TL;DR: When Plates Can Talk
Around 40% of food served in aged care goes uneaten, yet most facilities rely on guesswork to track what residents consume. This leads to undetected malnutrition, food waste, regulatory risk, and missed clinical cues.
Emerging AI tools—like AFINI?T—use overhead cameras and machine learning to analyse plates before and after meals, accurately tracking intake down to the nutrient level. These systems:
- Relieve staff of manual documentation
- Give dietitians real-time, pattern-based insights
- Help kitchens reduce waste and adjust menus based on actual consumption
- Support compliance with upcoming regulatory standards
- Offer residents more personalised, responsive care
Beyond the tech, this marks a deeper shift: from serving food to understanding nourishment. In a sector where small changes carry big consequences, AI-driven food tracking might quietly become one of aged care’s most transformative tools.
The Undernourished System
In the aged care sector, food is deceptively visible. It’s plated and served multiple times a day. It carries cultural expectation, operational cost, and clinical weight. It is one of the few remaining pleasures afforded to residents, one of the few daily rituals around which social life still clusters.
And yet, beneath this visibility lies a systemic blind spot: we don’t know what’s actually consumed.
We know what’s ordered. We know what’s plated. We know the budgeted cost-per-resident-per-day. But we rarely know, with any accuracy, what makes it from plate to body. That gap, between what is offered and what is absorbed, is where risk begins to accumulate.
Malnutrition is not just common in aged care; it’s often structural. Studies over the past decade have repeatedly found that up to 68% of aged care residents are at risk or already malnourished, even in facilities meeting all other care benchmarks. It’s not that food isn’t being served. It’s that food isn’t being eaten, or absorbed, or noticed when it isn’t.
Why? Because the system for monitoring intake is patchy, analog, and inconsistent. Staff rely on memory, estimation, and shorthand: “¾”, “ate well”, “refused”. These notes are usually made after the fact, sometimes hours later, and rarely calibrated across shifts. When staffing is tight, documentation suffers. When residents need texture-modified food or have cognitive decline, estimating intake becomes even more difficult, and even more crucial.
This isn’t just a clinical problem. It’s operational. Meals not eaten mean:
- Waste, both food and labour.
- Inaccurate forecasting, causing overproduction.
- Missed patterns, like someone who’s been declining in appetite for three weeks, unnoticed because no one had the data in one place.
- Avoidable costs, when a simple dietary adjustment could have prevented a costly hospital admission.
There’s also something deeper here, a dignity deficit. Imagine being served food you can’t chew or don’t enjoy, and having that silently cleared away without anyone asking why. Imagine having your preferences assumed or your hunger patterns ignored, simply because no one could see them.
Food is information. But in aged care, that information has remained stubbornly unavailable, until now.
Guesswork on the Menu
Spend a day in any residential aged care kitchen, and you’ll quickly see the complexity behind each tray. There are dietary codes, texture modifications, cultural preferences, allergy alerts, hydration goals. Meals must be safe, appealing, and nutritionally complete. But what happens after delivery is far more opaque.
Despite all this effort, once the plate is placed in front of the resident, the system goes blind.
Nursing staff or personal care workers may return to collect the tray and estimate what was consumed: “half eaten,” “refused meal,” “ate well.” These notes may be filed in a chart or logged in an app. They may or may not be passed on to the dietitian. They may or may not be accurate.
This is not a failure of care. It’s a failure of infrastructure.
The estimation process is vulnerable to:
- Subjectivity: What one carer calls “half eaten” another might call “mostly eaten.”
- Inconsistency: Shift changes, time pressures, and differing documentation cultures mean data can be patchy at best.
- Delay: Intake patterns often become visible only after a formal review, by which time weight loss or nutritional decline may already be entrenched.
Even well-trained staff, working under pressure, cannot track the nuances of resident consumption across time and context. Who’s skipping protein? Who’s stopped drinking fluids after 4pm? Who eats better when they sit near someone familiar?
None of this is easily knowable. But all of it matters.
For residents with dysphagia, diabetes, pressure wounds, or recovering from surgery, even minor nutritional gaps can have outsized clinical impact. But perhaps the most quietly devastating risk is cultural: when the system accepts not knowing as normal, it normalises invisibility.
To be malnourished in aged care is not to disappear overnight. It is to fade gradually, one unnoticed tray at a time.
Seeing the Invisible: AI as Nutritional Mirror
This is where a new class of tools, like AFINI-T, begin to shift the equation.
Developed in Canada by researchers at the University of Waterloo, AFINI-T (Automated Food Imaging and Nutrient Intake Tracking) is a camera-based, AI-powered system that measures actual food consumption with astonishing accuracy. A single mounted camera captures images of each plate before and after a meal. Machine learning algorithms then analyse the contents, calculating not just how much food was eaten, but which nutrients were consumed, in what volume, and how that compares to each resident’s nutritional needs.
