Work that shouldn't be done by hand.
We build AI systems that take over a specific job in your business from start to finish: answering enquiries, processing documents, or keeping your software in sync. Then we keep them running. You own the code and the workflows, your data stays in Australia, and nothing is offshored.
Using AI isn't the same as having it do the work.
Most Australian businesses now use ChatGPT and Claude to write emails and drafts. Almost none have changed how an enquiry gets answered, how a document gets filed, or how two systems stay in sync. The admin is still there; it's just better written. The hours are in the gap between "we use AI" and "this runs itself", and closing that gap means building a system, not writing a better prompt.
How a system does the job.
Every system we build does the same four things, in the order the job needs. Open "See what we built" on any of the five jobs below to see the steps as they run today.
- 01 Reads whatever arrives: an email, a form, a booking, a document, a webhook or one typed sentence.
- 02 Decides what should happen next, under the rules you set.
- 03 Acts in the software you already use: the calendar, the CRM, Xero, the inbox.
- 04 Asks a person before anything critical that touches customers or money goes ahead.
Five jobs we automate.
Each one is something we've already built, shown with a real system that runs it. If your version looks different, that's what the free assessment is for.
Claude & ChatGPT connectors
For teams who already use Claude or ChatGPT and want it to work inside their own software.
Your team asks their own Claude or ChatGPT in plain English, and it does the work inside your software. It changes the website, sends a client a report or updates a customer's record. We write the instructions that teach it how your business works, and connect it through a private connector (an MCP server) that can only do what you allow. Anything that can't be undone, like publishing a page, waits for a person's yes.
A set of instructions for Claude (a skill) that we wrote and keep up to date, next to the connector for Sanity, where the store's content lives. In the client's own Claude, new landing pages, wording changes, imported articles and A/B tests are all done by asking, and published only when the client says so.
- 01 A person
The client asks
"Make a landing page for the new campaign", typed into their own Claude.
- 02
Claude reads our instructions
the skill we keep up to date, with every block and field on the site and the rules for changing them.
- 03
A draft is built in Sanity
through Sanity's connector, with images uploaded by a small one of ours. Nothing is published yet.
- 04 A person
The client publishes
after checking the draft. Nothing goes live until they confirm.
The resultA new landing page needs a request, not a developer.
Enquiries & bookings
For tour operators, clinics, education providers, and anyone else whose inbox is a to-do list.
An enquiry arrives at 9:40pm. The system reads it, answers in plain English, checks real availability, holds the spot, writes the booking into the calendar and the CRM, and sends the confirmation. Anything unusual waits in a queue for a person, with a drafted reply attached.
A booking platform connected to Rezdy, their booking system, with an AI assistant for staff that drafts the daily supplier emails from live bookings and answers booking questions in plain English, using live data only.
- 01
Bookings come in
copied from Rezdy, their booking system, into a staff tool as they arrive.
- 02
Passengers grouped by supplier
every afternoon at 3pm, for tomorrow, with each supplier's rules applied: cut-off times, capacities, seasonal allocations.
- 03
Claude drafts every email
one for each of the fifteen or more suppliers, written from tomorrow's bookings.
- 04 A person
Staff check and send
they read, adjust and send each email, and every email is logged.
The resultA job that took three hours every afternoon is now a review.
Documents & reports
For brokers, advisers, agents, legal, licensing, and any service that runs on paperwork.
A client report that someone would put together by hand is built by the system instead: it pulls the data from the source, writes the document, finds the client's record, and either sends it or leaves it for one person to approve. The same works in reverse: incoming documents are read and checked against your rules, their key details go into the client's file, and anything missing is chased up.
A connector we built and host (an MCP server), which Joris added to the Claude and ChatGPT accounts he already pays for. He asks in a normal conversation, our tools do the work, and when we add a new ability there's nothing for him to install.
- 01 A person
Joris asks
"Send Sarah the report on 12 Smith Street", typed into his own Claude or ChatGPT.
- 02
The report is generated
on the property data service, and every detail is fetched back.
- 03
Sarah is found in HubSpot
where Joris keeps his contacts.
- 04
The report goes out
formatted and emailed to Sarah. Her HubSpot record is updated with a note and a follow-up to-do, and Joris is told it's done.
The resultOne sentence sends a client a property report and updates HubSpot, with no new app to log into.
“They built around how I already work instead of handing me another system.”
