Claude Implementation: A 30-Day Rollout Plan for Australia
A Sydney business rang us in March.
"We bought Claude Teams in February. It's six weeks on. Two people out of twenty are using it."
The problem wasn't Claude. And it wasn't that the staff were lazy or scared of technology. The problem was that they'd bought licences and called it a Claude implementation.
Buying the licences is step one of about ten. The rest is structure, training and change management, and there's no shortcut through it. The slow way is the only way that works.
So here's what actually happens, week by week, in a rollout that sticks.
What is a Claude implementation, really?
Three layers. Most organisations get through the first and park there.
Layer one: the technical foundation. Licences, admin console, privacy assessment, SSO if you need it. One to three days.
Layer two: training. Role-specific sessions, prompt setup, shared Projects with real context in them. Two to four weeks.
Layer three: culture. Internal champions, measurement, iteration. This one never really finishes.
Organisations that stop at layer one sit under 20 per cent adoption after three months. The ones that get through all three tend to land at 60 to 80 per cent in the same window1.
Why the gap? Because technology without training is just a cost line. And training without culture evaporates in a fortnight. That's the whole thesis.
Days 1 to 3: the technical foundation
Fast, unglamorous, and worth doing properly.
Pick the plan first. For most Australian SMBs that's Claude Team Standard at US$20 per user per month billed annually, roughly A$30, minimum two seats, which gets you the admin console, SSO and shared Projects. Larger organisations with SAML or procurement requirements go Enterprise. If you'd rather have AUD invoicing and someone handling the administration, satori-claude sits on top of the licence. Full numbers are in our Claude pricing guide, because this article is about the process, not the price list.
Then, in the admin console, in this order:
- Invite users by CSV or manually
- Set up billing, and decide now whether you want USD direct or AUD via a local partner
- Configure SSO if you need it
- Create shared Projects per department
- Load your foundational context into them: product info, FAQs, tone guide, process docs
- Set user policies, who can do what
- Verify every single person can log in and reach their Project
That last step sounds trivial. It is not. A rollout where three people quietly couldn't log in on day one is a rollout that's already lost those three people.
Privacy Act groundwork
Do this now, not later. The OAIC expects organisations using commercially available AI products to take a privacy-by-design approach, including a privacy impact assessment before deployment2. Practically, that means three things: assess the product against the Australian Privacy Principles (APP 8 matters here, because data crosses borders), document Claude in your AI register, and tell your staff in writing that they're using an AI tool and what they may and may not put into it.
The National AI Centre's Guidance for AI Adoption, the six practices known as AI6, is the reference to work from. It's voluntary guidance that complements the Privacy Act rather than adding new obligations3. Half a day of work. Skipping it turns a technical rollout into a legal problem later.
We go deeper on all of this in our guide to Claude and the Australian Privacy Act.
Days 4 to 10: the pilot team
This choice matters more than any other in the whole rollout.
Start with: sales, marketing or customer service. They see results fastest, they talk to everyone in the business, and they become your best internal advocates without being asked.
Don't start with: IT or engineering. Not because they shouldn't use Claude. But their use cases run deeper, and the results don't spread through the organisation the same way.
What the pilot fortnight looks like:
- Build a department Project with real context in it: product sheets, FAQs, exported customer profiles, proposal templates
- Run a two-hour workshop focused on three tasks the team already does every single day
- Set a weekly "Claude hour" where people share what they tried and what actually worked
- Measure at the end of week two: how often are they logging in, on what tasks, and where do they get stuck
The key move: always start from their daily work, never from Claude's feature list. Ask "what eats the most time in your week?" and begin there. A team that has automated one genuinely annoying task is converted. A team that's been shown twelve features is exhausted.
Days 11 to 20: scaling out
The method is "teach it forward". The pilot team runs a one-hour session for the next department. They know the real use cases better than any consultant does, and they're colleagues rather than outsiders. That difference is bigger than it sounds.
Every new department gets three things:
- Its own Project with department-specific context loaded
- Role-specific prompt templates, building into a shared internal library everyone contributes to
- A named internal champion who is the go-to person for questions
Measure weekly. Active users, frequency, which tasks, where people stall. Measurement isn't bureaucracy. It's the only way to know whether any of this is working while you can still change course.
Format and pacing for the training itself is a topic of its own, and we've covered it in the Claude training guide.
Days 21 to 30: measurement and what comes next
By day 30 you want to know three numbers.
Adoption: what share have logged in over the last seven days? Target 60 per cent or better for the core team.
Frequency: how often does the average user reach for Claude in a week? Daily for the roles where it fits best.
Use cases: which tasks are people actually solving? That list tells you where the value really is, and it's almost never where you predicted.
Then you can plan the next phase honestly. Connecting Claude to your CRM, database or project tools through MCP so it answers questions about your own data. Putting Claude into Excel and the rest of Microsoft 365 so it lives where people already work. Or Claude Code for the engineering team. Integration work like this we scope and quote through satori-builders, once the foundation is stable and not before.
Why do Claude implementations stall?
Five reasons, and we see the same five over and over.
"We'll send out the logins and see what happens." Almost nothing happens. Without a concrete starting point people open the tool, look at a blank box and close it. Always plan a kickoff.
Rolling out four AI tools at once. Analysis paralysis. Start with one. Add the next when the first is genuinely embedded.
Generic training for everyone. A salesperson and a developer share nothing in how they use Claude. Role-based training runs roughly three times faster to adoption.
Skipping the privacy step. Unclear data handling and no policy isn't a technical issue. It's a legal one, and it surfaces at the worst possible moment.
No internal champion. If every question routes to IT or to your consultant, momentum dies inside a month. One enthusiast per department, with actual time allocated.
How Satori runs it
We run this as a fixed-scope engagement, not an hourly meter.
satori-launch covers the full implementation at A$990 per user, minimum eight users: plan, technical setup, role-based training, AI policy and the measurement structure. satori-claude handles licences, AUD invoicing, onboarding and ongoing support at A$850 setup plus A$15 per user per month. And satori-run keeps the thing improving after go-live, from A$400 a month.
The question we get most often is "couldn't we do this ourselves?" Yes. Genuinely. But rollouts run without support take substantially longer and land at about half the adoption, and by then the licences have been burning for a quarter.
Where to start
You've got the plan. The question was never whether to implement Claude. It's whether you do it in a way that produces something.
Pick one department. Pick three tasks they do every day. Give them two weeks and someone to ask questions of. That's the whole thing, honestly.
Want a hand with the sequencing? Get started and we'll map it to your size and your systems, or read more about how we work as a Claude consultant.
Sources
Footnotes
-
McKinsey & Company (2025). The state of AI: How organisations are rewiring to capture value. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai ↩
-
Office of the Australian Information Commissioner (2024). Guidance on privacy and the use of commercially available AI products. https://www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/guidance-on-privacy-and-the-use-of-commercially-available-ai-products ↩
-
National AI Centre, Department of Industry, Science and Resources (2025). Guidance for AI adoption: foundations. https://www.ai.gov.au/staying-safe-and-responsible/essential-ai-practices/guidance-ai-adoption-foundations ↩