An AI operating system for your business.
The written layer that tells AI how your business works. What you do, what you are aiming at, how the work gets done, which tools are approved, what data may go into them, who checks the output, and what gets automated first. We write it with you, you own it, and every AI tool in the business reads from it. Built by the founder, priced in AUD.
What an AI operating system actually is
It is the written layer that tells AI how the business works. Not software, not another dashboard, and nothing to log into. One place that holds what you do, who you sell to, what you are trying to achieve, how a job actually runs from enquiry to invoice, the words you use and the ones you never use, which tools and models are approved, what data may go into them, who checks the output before it reaches a customer, and which job gets automated first. Some consultants call the same idea an AI operating model.
Every AI tool in the business reads from it. That is the whole point. Right now your team is typing the same background into a chat box twenty times a week and getting twenty slightly different answers, because nobody has written the business down anywhere a tool can use. Write it once and the drafts start coming back in your voice, with your prices, inside your rules.
We write it with you. Every job is led by the founder, so the person who maps how you work is the person who writes it up. It lives in your systems, in your language, and you own it. If you never speak to us again it still works.
Why AI keeps giving you generic answers.
It is almost never the tool. It is that the tool knows nothing about you, and nobody has decided anything, so everyone decides for themselves.
Every chat starts from nothing
Your team re-types the same background every time, badly and differently. The tool does not know what you sell, how you price it, or what you already promised that client last month.
Three people, three tools, three answers
Someone is on ChatGPT, someone put a second tool on the company card, and the same question gets three answers. You are paying for the overlap and trusting none of it.
Rules written after the fact
Nobody has said what staff may put into an AI tool, so they guess, usually generously. You find out what went in when a client asks about it.
What goes into it.
Twelve things, written plainly, short enough that people read them. Not every business needs all twelve. We tell you which ones yours needs on the first call.
Business context
What you sell, who buys it, how you charge, what a good customer looks like, and what you will not take on. The facts a new starter would need in week one.
Goals and the numbers behind them
What the business is actually trying to do this year and the handful of numbers that say whether it is happening. AI with no goal optimises for sounding good.
Competitors and how you differ
Who you lose work to and why you win the jobs you win. Written down, so a draft stops sounding like it could have come from anyone.
Knowledge in one place
Quotes, past jobs, policies, product detail, the answers you give on the phone. We find where it lives, what is out of date, and what has to be tidied before any tool can use it.
How you sound
The words you use, the words you never use, how formal you are with a client versus a supplier. This is what makes a draft usable instead of a rewrite.
How the work actually gets done
Your main jobs mapped end to end, as they really run rather than as the manual says. We redesign the task before automating it, so you do not automate the slow version.
The rules for using AI
What may go into a tool, what never can, what has to be disclosed, and what happens when an output is wrong. Written against the National AI Centre Guidance for AI Adoption so you are not starting from a blank page.
Approved tools and models
Which model does which job, which platform runs it, and which of the tools you already pay for stay. Decided once and written down. We are not resellers and not tied to anyone.
Who checks what
Every output gets a named reviewer or an explicit note that it can go straight out. Anything that reaches a customer gets a person. Not a vague promise that someone will look.
Who owns which process
One name against each automated job, so when it drifts there is somebody whose job it is to notice. This is the part most businesses skip and then wonder why nothing gets fixed.
How it gets measured
Two or three numbers per job and where each one comes from. If we cannot tell you where the number comes from, it is not a number.
The order it rolls out in
A ranked list with dates and names against it. First job is the one with the best ratio of what it saves to what it takes, not the most interesting one.
AI strategy and governance, without the boardroom version.
Strategy and governance are two sections of the operating system, not separate projects. Strategy is which jobs AI should do and in what order. Governance is the rules it runs under. For a business under 200 people both fit on a couple of pages, and both are useless unless the people doing the work will read them. There is no Australian AI Act, so we write your rules against the Privacy Act, whatever your industry adds, and the National AI Centre voluntary guidance. A practical starting point, not legal advice.
Who is accountable
One named person for AI decisions and one for each automated job. Not a committee and not everybody, which means nobody.
What may go in
Approved tools, and for each one exactly what data is allowed and what never is. Business tier tools that do not train on your inputs, never the free consumer version.
Who checks the output
A named reviewer on anything that leaves the business, and an honest note on the low-risk jobs that can go straight out.
What happens when it drifts
Models change, prices change, your business changes. We set the review rhythm and tell you what to watch for.
How it gets written.
Five steps. You will know whether it is worth continuing after the first one, which is free and costs you half an hour.
The first call, 30 minutes
Free, nothing to sign. How the business runs, what is eating your week, what you have already tried. You get a straight answer on whether this is worth doing.
Mapping how you work, 3 to 5 days
We follow your main jobs end to end, look at the tools you already pay for, and find where AI would actually save money rather than look clever.
Writing it up
The context, the knowledge, the rules, the tool and model choices, who checks what, and the rollout order. Short enough that you will read it, specific enough to act on.
Putting it to work
We wire it into the tools, get the first job on the list running against it, and train whoever has to live with it.
Review at 30 days
What is actually being used, what is not, and why. Then you decide what is next, if anything.
What you end up with.
Yours to keep, in your own systems, on accounts in your name. You maintain it, or we review it with you on a rhythm. Either way it is not ours.