Accuracy rates range from 89% to over 95%, depending on meal complexity. The AI recognises and categorises multiple food types on the same plate. It estimates volume and caloric intake. It tracks macronutrients. And critically, it does all of this without adding extra work for staff.
Other pilots are emerging globally:
- In Singapore, Alexandra Hospital trialled a similar system where trays are scanned to assess intake. Results were cross-checked against manual nursing logs and found to be more consistent and time-efficient.
- In the US and Europe, RGB-D imaging systems (which combine colour and depth data) are being explored to improve nutritional tracking in hospitals and aged care environments.
These aren’t theoretical concepts, they’re in active deployment. And they’re not designed to impress tech departments. They’re designed to solve a low-tech, high-impact problem: not knowing who is eating what.
Importantly, these tools do more than record, they reveal patterns.
Over time, AFINI-T and similar systems can:
- Alert when protein consumption drops below threshold over three days
- Flag residents who repeatedly reject certain textures
- Inform menu redesign based on collective uptake data
- Empower dietitians with trend lines instead of one-off notes
It’s not just surveillance. It’s feedback. It’s a nutritional mirror, one that shows the system what it’s missing and gives it a chance to adjust.
And because the data is continuous, not episodic, it shifts aged care from reactive to anticipatory. That alone marks a structural change in how care is imagined.
Why This Matters, Beyond the Meal
If food intake data were only about nutrient numbers, the benefits of AI tracking might feel limited to dietitians and auditors. But in practice, the implications stretch across every layer of the aged care ecosystem, from resident wellbeing to kitchen efficiency to systemic reform. What begins as a technological enhancement reveals itself as a structural upgrade.
1 For Residents: From Data to Dignity
Most aged care residents don’t just want to be fed. They want to be known. And one of the simplest, most persistent ways the system misses them is through misaligned meals. The wrong texture. The wrong flavour. The wrong portion. The wrong time of day.
AI doesn’t make these decisions. But it makes awareness possible.
With tools like AFINI-T:
- Residents who’ve stopped finishing their lunch can be gently asked why, before weight loss sets in.
- Someone consistently skipping meat can be offered enriched alternatives, not blamed for “refusing.”
- Fluid intake patterns can be tracked and adapted without relying on hourly manual tallies.
This isn’t just about reducing risk. It’s about restoring a sense of responsiveness.
Because when someone notices that you eat better with soup and bread than with dry chicken and potatoes, you’re no longer just a bed number or dietary code. You’re a person. One with preferences, rhythms, and needs that deserve attention.
Even in the quiet decline of late life, dignity can look like a bowl of the right food, served at the right time, in the right form.
2 For Providers and Caterers: From Insight to Efficiency
AI-enabled intake tracking offers food service operators what they’ve long lacked: a closed loop.
Traditionally, kitchen data stops at production, how many meals were prepared, for which diets, at what cost. What happens next is opaque. Were they eaten? Enjoyed? Wasted?
Now, providers can:
- Track which meals are consistently under-consumed and adjust menus accordingly.
- Refine portion sizes to align with actual consumption, not imagined need, reducing waste and saving cost.
- Streamline staffing, as manual intake logging becomes less necessary, freeing time for care tasks that truly require human judgment.
Waste reduction alone has powerful downstream effects. One Canadian facility using AFINI-T reported a 27% drop in plate waste within months of aligning portions to real resident patterns. That’s not just savings in food. It’s savings in labour, storage, procurement, and waste disposal.
And at audit time, compliance becomes clearer. Rather than presenting handwritten logs or anecdotal summaries, facilities can produce system-generated reports that demonstrate consistent monitoring, dietitian engagement, and responsive adjustment. This makes Standard 6 compliance not just easier, but meaningful.
For caterers, this is gold. It becomes possible to demonstrate performance, adapt rapidly, and partner with facilities in a way that goes beyond delivery. It’s not just about providing food, it’s about co-producing outcomes.
3 For Governments and System Designers: From Cost to Care Intelligence
On a policy level, AI-enabled intake data opens the door to more intelligent funding and regulation.
Currently, many governments rely on proxies to assess quality in food service, per-resident food spend, menu cycle reviews, or periodic assessments. But these tell you what was planned and budgeted, not what was actually delivered, or absorbed.
Imagine a system where:
- Funding bonuses or alerts were tied to actual nutrient intake trends, not reported spend.
- Food-related hospital admissions could be cross-referenced with AI-derived data to identify where early interventions might have prevented escalation.
