PrizmaCore
People ask me why I started PrizmaCore instead of just getting another job. Most of them want the polished answer. There is a polished answer. There is also a real one.
The polished version is on the About page. Twenty years across hospitality, trades, and RTOs. Wrote two nationally recognised courses. Coded HTML since 2006. AI hit a usable level and I went deep. All true. None of it really explains the part of the decision that actually mattered.
The shorter, more honest version is this.
I watched smart operators drown in admin their entire careers and nobody was solving it. Not really. Software companies were selling them more software that created more admin. When AI actually got good enough to change that, I couldn't not do something about it. PrizmaCore is what happens when someone who has actually done the work builds the tools instead of someone who just read about it.
That is the bar version. The rest of this is just the longer way of saying the same thing.
What I kept seeing
I spent two decades inside operationally messy businesses. Hospitality venues. Workshops. Mechanical and transmission work. A long stretch inside a hospitality RTO writing course material from scratch. Different industries, different uniforms, same pattern every single time.
Good operators, drowning. Owners working seventy hour weeks because the system was their brain. Office staff stitched together by sticky notes and verbal handoffs. Quotes going cold in inboxes. Job sheets lost between the workshop and the front desk. Customers slipping through the cracks while the team ran flat out trying to keep up.
None of these businesses were poorly run by careless people. They were run by good operators who had been told for years that the answer to operational chaos was buying more software. So they bought more software. Now they had a CRM, a quoting tool, a project management platform, a scheduling app, a separate inbox for support, and an accounting package. Each one promised to make life easier. The sum total was that everyone in the business spent two extra hours a day moving information between tools that did not talk to each other.
The pattern was always the same. The software industry was profiting from the chaos. The operators were paying for both the chaos and the half measures sold to fix it.
The moment it clicked
I had been watching language models since the GPT-3 days. Curious, but not convinced. Most of the early demos were party tricks. Write me a poem about my dog. Useful for nothing inside a real business.
The shift was when the models got good enough to read messy real world inputs and act on them with judgement. Customer emails written by humans with poor grammar and missing context. Quote PDFs that all look slightly different. Photos of a job site. Voicemail transcripts. The kind of inputs that broke every "automation" tool I had ever seen, because automation is dumb and humans are not.
When the models started handling that layer properly, the whole picture changed. The reason small business automation had always been so weak was because every input that mattered required interpretation, and old software could not interpret anything. That bottleneck broke and most operators did not notice, because the AI conversation in their world was still about chatbots and content generation.
That was the click. The same operators I had spent twenty years watching get crushed by admin could finally have tools that did the interpretation step for them. Not by replacing them. By doing the part of the job that should never have been their job in the first place.
I built a few tools to prove it to myself. Broke them. Rebuilt them. Showed them to operators I respected. Watched their faces when the thing did what they had spent ten years doing by hand. Decided.
Why boutique, and why Adelaide
Boutique because the agency model breaks for this work. Agencies need to scale, which means productising the offer, which means selling the same package to every business regardless of fit. Operational work does not run on that. Every business is built differently because every owner built it differently. The work that actually moves the needle has to be bespoke or it does not move the needle at all.
So. One operator. Full stack capability. No offshore handoffs. No account manager sitting between the work and the person paying for it. The whole engagement runs through one person who has actually done the trade, the shifts, and the diagnostic work. That is the model.
On Adelaide specifically.
It's home. The market here is full of genuinely good operators who are behind on this stuff through no fault of their own. Nobody was talking to them about it at a level that made sense. That felt like the right place to start.
The plan was national from day one and global where it fits. But if you cannot be useful to the people in your own city you cannot really be useful anywhere.
What this is actually building toward
The day one offer is workflow reviews, automation builds, AI consulting, custom software, and apps. Real and useful and what most clients need first. That layer is the work, and it stands on its own.
The longer game is sovereign Australian AI compute. Right now if you build a useful AI tool for an Australian business, the heavy lifting almost always happens inside someone else's data centre in another country, under that country's data handling rules. For a workshop running a customer database, that is fine. For a business handling sensitive medical, legal, financial, or compliance data, it is increasingly not fine. The regulatory ground under Australian data is shifting. The companies that will quietly own the next decade in this space are the ones positioning to run real AI infrastructure on Australian soil.
That is where PrizmaCore is going. Custom small language models. NVIDIA hardware sitting in Australia. Sensitive workloads handled inside Australian jurisdiction without compromising on capability. Boutique infrastructure for businesses that take their data seriously.
Everything we ship today is a step toward that.
So why not just get another job
Because I have done that and it does not solve the actual problem. The actual problem is that the gap between what AI can do for an Australian small business and what most Australian small businesses are actually using is huge, getting bigger, and almost nobody on the ground is closing it for them in plain language.
I would rather be the person closing that gap than someone watching it widen from inside another payroll.
If you want to see what closing it looks like in practice, the Workflow Review is the cheapest way; the longer version of all this lives on the about page.