- The business written down once, in a form every AI tool can read
- Your goals and the few numbers that say whether AI is helping
- Your knowledge gathered into one place, with the out-of-date material flagged
- How you sound, written down, so drafts come back in your voice
- Your main jobs mapped as they actually run
- Your AI rules in plain English, short enough that people read them
- The approved tools and models, with what may go into each one
- A named reviewer against anything that reaches a customer
- A named owner against every automated process
- The rollout order, with dates and names against it
- The accounts, logins and documentation in your business name
Businesses this suits.
Roughly five to two hundred people, Australian, and past the point of experimenting with AI one person at a time.
- Owners who have tried a few AI tools and want to stop guessing
- Teams where three people bought three different tools
- Practices and firms holding client data with professional obligations to meet
- Businesses with one or two automations working and no way of working around them
- Owners about to hire for admin and wondering whether they should
- Owners who want AI spending they can explain to a partner or a board
When we would tell you not to bother.
We do not do AI for the sake of AI. If the honest answer is a better process and a cheaper tool, that is what you will hear on the first call, and the first call is free.
- You want a document to show the board, with nothing changing afterwards
- You want sign-off that you are compliant. We are not lawyers and this is not legal advice
- You need a security certification or a penetration test. That is a different trade
- One obvious job is the whole problem. Skip this and just build that job
- Nobody can give us a couple of hours a week while we map the work
- You have already chosen the tool and want it justified
What changes, and what we will not claim.
We are not putting somebody else’s percentage on this page. Any time-saving figure we give you gets measured in your business, from a baseline we take before we start. If we cannot show you the working, we do not say it.
You stop paying for overlap
One list of approved tools with a reason against each. The duplicates get cancelled and the actual gap gets filled.
Decisions get made once
Which tool, which model, who reviews what, what data goes where. Decided, written down, and not re-argued every time somebody has an idea.
You can see whether it is working
Each job has numbers against it and a place to read them, so whether this is worth it stops being a feeling.
AI operating systems: frequently asked questions.
What is an AI operating system for a business?
It is the written layer that tells AI how your business works. One place that holds what the business does, who it sells to, what it is trying to achieve, how the work actually gets done, what language it uses, which tools and models are approved, what data may go into them, who checks the output, and which job gets done first. Every AI tool and agent in the business reads from it, so they all answer from the same facts instead of guessing. Some consultants call the same idea an AI operating model.
Is this software we buy from you?
No. There is nothing to install, no dashboard and no licence. It is written material that lives in your own systems, in whatever your team already uses for documents. The value is in the thinking and the writing, not in a product. If a job later needs something built, that is quoted as its own piece of work.
How is this different from an AI strategy?
Strategy is one part of it. A strategy says which jobs AI should do and in what order. An operating system also holds the context the tools need to do those jobs well: the business facts, the knowledge, the house language, the rules, who reviews what, and how it is measured. Strategy without the context tends to sit in a folder. Context without the strategy means everyone automates whatever they feel like.
Who owns it, and who keeps it up to date?
You own it. It is written in your business, in your language, in accounts and folders in your name. Most owners keep it themselves once it exists, because the hard part is writing the first version, not maintaining it. If you would rather we kept it current and reviewed it with you on a set rhythm, that is the Fully Managed option.
Do we legally need an AI policy in Australia?
There is no Australian AI Act. The National AI Plan, released on 2 December 2025, confirmed AI will be handled through existing laws and sector regulators rather than a standalone one, so the mandatory guardrails proposed in 2024 were not enacted. What already applies to you is the Privacy Act and the Australian Privacy Principles, plus whatever your industry has. The National AI Centre publishes voluntary guidance, Guidance for AI Adoption, built on six essential practices, and we write your rules against it. That is a sensible starting point and not a legal opinion. If your situation needs one, we will tell you to put it in front of your lawyer.
Can we put customer data into AI tools?
Not into the free consumer versions. The OAIC guidance on commercially available AI products says privacy obligations apply both to the personal information you put in and to what the tool produces, and as best practice it recommends not entering personal information, particularly sensitive information, into publicly available generative AI tools. So the operating system writes down which tools are approved, exactly what may go into each one, and what never can. That gets settled before anything real touches a system.
How long does it take to write?
Mapping how you work takes three to five days of our time, usually spread over a week or two around your availability. The written layer follows a few days after that. Then we set up the first job it supports and train whoever has to live with it. You get a date before anything starts.
What does it cost?
Consulting to get the right tools set up properly typically starts from a few thousand AUD. If a job inside the rollout needs a custom build, that is quoted separately before it starts. We quote in AUD, up front, with no surprise add-ons, and the first call is free.
We already use ChatGPT and Copilot. What would this add?
Most businesses in that position are using a fraction of what they pay for, because every person starts every chat from nothing. The tool has no idea what you sell, how you price, what you have promised a client, or what your team is not allowed to put into it. Writing that down once, in a form the tools can read, is the difference between a clever assistant and one that knows your business.
How small is too small for this?
If you are a one or two person business with one obvious job to automate, skip this and just build that job. It starts earning its place once more than a couple of people are using AI, because that is the point where everybody is deciding things on their own and the answers stop matching.
Where to go next.
AI for Australian Small Business
The broader picture: where AI is worth it for a small team, and where it is not.
Read moreCan I just build this myself?
The honest version of the build-versus-hire question, before you decide either way.
Read moreWhat is RAG?
How a tool reads from your own knowledge instead of making an answer up, in plain English.
Read moreLet’s work out
what AI should do here.
Book a free 30-minute discovery call. We’ll go through how the business runs, tell you which job is worth doing first, and whether it is worth doing at all.