I'm in Brisbane this week for ITAC. On Thursday I'm part of the closing fishbowl panel, Aged Care 2076. If you're at the conference, come say hi.

Speaking of the future — I came across something quite innovative. A startup that builds robotics for commercial kitchens has just published a paper on an AI model they developed, called Epicure. The model was trained on “recipe co-occurrence” — which ingredients chefs combine with which. It then identified the five basic tastes, clustered cuisines by culture (what makes a dish Thai or Lebanese), and deduced sweetness and saltiness entirely from recipe patterns.

The research is very dry and technical. I'd suggest skimming the introduction, then trying the Epicure tool yourself. I uploaded a photo of my fridge and asked it to give me a recipe with the ingredients that I had available and that is suitable for an older Italian person with high cholesterol cooking at home. You can see the full PDF output here. It’s not perfect (it has calculated a serving at 640 grams!), but it’s still very impressive.

The implication that interests me is for aged care kitchens, and for the many older Australians cooking at home. A model that understands flavour, texture and culture could help build menus across dozens of cultural backgrounds, and quickly design products with the texture profiles older people need.

This week I sat down with Elyssia Clark, General Manager of Marketing, Insights and Customer Experience at Benetas, to talk about what aged care looks like through a customer experience lens.

Elyssia came into aged care from SEEK and the consumer insights world. She joined Benetas in 2022 and built one of the few in-house insights functions in the sector.

Two threads from the conversation worth focusing on:

1 - The person making the decision is often not the person receiving the care. Elyssia describes a pattern she sees regularly — the eldest daughter doing the research, fact-checking, synthesising, then presenting the recommendation to other family members. Providers designing for "the customer" without recognising who is actually doing the research are designing for the wrong person.

2 - Families are increasingly using ChatGPT, Gemini and Claude to research providers. SEO is no longer the only game. Elyssia talks about GEO and AEO — generative engine optimisation and answer engine optimisation — making sure your website content is structured so that AI models cite it accurately when families ask.

👉 There's more in the full interview, including her answer to which single practice from outside aged care she'd embed across the sector tomorrow.

Tuesday 19 May | 10.30am AEST | Online | Free

A live Q&A with Elyssia Clark (Benetas), Justina Gardiner (LightSpeed Consulting), and Hanaa' Grave (culturalQ) on who's actually making aged care decisions in 2026, how cultural diversity shapes the consumer base, and where AI fits.

The aged care consumer in 2026 is more than one profile. Decisions move through families with different priorities and unequal dynamics. Cultural diversity is shifting too, and consumers are arriving more informed — 66% of families say online reviews shape their decision. The combination is what's new, and the pace it's changing at.

A new report from New Zealand's Koi Tū Centre for Informed Futures — People, Place and Prosperity — looks at the combined effect of an ageing population, a shrinking workforce, and growing regional pressure on smaller communities. The headline finding from the aged care side of the response is that older New Zealanders requiring dementia care are now being moved hundreds of kilometres from family — Dunedin to Nelson, in one example — because that's the closest available placement.

Heather Heap, a researcher at Aberystwyth University, writes about how humour shifts across the lifespan. The popular framing is that older adults lose their sense of humour. The research suggests the change is in how they use and enjoy it — what they find funny, what role humour plays in relationships, how it helps regulate emotion. Worth reading if you run lifestyle programs, or design social activities.

The Hangzhou Intermediate People's Court established a meaningful precedent late last month. A company tried to slash an employee's monthly salary from 25,000 to 15,000 yuan, citing AI-driven optimisation, then dismissed her when she refused. The court found the dismissal illegal. Replacing a worker because AI can do the job cheaper isn't an "objective major change" that justifies voiding a contract.

The court went further. Companies pursuing AI restructuring should prioritise retraining staff into roles requiring human intervention. China is signalling that the legal default should be retraining, not replacement.

I sat down with Tania Gomez on The AI Apprentice podcast to talk about what AI actually changes for older Australians ageing at home. The main shift I see is care moving from reactive to proactive.

Watch on YouTube | Listen on Spotify

Tania's work is worth following if you're in the NDIS or aged care space. She also runs an AI community of practice that's worth joining if you want to keep learning alongside others working through the same problems.

If you're using the paid version of Claude and you keep pasting the same context into every new chat — your organisation's tone, your template format, the policies you reference, you should use a Project instead.

A Claude Project is a workspace you set up once. You upload your style guide, your last few board papers, your communications templates, the documents that describe how you do things. Every conversation inside that Project starts with all of it already loaded.

The setup takes about ten minutes. Click + New Project, give it a name (e.g. "Board papers" or "Comms drafts"), drop in three or four reference documents, and write a short description of what the Project is for and what the output should sound like.

From then on, every time you open that Project, Claude is already calibrated. The first draft comes out closer to what you actually want, and you spend less time (and tokens) editing voice and structure. ChatGPT's equivalent is a Custom GPT; same idea, different name.

Each week, I review developments in ageing and aged care and what they mean in practice. If this was useful, forward it to someone in the sector who'd appreciate it.

George Gouzounis