- Research into optimal diets for ageing bodies could be powered by real-world data, not self-reported studies or idealised trials.
This isn’t about surveillance. It’s about building a feedback system that matches the complexity and variability of human need.
And it’s about future-proofing the aged care system for a demographic wave that will be larger, more diverse, and more digitally integrated than any before it.
The Bigger Pattern: When Food Becomes Infrastructure
Food is often discussed as a cost, a compliance item, a cultural amenity. Rarely is it framed as infrastructure, something that underpins systemic function and carries strategic consequence.
But AI-driven intake tracking invites us to reimagine aged care kitchens not just as sites of meal preparation, but as sensor-rich environments that quietly generate insights about wellbeing, decline, rhythm, risk, and responsiveness.
If we can measure what is eaten (and what is not), we can:
- Track appetite as an early indicator of illness, depression, or social withdrawal
- Adjust service models, meal timing, texture, and social settings, to suit actual human patterns
- Build closed-loop systems where feedback doesn’t rely on subjective recall or periodic audits but emerges daily, quietly, and precisely
This turns kitchens into something more than culinary units. They become data nodes in a larger network of care.
And when paired with other systems, fluid tracking, medication timing, mobility patterns, a more dynamic model of eldercare begins to emerge. One in which nourishment is not isolated from health, but deeply embedded in its maintenance and forecasting.
We talk a lot about smart homes. Smart beds. Smart alerts. But we rarely talk about smart meals. About kitchens that learn. About menus that adapt. About dining experiences that sense, not just serve.
This is not surveillance. It’s situational awareness for food, an area that has long operated in a fog of best guesses.
And like all good infrastructure, when it works, it becomes invisible. It just makes things smoother, safer, more human.
Resistance and Readiness
Of course, even elegant solutions encounter friction. Not every facility is ready for AI intake tracking. Not every kitchen team is comfortable with digital overlays. And not every leadership team sees food as a frontier for innovation.
There are real barriers:
- Upfront cost, especially for smaller providers or those in rural settings
- Cultural resistance among staff who worry tech will displace care
- Privacy concerns, particularly around image capture and resident autonomy
- Integration challenges, where new systems must dovetail with legacy software and workflows
But these barriers are not insurmountable. And they’re not unique to this space.
Ten years ago, many facilities resisted electronic medication charts. Twenty years ago, few imagined family members would check in via FaceTime or book care visits online. Progress in aged care often comes in waves, first resisted, then tolerated, then expected.
The key is to introduce technology that amplifies human insight, not replaces it.
When an AI system like AFINI-T quietly highlights that Mrs. Green has consumed less than 60% of her protein over three days, it’s not dictating care. It’s offering a nudge. A clue. A thread to follow.
When it helps a kitchen discover that fish dishes are routinely under-eaten, it’s not criticising the chef, it’s offering feedback, the kind that’s hard to gather otherwise.
Over time, these systems reduce rework. They reduce waste. They reduce escalation. But most importantly, they increase the system’s ability to notice early and adjust gracefully.
And in aged care, that grace, of timely attention, of tailored nourishment, of quiet knowing, may be one of the most powerful innovations we can offer.
Listening to the Plate
For decades, we’ve built aged care systems around what can be scheduled, standardised, and recorded. Meals have fallen somewhere between hospitality and hygiene, prepared with care, delivered on time, documented loosely, and often discarded quietly.
But what if we treated the plate not as a one-way delivery, but as a conversation starter?
Noticed what was returned untouched. Observed what was consumed eagerly, or reluctantly, or not at all. Tracked not just the nutrients, but the changes, the subtle drift in appetite, the growing aversion to texture, the quiet cue that something is shifting.
This is what AI can do when applied with care. Not dazzle. Not dominate. But listen, at a scale and consistency that humans alone can’t sustain.
It helps surface the things we’ve long suspected but never been able to prove: that certain residents eat better when food is softer, earlier, warmer; that portion sizes don’t need to be uniform; that intake can predict decline before weight loss appears on a chart.
It shifts our food systems from serving meals to tracking nourishment. From compliance to conversation. From waste to wisdom.
The tools now exist. The technology is real. The pilots are promising. What’s needed next is cultural readiness, a willingness to see food not just as a cost to manage, but as a data-rich, dignity-anchored medium of care.
Because in the end, the plate does talk. It always has. The question is whether we’re ready, finally, to listen.
If you found this article interesting, you might also enjoy reading CHRISY.com.au — a platform focused on aged care, health, disability, and rural equity, with grounded insights from inside the system.
For more pattern-driven writing on culture, systems, and diagnostics, you can subscribe to my Substack, Throughline, at kimhatton.substack.com.