How One Restaurant Cut Food Waste by 33% in a Month with an AI Camera
Sydney Opera Bar installed a single AI-powered device above their kitchen bin. Within weeks, food waste per cover dropped by a third. A look at what happened and why it matters for your programme.
The starting point
Sydney Opera Bar is a single-venue restaurant and bar at the Sydney Opera House. It is not a chain. It does not have a data science team. It is a busy hospitality venue with a kitchen that, like most kitchens, was throwing away more food than it needed to.
The problem is common across the industry. According to the UN Environment Programme, the food service sector accounts for roughly 28% of all food waste globally. Most kitchens know they waste food. Very few know exactly what they waste, when they waste it, or why. Without that data, improvement is guesswork.
In January 2025, Sydney Opera Bar partnered with KITRO, a Swiss startup, and End Food Waste Australia to find out what was actually going in the bin and do something about it.
What they did
KITRO makes a device called TARE. It is a small unit that sits above a kitchen waste bin and contains a camera and a scale. Every time something is thrown away, the camera takes a photo and the scale records the weight. The image is uploaded to the cloud, where computer vision identifies what was discarded: carrot peelings, bread, leftover pasta, whatever it is.
The data feeds into a dashboard that shows the kitchen team exactly what is being wasted, how much it costs, and when it happens. KITRO's account managers then work with the kitchen to identify specific changes: adjusting portion sizes, refining prep workflows, rethinking buffet quantities.
The critical thing is that the device does not interrupt kitchen operations. The executive chef at Sydney Opera Bar described it as "set and forget." The camera triggers automatically. There is nothing for the team to scan, log, or remember to do during service. The data just appears.
The implementation was funded partly through a Bin Trim grant from the NSW EPA, which covers up to 50% of eligible equipment costs (up to A$50,000). That matters because it means the upfront cost was genuinely manageable for a single venue.
The results
Within the first month, food waste per cover dropped by 33%. Kitchen-generated waste fell as a proportion of total waste, and the team reported a "mindset shift" in how they thought about food preparation and portioning.
The executive chef put it simply: "I am just a chef, not a sustainability professional, so being able to collaborate and get help breaking down data is incredible."
KITRO's broader data across hundreds of kitchens in over 20 countries tells a consistent story:
| Metric | Range |
|---|---|
| Food waste reduction | 23–51% |
| Annual food cost savings | 2–8% |
| Typical ROI | 200–600% in year one |
| Payback period | Under 6 months |
An academic study published in the Waste Management journal in 2025 validated these figures across sites in Germany, Switzerland, and Greece, finding that AI-based waste tracking reduced cost of wasted food per meal by up to 39%. The study noted that the biggest reductions came from two areas: food preparation waste and overproduction.
For context, the Zurich Marriott Hotel, a larger deployment, saved 24 tonnes of edible food waste annually and reduced food costs by CHF 140,000 using the same system.
What makes this case interesting
It is genuinely small scale
This is not a chain-wide rollout or a multi-million pound technology investment. It is one device, in one kitchen, producing measurable results within weeks. The barrier to entry is about as low as it gets for an AI deployment.
The AI is invisible to the team
Nobody in the kitchen had to learn a new system, change their workflow during service, or become a data analyst. The device observes. The dashboard reports. The conversations about what to change happen after service, with data in hand. That is how adoption works in operational environments where people are busy.
The ROI is immediate and tangible
Food cost is one of the largest controllable expenses in hospitality. A 2–8% reduction in annual food costs, achieved within weeks and sustained over time, is not a marginal improvement. For a venue doing £500,000 in food purchases annually, that is £10,000–£40,000 back on the bottom line.
It proves the experiment-first approach
Sydney Opera Bar did not commission a feasibility study or build a business case for months. They installed a device, measured what happened, and made changes based on the data. The entire cycle from installation to measurable results was under four weeks. This is what the experimentation phase of an AI programme should look like.
Lessons for your programme
You do not need a big budget to start. KITRO's device costs a fraction of what most organisations spend on a single consultant. The Sydney Opera Bar case was partly grant-funded. If you are waiting for a large budget to begin experimenting with AI, you are waiting for the wrong thing. Section 7: Experimentation covers how to design and run low-cost experiments that build the evidence base for larger investment.
Measure before you optimise. The kitchen team did not know what they were wasting until the data showed them. Most operational improvements start with measurement, not action. If you cannot see the problem clearly, you cannot solve it efficiently. Section 3: AI Readiness includes a data foundations assessment that helps you understand what you can and cannot currently measure.
Make adoption effortless. The "set and forget" design of KITRO's device is not a feature. It is the reason the deployment succeeded. AI tools that require behaviour change during high-pressure operations fail. Tools that observe, analyse, and report without adding friction get used. Section 11: Change Management covers how to design for adoption rather than resistance.
Start with one site, one problem. Sydney Opera Bar proved the concept in a single kitchen. KITRO's broader client base shows the same approach scales to hotel chains and corporate caterers. But it started small. Your programme should too. Section 7: Experimentation has the framework for structured experiments that build confidence before you commit at scale.
Sources
- How Sydney Opera Bar significantly reduced food waste (KITRO, 2025)
- Reducing food waste in the HORECA sector using AI-based waste-tracking devices (Waste Management journal, 2025)
- Less food waste thanks to artificial intelligence (Innosuisse)
- KITRO Solution Overview (KITRO)