Joris Cuesta, Joris Buyers Agent review on Clutch Systems kept in sync
For businesses juggling a booking tool, accounting software, a CRM and spreadsheets.
A booking changes in one place, and the invoice updates in Xero, the CRM record follows, and the operations spreadsheet stays correct, without anyone typing anything twice. Nobody has to ask "which one is right?" any more.
A portal where hundreds of makers see their own sales, pulled live from Shopify, and where staff run the fortnightly payout. Whether a sale happens in store or online, nothing is typed in twice.
- 01
Every sale is recorded
in store and online, in Shopify, with the maker who made the product.
- 02
Each payout is worked out
every fortnight, from the cost of goods, adding GST where the maker is registered.
- 03 A person
Staff confirm the figures
and adjust anything that needs it, before any money moves.
- 04
Everyone is paid
each maker is emailed their report, and the bank file is created for one batch payment.
The resultWhat took a day of spreadsheets is now one click, with the numbers logged.
AI apps
For businesses with a job that needs watching every day, like ad spend, stock levels or prices.
Some work isn't one task but a whole job that needs checking every day. An AI app takes it over. Every morning it reads the numbers, makes the changes it's allowed to make and sends a short report, with a yes-or-no question when something needs a person. The spending limits sit in separate code with no AI in it, so they hold whatever the AI decides.
The system that has run our own Google Ads. Claude works on a schedule, connected to Google Ads and Google Analytics, next to a separate guard that enforces the spending limits and can only hold spending back, never raise it.
- 01
The account is read
every morning: yesterday's spend, searches and leads from Google Ads, and what each visitor did on the site.
- 02
Claude makes the changes
it blocks searches that bring the wrong people, pauses keywords that cost money without results and adjusts bids, inside set limits.
- 03
A guard checks the spending
separate code with no AI, every six hours. It pauses the ads at the monthly cap, or once a set amount is spent without a lead.
- 04 A person
The owner gets a report
every morning, in plain English, by phone and email. The budget only goes up with their yes.
The resultA job nobody had time for is done every morning, and the budget only goes up with a yes.
Not on the list?
The five above cover most of what we're asked to build, not everything that can be built. If a task is repeatable, follows rules and takes up hours, it can probably be automated. The free assessment is how we find yours: a 60-minute call about the manual work in your business, then a written proposal within three business days that ranks each process by what it costs you today.
Assessment. Build. Managed.
You don't have to arrive knowing what to buy. That's what step one is for.
- Step 1 · a 60-minute call Free
Assessment: decide what's actually worth building.
Book a free assessmentA 60-minute video or phone call about the work that takes up your team's hours: what arrives, who handles it, how long it takes and what goes wrong.
Within three business days, you get a written proposal: the jobs worth automating first, what each one costs you today, and a fixed price for each build. If the honest answer on something is "don't automate this", the proposal says that too.
We don't charge for it, because the proposal is how we win the work.
- Step 2 · 3 to 8 weeks
Build: one process, start to finish.
This takes three to eight weeks. We connect the system to the software you already use and run it alongside your current process, so your team can test it on real data before anything goes live. After launch, we fix any bug in what we built at no charge for 12 months.
- Step 3 · monthly
Managed: keep it working.
AI systems need looking after: the AI models change, the software they connect to changes, and your business changes. We watch the system and fix it when something shifts. Improvements are quoted as they come up. Every month, you get a report of what it handled, what it saved and what it got wrong.
Built so you're never stuck with us.
-
You own the code and the accounts.
The code and the cloud accounts are in your name, there are no per-task fees, and you can leave any time.
-
Data stays where you say.
Your data is hosted in Australia by default, or on your own servers when it needs to be. We'll tell you plainly which AI providers see what.
-
Never offshored.
Designed, built and looked after in Australia, by a small studio with a network of Australian specialists.
-
A person approves what matters.
Anything that touches customers or money waits for a person's approval until you say otherwise. Every action is logged, and every decision can be explained.
-
Built for regulated businesses.
We've run live systems for ASX-listed groups since 2021. Access controls, reviewed changes and audit trails are built in from the start.
-
Connects to what you have.
You keep your CRM, accounting, booking tool and inbox. We connect them.
-
12 months of bug fixes.
For a year after launch, we fix any bug in what we built and apply critical updates, at no charge. The managed plan covers what changes around it.
Works with the software you use.
