Start with one thing. Prove it. Then expand.
Most businesses do not have an AI problem. They have a starting problem: too many options, too many people paid to say yes, and no way to know what worked. Our method is built to remove that, one step at a time, with the criteria written down before the work starts.
The steps
Four steps, each one earning the next. You are never committed to a step until the previous one has shown its result.
- Step 1 · 15 to 30 minutes · no charge
A conversation
You tell us the one thing you would like AI to do for your business, or the cost you suspect is too high, or the decision you cannot get comfortable with. We ask the questions an operator asks. No deck, no discovery form, no pitch.
What you hold at the endA clear statement of the one thing, in your words, that we both agree is the right place to start.
- Step 2 · Within three business days · no charge
Written scoping
We write down what we would do, what it costs, what it should save or earn, how we would measure it, and what would make us tell you to stop. You read it in your own time. There is no charge for this and no obligation attached to it.
What you hold at the endA one to two page scope with a fixed fee, a measure and a kill criterion. You can take it to anyone.
- Step 3 · Two to six weeks · A$2,500 to A$7,500 fixed
A paid pilot
Small, real and measured. The pilot runs on your actual data and your actual process, with the success criteria, the budget cap and the kill criteria locked before the first hour is billed. If it does not move the number we agreed, you hear that plainly, and you have lost a small fixed fee rather than a year.
What you hold at the endA working result in your business, the measurement that proves it, and a written recommendation: expand, adjust, or stop.
- Step 4 · Priced per step · pilot fee credited
Implement and expand
What worked gets built to last: documented, trained into the team, monitored, and connected to the systems you already run. Then the next thing, chosen from evidence rather than a roadmap somebody sold you. Each step is priced on its own, so you are never committed to the step after it.
What you hold at the endAn operation that improves each month, owned by you, with a named person responsible for keeping it working.
In writing, before we start
Three sentences that hold on every engagement, written into the scope before anything is billed. They are the method as promises, and they are the reason the method holds.
The measure, the budget cap and the point at which we stop are set in the written scope before anything is billed. Nothing moves the goalposts afterwards.
Everything we build runs in your accounts and is yours: code, configuration, documentation and measurements. We hand over; there is no dependency on us.
If the pilot does not hit the measure we agreed first, we keep working at no further cost until it does, or we stop and say so in writing. You will not be sold the next step by the first one.
The rules
Three rules we do not bend, because each one exists to protect you from a way this usually goes wrong.
Criteria before work
Success criteria, budget cap and kill criteria are agreed and written before anything is billed. Goalposts that cannot move are the only defence against a project being called a success afterwards regardless of what happened.
Measured, not admired
A pilot is judged on the number we agreed, in your systems, not on a demonstration. We report the number even when it is against us. A tool that is not working gets switched off and said so.
Stop is a real answer
Stop is a real answer, and it is written into the scope as one. When the measure says stop, you hear it plainly and you have spent a small fixed fee rather than a year. We are paid for the right call, not for the build.
What the paperwork looks like
What a client receives at each step, in writing.
- A written scope: the fee, the measure and the kill criterion on one page
- A signed pre commit sheet before any pilot is billed
- A pilot measurement page, before and after, in your own numbers
- A monthly operations report you can put in front of a board
Fees and terms
Conversations and written scoping are at no charge. Pilots run from A$2,500 to A$7,500 fixed depending on the workflow and the systems involved; most first pilots sit in the lower half. Reviews from A$1,800, training from A$1,400 a half day, the decision verdict A$750, the content engine from A$3,800 to set up and A$680 a month. Every fee is quoted in writing before work starts and does not move. Ongoing operations are a monthly retainer from A$900, itemised, that you can end with a month’s notice.
The pilot fee is credited in full against implementation if you proceed within ninety days. We publish these figures because every credible small firm in this market does, and because a buyer who has to ask the price has already been told something about the seller.
Your data stays in your systems and in Australian or agreed jurisdiction infrastructure; personal information never crosses into an AI model without being pseudonymised first, and we work within your Privacy Act 1988 obligations.
Everything we build, you own: the code, the configuration, the documentation and the measurements. If you decide to take it elsewhere, it goes with you.
Questions
In four parts: getting started, pricing, how it works, and the objections we hear most.
We cannot name the one thing. Where do we start?
With the AI Opportunity Audit, about fifteen minutes, which names three. Or with a conversation, where we ask the questions an operator asks and usually find it in the first ten minutes.
How much of our time does a pilot take?
An hour to scope, an hour a week during the pilot from the person who owns the process, and a training session at the end. We do the rest.
Why publish prices?
Because a buyer who has to ask has already been told something about the seller. The band is real: most first pilots sit in the lower half, and the quote is fixed before work starts.
What is credited against what?
The pilot fee, in full, against implementation of the same workflow within ninety days. Reviews are credited against any implementation in the same period.
Who does the work?
ProjxAI, with the systems it runs for itself. A person is accountable for every scope, every pilot and every verdict, and that person is named on the scope.
What do we own at the end?
Everything: code, configuration, documentation, measurements, accounts. If you take it elsewhere, it goes with you and keeps running.
Where does our data go?
It stays in your systems. Anything that crosses to an AI model is pseudonymised first; personal information never crosses. We work within your Privacy Act 1988 obligations and will say plainly where a tool’s terms do not meet them.
Our IT provider says they can do this.
Often they can, and we are happy to scope it for them. Where an AI step has to read or draft, or where a measure has to be agreed and reported, that is usually the part we build.
We tried AI and it did not stick.
Almost always because nobody owned the process, nobody trained the people, and nobody measured it. Those three things are the pilot.
Is this just ChatGPT with a consultant?
The model is the smallest part. The scope, the measure, the human checks, the training, the record and the person accountable are the parts that make it work, and they are what you are paying for.
Bring us the one thing you would like AI to do.
Say it in your own words. We come back with it written down: what we would do, what it costs and what it should save. You decide from there.
- 1A conversationFree · 15 to 30 minutes
- 2Your one thing, in writingFree · what, cost and saving
- 3You decideNo obligation
I would like AI to answer after hours calls and book the callback.