Two Adelaide doctors built an AI that predicts hospital discharge readiness, saving millions and cutting readmissions.

A prospective trial at Lyell McEwin Hospital found median stay cut from 3.1 to 2.9 days, readmissions dropped from 7.1% to 5.0% resulting in projected savings of $9.5M/year and is now rolling out across SA hospitals.

South Australia is quietly becoming Australia's AI proving ground.

Two Adelaide doctors, a neurologist Dr Stephen Bacchi (32) and surgeon Dr Joshua Kovoor (28) built an AI system called The Adelaide Score that predicts when hospital patients are medically ready for discharge. Not a chatbot. Not a copilot. A narrow algorithm solving one specific operational bottleneck.

They ran a prospective trial at Lyell McEwin Hospital in Adelaide's northern suburbs. The results, published in the ANZ Journal of Surgery, are impressive:

While the global conversation fixates on frontier models, billion-dollar compute clusters, and artificial general intelligence, the AI that's actually delivering measurable ROI is narrow, specific, and addressed something tangible. It doesn't have to generate text or images. It just has to make one decision better.

Every business has a "discharge bottleneck”. A decision point where delays compound and costs accumulate. It might be project approvals, quality checks, inventory replenishment, or scheduling. The Adelaide Score didn't try to reinvent hospital operations. It found one high-cost decision and made it faster. That's where the ROI lives, not always in platforms, but in removing bottlenecks.

ANZ Journal of Surgery — Kovoor et al. | Healthcare IT News | University of Adelaide

Think of an AI’s context window like a whiteboard in a meeting room.
Everything written on that whiteboard is what the AI can “see” at any one time. Your instructions, the conversation history, any documents you’ve pasted in, and its own previous responses. The moment something gets erased to make room, the AI forgets it completely. Not archived, just gone.

The size of that whiteboard varies. Older or cheaper models might give you a few thousand words of space. Modern tools like Claude or GPT 5.4 can hold the equivalent of a short novel, but it’s still finite.

Front-load your context. Put the most important instructions and background at the start of your prompt, before you ask your question. AI reads it all, but weighs early context heavily.

Paste in your documents. Got a policy, a contract, or a report you want the AI to work from? Paste it in. The AI will reason over the actual content rather than guessing from general knowledge.

Keep long chats fresh. In extended chat sessions, periodically paste in a quick summary of what’s been decided so far or get AI to summarise where you’re at. Don’t assume the AI “remembers” something from 40 messages ago, it may have been pushed out.

Give it a role and rules upfront. “You are helping me write for SA small business owners. Keep it plain English. No jargon.

Don’t stuff it with irrelevant content. Pasting in a 40-page document when you only need page 3 wastes window space and can actually dilute the quality of the response.

Don’t bury your real question at the end of a wall of text. If your actual ask is a single sentence hiding in paragraph seven, the AI may not weight it properly.

The AI is only as smart as what you put on the whiteboard, so be deliberate about what goes on it.

Next week we will go a bit deeper. Should you spend $80k on a junior analyst, $120k - $150k on a senior analyst when you have AI for $30 per month?

Keep and eye out for next weeks issue to see the debate

A oldie but a goodie. Fleet Space is well known as a South Australian success story and they held hackathons.

For those who are unaware, it’s common for a hackathon to have a problem(s) in mind, teams come together for an all out rapid thinking, planning and executing event over 8 hours or to a few days. The outcome is an unconstrained, and low distraction number of solutions to problems.

Why Fleet Space make this weeks story cut is because they are both a unique pinnacle of success, but also a business just like you.

Hold yourself a 1 day session, have a good facilitator, and be ready to smash out ideas. All of which can be pushed through AI in many ways.

Oh, and what did Fleet Space achieve? Multiple dashboards, reports and creative solutions to their operational challenges. Things that would have taken their teams and analysts many weeks or months to achieve in their day jobs.

Take the day, your team will surprised you.

Two Adelaide doctors found a bottleneck and saved a hospital $9.5 million. Your bottleneck probably isn't hospital discharges. It might be quoting, rostering, compliance checks, or approvals, but the playbook is the same:

Share with us your challenges or wins to be featured in our upcoming letters