If your software has an API (a way for other programs to plug into it), we connect through that. If not, we can usually find a workaround.
- Xero
- MYOB
- HubSpot
- Pipedrive
- Google Workspace
- Microsoft 365
- Calendly
- Shopify
- Airtable
- Slack
Plus older custom systems built just for your business.
What it costs.
- Automation Opportunity Assessment a 60-minute call · a written proposal with prices within 3 business days
- free
- Automation build one process or an AI app · typically $10–30k
- from $8,000
- Claude & ChatGPT connector your own Claude or ChatGPT, connected to your software · custom connectors typically $3–8k
- from $1,500
- Managed monitoring, fixes, monthly report · changes quoted
- from $250/month
What affects the price:
- how many systems we connect
- whether steps need a person's approval
- how messy the source data is
- how many enquiries, documents or bookings it handles
Why not cheaper?
A system that saves 15 hours of admin a week replaces $60–80k a year in wages and on-costs. We price against that saving, not our hours, and the managed fee is what keeps it working months after launch.
These are guide prices. You get a fixed price after the assessment.
Five systems already running.
Ask away.
Fourteen questions people ask before they book. Have another question? Ask us.
We already use ChatGPT. What's different?
ChatGPT helps a person do a task. We build a system that does the task itself, connected to your calendar, CRM, inbox and files, and it only asks a person when it should. When it makes sense, we connect our tools to the ChatGPT or Claude you already pay for, through a connector we host (an MCP server), so you ask in plain English and the system does the work.
Is this workflow automation?
Partly. Workflow tools like Zapier, Make and n8n move data between apps when something happens, and we use them where they fit. What we build goes further: a system that reads the enquiry, document or booking, decides what to do under rules you set, acts across your tools, and asks a person when it should. A workflow tool is one part of it.
Is this the same as AI agents?
Mostly, yes. An AI agent is software that reads, decides and acts on its own, and the systems we build do that. The difference is the limits we give them. They work inside your own software, follow the rules you set, and ask a person before anything that touches customers or money.
Do you build chatbots or AI receptionists?
No. Website chatbots and AI receptionists are ready-made products sold by the month, and for a simple job one of them may be all you need. We build the work that happens after the conversation, such as the booking written into your calendar and CRM, the document filed or the invoice matched. If a ready-made product would do the job, the assessment says so.
What if the AI gets it wrong?
It will, sometimes. That's why anything important waits for a person's approval, every action is logged, and the managed plan exists. We measure and report how often it makes mistakes, rather than guessing.
Where does our data go?
It's stored in an Australian cloud region by default, or on your own servers if you need that. We document exactly which providers handle what, and we never use your data to train AI models.
Do we have to replace our systems?
No. We connect to the systems you already use.
Doesn't Xero or MYOB already do this?
For some jobs, yes. Xero and MYOB can already read a bill and suggest a bank match, and we won't rebuild what your software already does. We automate the work between your systems, like the job that starts in your inbox, needs your CRM and ends in Xero, which no single product covers. Supplier invoices, for instance, can be matched to the purchase order, so accounts payable only sees the ones that don't match. Sales from Shopify, Square or Stripe can be reconciled in Xero where the standard connector falls short. And a deal won in HubSpot can become a Xero invoice without anyone retyping it.
What does this mean for our staff?
The system takes over the copying, checking and chasing, the parts nobody was hired for. Judgement calls, unusual cases and customers stay with your team. In the small and medium businesses we work with, the usual result is not needing the next admin hire, rather than letting people go.
How long until something's live?
The assessment is a 60-minute call, and the written proposal follows within three business days. The first process is usually live 3–8 weeks after you give the go-ahead.
What does it cost to run?
AI usage fees are usually tens to a few hundred dollars a month for a process at a small or medium business. The managed fee covers our time, and you'll see both on the monthly report.
Can we get funding for this?
Sometimes. Federal and state programs exist and change often. We keep a current list on the site and will tell you what applies.
What if we outgrow it, or want to leave?
You own everything. We hand it over with full documentation, and it keeps running without us.
Do you do the "AI strategy" thing?
Yes, through the assessment: it's an AI strategy with prices attached, and it's free. Then we build it.
The long answers
Book a free assessment.
Tell us what takes up your team's week. After a 60-minute call, you get a written proposal within three business days: what to automate first, what it saves, and a fixed price for each build.