How We Cut Rostering From a Week to 2 Minutes With AI (Case Study)

Every week, the same job came around for one of our clients, a Victorian NDIS provider. Someone had to sit down and build next week's roster: who supports which participant, on which day, at what time. It sounds simple. In practice it took about a week of calculations - checking what happened in past weeks, matching regular workers to regular participants, working around overnight and 24-hour shifts, and making sure nobody was left without support.

We built an AI integration inside their NDIS CRM that now does the heavy lifting. It reads the rosters from previous weeks and builds a draft roster for the next week in about two minutes. The coordinator reviews it, fixes anything flagged and publishes. The client is happy for us to share the story, so long as we keep their name private - and everything you see below uses sample data, not their real roster.

The problem: a week of calculations, every week

Rostering in disability support is harder than rostering in most businesses. Each participant has their own routine and their own preferred workers. Some homes need support around the clock. Workers have their own availability, and things change all the time.

Before automation, building the roster looked like this:

• Open the last few weeks' rosters and work out the regular pattern for each participant
• Copy shifts forward one by one, adjusting times and dates
• Match each shift to a worker who usually supports that participant
• Check overnight and 24-hour shifts carefully, because they cross into the next day
• Spot the gaps and start phoning around to fill them
• Double-check everything before sending it to the team

None of this is difficult on its own. It's the volume and the repetition that eat the time. And because it's done by hand, small mistakes creep in - a missed shift, a double booking, a worker rostered on a day they asked off.

What we built

The idea was simple: most of next week's roster already exists in last week's roster. Participants tend to have regular supports, and they usually prefer regular workers they know. So instead of starting from a blank week, we let AI start from the history.

The coordinator now presses one button - Build roster using AI - and the system:

1. Reads the rosters from previous weeks
2. Works out the repeating patterns - which participants are supported on which days and times, and by whom
3. Creates next week's shifts as drafts, so nothing is sent to workers yet
4. Flags anything that needs a person's attention, such as a shift with no available worker
5. Waits for the coordinator to review the week and publish it

What the coordinator sees

Here's a sample of what an AI-drafted week looks like. Every shift starts as a draft. Shifts with a warning can still be published once someone has checked them. Shifts that must be fixed are shown in red, and the week can't be published until they're sorted.

This is the important part: AI does the repetitive maths, a person makes the decisions. The coordinator can change any shift, swap a worker, add a one-off support or clear the drafts and start again. When they're happy, they publish the week, and workers can accept their shifts as usual.

Why keep a person in the loop?

We could have built a system that publishes the roster on its own. We didn't, on purpose. Rostering in disability support affects real people. A participant might have an appointment this week that wasn't in last week's roster. A worker might be unwell. A family might have asked for a change. A coordinator knows these things; a pattern in old rosters doesn't.

Keeping a person in control also fits how NDIS providers are expected to work: supports should respect each participant's choice and control, and the provider stays responsible for the quality and safety of the supports it delivers. AI is a tool that helps the team do that - it doesn't replace their judgement.

The result

The time saved doesn't disappear. It goes back into the work that actually needs a person: talking to participants and families, supporting workers, and dealing with the changes that always come up during the week.

Could this work for your business?

This case study is about NDIS rostering, but the same idea works anywhere a team rebuilds the same thing every week from last week's version: cleaning rounds, delivery runs, clinic schedules, shift-based hospitality and trades crews. If your team says "it's mostly the same as last week, but it still takes ages", that's a strong sign AI can help.

Not sure? Read how AI rostering works for NDIS providers, see how SIL rostering can work without spreadsheets, or work out what manual rostering is costing you. Or skip the reading and book a free check - we'll look at your current process and tell you plainly what's worth automating.