The Live Your Best Life story

The provider getting thirty hours a day back from admin

Live Your Best Life is a disability support provider in NSW. This is the honest story of their organisation-wide implementation of Claude, the AI assistant made by Anthropic: what it replaced, what we built, and what changed. Published with their knowledge, told without invented numbers.

Where they started

Every provider will recognise the picture. Support work generates paper: progress notes after every shift, incident reports that have to be right, service agreements, funding letters, audit evidence, HR records. All of it matters. All of it competes with the actual work of supporting participants.

The sector's maths makes it worse. The NDIA's own cost modelling leaves providers a margin of a few per cent, so every admin hour is paid for out of almost nothing. And the quiet risk had already arrived, at LYBL like everywhere: staff wondering whether ChatGPT could take some of the load, with no rules protecting participant privacy if they tried.

LYBL's leadership did not want to ban AI. They wanted it done properly.

The old way

It is 8:40pm. A support worker is still typing up the day's notes from memory, deciding how much detail is enough. An incident report is due and it starts from a blank page, written the way that staff member happens to write. Tomorrow, the coordinator will reformat half of it, and the quality manager will quietly fix the rest at audit time.

The new way

The same worker opens the incident project, pastes de-identified shift notes, and a complete draft appears in LYBL's own template, asking for the two details it still needs. The human reads it, corrects it, owns it, submits it. The writing took minutes. The judgement stayed where it belongs: with the person who was there.

What we built, and what each piece changed

An implementation is only as good as the outcomes it moves. Here is the build, paired with what it actually changed at LYBL.

Nine role-specific Claude projects
The blank page is gone. Every role starts its documents from a draft that already knows the job, the format and the rules.
Forty-two custom skills
The documents the team repeats every week come out in LYBL's format the first time, whoever writes them, however tired they are.
Knowledge bases from LYBL's own policies
Answers reflect how LYBL actually works, in LYBL's voice, not generic sector advice from the internet.
Privacy guardrails and organisation-wide rules
Participant personal information stays out of AI tools, de-identification is built into the workflow, and a human is accountable for every document. Shadow AI stopped being a risk because safe AI became easier than the workaround.
Role playbooks and per-person checklists
The setup survives staff turnover. The knowledge lives in the system, not in one person's head.

The training day

We trained the whole team in person, in one day, starting from zero. What AI is and is not, in plain language. The Claude app, button by button. The privacy rules and why they matter more in disability services than almost anywhere else. Then the part that changes minds: every person practising on their own real tasks, inside their own role's project.

People walked in ranging from sceptical to nervous, which is exactly right. Scepticism in this sector is earned. The turning point is always the same moment: watching a first draft of a document they hate writing appear in their organisation's own format, and realising the job just got lighter, not smaller.

What changed

Documents start at eighty per cent done

The blank page is gone. Notes, reports and letters begin as a strong first draft in LYBL's format, and staff spend their time on judgement and accuracy instead of typing.

Admin hours flow back to participants

Every hour AI carries is an hour that returns to support work, coordination and the human parts of the job. On provider margins, that is the whole game.

Shadow AI became safe AI

Instead of staff quietly experimenting with free tools, LYBL has business-grade AI with participant privacy guardrails, clear rules and accountable humans on every document.

Consistency a quality manager can trust

The same document type now comes out the same way regardless of who wrote it, which is exactly what auditors want to see.

"Having the team learn Claude has been incredible. I was really overwhelmed with how to use AI, and the change it has made cannot be overstated. The way Storm helped us set it up as a business means it is idiot-proof, which is exactly what I need. Our overheads are reduced by 35%, and the team is so much more productive, doing things they actually like doing. It has been a huge game changer for us."
Kahlia, General Manager, LYBL Disability Support Services

The number so far: thirty hours a day, team-wide

LYBL reports a team of ten staff, each saving a minimum of three hours a day on admin since the implementation. That is thirty hours back every working day across the team — roughly $6,750 a week and on track for $300,000+ a year, at their average wage of $45 an hour, if it holds. The arithmetic is 10 staff × 3 hours a day × $45 an hour, shown in full because we would rather you could check it than be impressed by it. Two things worth knowing about how we handle numbers: this figure is reported by the client, not invented by us, and a formal measurement is under way. When it is done, this page will publish the method alongside the result. The AI-for-NDIS market is full of precise-sounding numbers nobody can check. We will not add to them.

Why we tell it this way

You may have noticed what this page does not say. No invented percentages, no guaranteed audit outcomes, no anonymous miracles. We would rather tell you exactly what was built and let you judge. It is the same honesty you would get in an engagement, which is rather the point.

The model transfers to any provider: understand the roles, build a project per role on your own policies, add the guardrails, train the whole team, support it until it sticks. If you want it mapped to your organisation, that is what the first call is for.

Your organisation could work like this

Thirty minutes with Storm, and we will map what a role-based Claude setup looks like for your team.

Book a call