Claude, set up for a trade business.
Claude is the AI assistant our own tools run on. We set it up for your team: Projects that hold your price lists and procedures, Skills for the jobs you do every week, and connections to the systems you run, read-only first.
Your documents. Your jobs. Your rules.
Projects for your documents
Price lists, procedures, warranty terms and templates in one place, so answers come from your business, not the internet.
Skills for the weekly jobs
Written steps Claude follows every time: the monthly revenue question, the quote follow-up, the job summary for the customer.
Connected to your systems
ServiceM8's MCP connection and HighLevel's, set up read-only first, so Claude can answer from your real jobs and customers.
Rules for the team
No customer data in free tools, model training switched off, an AI policy everyone has read, and a list of the AI tools in use.
How we use it ourselves.
125 calls down to 3
A Claude skill answers a revenue question from ServiceM8 in 3 API calls instead of 125: about 3 minutes instead of 28, in a two-case test at Tune Cycles.
Answers without the secrets
Tune Cycles' staff handbook is rebuilt every week behind a guard that holds back costs, supplier terms and financials, and the team asks an assistant built on it.
You approve. AI earns its way in.
Every build starts small and read-only, and gets write access one step at a time.
Dry run first
Every tool shows what it would change, and changes nothing until a person says yes.
You approve what matters
Quotes, refunds, anything sent under your name: a person signs off before it goes.
Customers know it's AI
Chat and voice agents say they are AI, offer a person, and announce recording before a call is recorded.
Your data stays yours
It lives in your own Simpro, ServiceM8, CRM and AI accounts. Nothing is copied into a system we own.
Built to Australian rules
Call-recording consent, the Spam Act for texts and email, the Do Not Call Register for AI voices, and no AI-written reviews.
Handed over properly
Every build ends with an AI policy and a register of the AI tools you use, with one named owner.
General information, not legal advice.
One job first. Measured.
We don't start with a platform. We start with the task that eats the most hours, and prove it.
- 01
A fifteen-minute call
Which job eats your week? Which system holds it? We tell you straight if AI is the wrong tool.
- 02
We time it
Minutes per task, on your own jobs, before anything changes. That number is what we have to beat.
- 03
We build one thing
It runs as a dry run first. You see every change it would make, and you approve it.
- 04
We train the people who use it
Role by role, on your own system, with your own jobs. The go-to person in the office first.
- 05
A month later, we check
Is the task faster? Who still uses it? What needs changing? Then we pick the next job.
Every build is quoted before it starts. No platform plans, no lock-in.
Claude questions.
Claude or ChatGPT?+
Both are good. Our tools run on Claude, and that's what we set up and train on. If your team already uses ChatGPT or Microsoft 365 Copilot, we'll work with that.
Which plan do we need?+
You hold the subscription, and we help you pick between Claude Pro, Max and Team for each person. We do not resell it.
Is our data used to train the model?+
We set it up on a plan and settings where your conversations aren't used for model training, and check that at set-up.
Find the hours first.
Fifteen minutes, no pitch deck. Tell us which job eats your week and we'll tell you whether AI is the answer yet.
Hayden Nosatti · Leadout · Sydney