the point where a permanent conversation is fairer and cheaper?
What Claude does with a Deputy export
Deputy lets you export rosters, timesheets and shift data as CSV or Excel. Once that file is in front of Claude, you can ask questions in ordinary language and get an answer with the working shown. Claude reads the columns, does the arithmetic, and explains how it got there so you can sanity-check the logic rather than trust a black box.
A single mid-size venue might roster 40 staff across 300 shifts a fortnight. Reviewing that by eye to catch cost creep is a couple of hours of an owner's week. Claude turns the same review into a two-minute question. If it surfaces even one over-rostered day worth $1,200 in avoidable wages a week, that is roughly $60,000 a year of margin you were quietly handing back.
A worked example
Say a hospitality operator exports a fortnight of timesheets and asks Claude to compare rostered cost against takings. Claude flags that Tuesday nights across two sites are staffed for a Friday-sized crowd, adding about $800 a week in wages against soft sales. It also notices a casual who has worked steady 30-hour weeks for three months, which is worth raising with your adviser. The owner trims two Tuesday shifts and keeps the roster honest. On an annual wage bill of, say, $620,000, small corrections like these compound quickly.
Where the awards make this harder than it looks
Rostering in Australia is not just arithmetic. It sits under Fair Work and the relevant modern award, whether that is the Hospitality Industry Award, the General Retail Award or another. Penalty rates, casual loadings, minimum shift lengths and break rules all change what a shift really costs. Claude is genuinely useful for reasoning about these, but treat it as a fast, well-read assistant, not the final word.
Keep these limits in mind:
Award rates and rules change; confirm current figures against the Fair Work Ombudsman or your payroll system before acting.
Claude works from the file you give it, so a stale or partial export produces a confident but wrong answer.
Compliance calls, terminations and casual conversions are decisions for you and your adviser, not the model.
Deputy has no live API in this setup, so numbers are a point-in-time snapshot, not a real-time feed.
Keeping it safe and repeatable
Rostering data can include names and hours, which are personal information under the Privacy Act. You rarely need identities to answer a cost or coverage question, so strip names to initials or staff IDs before exporting where you can. Ask the questions you would be comfortable putting in an email, and keep the exports somewhere your business already controls.
There is also a governance benefit. Because you ask each question in plain language and Claude shows its working, anyone in the business can repeat the check next week without special training or a licence to a reporting tool. That keeps the knowledge in the business rather than locked in one person's spreadsheet, and it means a relief manager can spot the same cost creep an owner would.
Done well, this becomes a weekly habit rather than a one-off. Export the roster, ask the same five questions, act on what stands out. If you want help turning that into a repeatable routine for your own Deputy setup, book a brainstorm with us and we will map it to how your venue actually runs.
