The Real Problem
Yuki runs a sushi train in Newmarket. Every night at closing, she watches $150 to $200 worth of unsold sushi go into the bin. Salmon nigiri that nobody ordered in the last hour. Rainbow rolls that circled the conveyor until the rice dried out. Prawn tempura rolls that looked perfect at 6pm and looked tired by 8:30pm.
Sushi restaurants have a waste problem that most other hospitality businesses don't face. You prepare food in advance -- not when a customer orders it, but before they walk in the door. A busy sushi train might have 200+ pieces circulating at any given time, each with a strict freshness window. Once a plate has been on the belt for 45 to 90 minutes, it gets pulled and thrown away. There's no reheating it. There's no serving it tomorrow.
Industry estimates put sushi restaurant food waste at 15-20% of prepared food daily. For Yuki, that's roughly NZD 4,000-5,000 a month going straight in the bin.
The real pain is that the waste is unpredictable. On a rainy Tuesday, Yuki over-prepares and bins 60 plates. On a sunny Friday, she under-prepares and runs out of salmon nigiri by 1pm -- losing sales to hungry customers who came specifically for it. She can feel the pattern, but she can't predict it reliably enough to prep the right quantities at the right times.
And the ingredients are expensive. NZ sushi restaurants import specific fish grades, nori, and specialty rice. Every wasted plate isn't just lost food -- it's lost margin on some of the most expensive ingredients in hospitality.
Why Existing Tools Don't Solve This
Your POS system (Lightspeed, Square, Eats365) tracks what you sold. It tells you that you sold 120 salmon nigiri last Tuesday. But it doesn't tell you how many you should prepare next Tuesday at 11am versus 2pm versus 6pm.
Generic hospitality platforms like MarketMan and WISK handle inventory and food costing for standard restaurants. But sushi trains have a unique dynamic: food is prepared in batches throughout the day, placed on a conveyor with a time limit, and removed when it expires. No off-the-shelf tool models this prepare-circulate-expire cycle.
There's no NZ-specific tool that does demand forecasting for sushi. So most sushi restaurant owners rely on instinct and experience -- which works most of the time, but costs thousands a month when it's wrong.
How AI Solves This
You share your POS sales data with the AI -- typically six months of transaction history. The AI analyses what sold, when it sold, how much was prepared, and cross-references that with external factors: day of the week, weather, school holidays, nearby events, public holidays.
Every morning, you get a prep schedule:
Prep Schedule -- Tuesday 8 April 2026
Weather: Overcast, 16C. Expected: moderate foot traffic (Tuesday average -12% due to weather).
10:00am - Opening batch:
- Salmon nigiri: 30 pieces
- Tuna nigiri: 18 pieces
- Prawn: 15 pieces
- Teriyaki chicken: 22 pieces
- Vegetarian rolls: 12 pieces
11:30am - Pre-lunch batch:
- Salmon nigiri: +25 pieces
- Tuna nigiri: +15 pieces
- California roll: +20 pieces
1:30pm - Post-lunch taper:
- Reduce all prep by 40% from lunch quantities
- Focus on: salmon nigiri, chicken katsu (afternoon bestsellers)
5:00pm - Dinner prep:
- Full evening batch -- see detailed schedule
The AI doesn't just tell you what to make for the day -- it tells you what to make each hour. This is critical for sushi, because freshness windows are short. Preparing 100 salmon nigiri at 10am for the whole day guarantees waste. Preparing 30 at 10am, 25 at 11:30am, and adjusting from there keeps everything fresh and reduces what ends up in the bin.
Learning From Results
After each day, the AI compares its predictions to actual sales:
End of Day Report -- Tuesday 8 April
Total waste: 14 plates (was averaging 45 plates on Tuesdays) Stockout events: 1 (ran low on prawn at 12:45pm -- adjusting tomorrow) Prediction accuracy: 88%
Adjustment: Increasing prawn allocation by 8 pieces for rainy weekday lunches.
Over time, the model gets more accurate. It learns that your Newmarket location gets a lunch rush from the nearby office buildings on weekdays but a slower, longer lunch on weekends. It learns that salmon nigiri outsells everything else 2:1 on Fridays. It learns that the first week of school holidays drops weekday lunch traffic by 20%.
How We Set This Up
None of this works if the AI is just a standalone chatbot with no connection to your actual business. That's why BestAI builds a custom integration program -- a piece of software that bridges your AI assistant with the systems you already use.
For this kind of setup, that means:
- Connecting to your POS system to pull historical sales data and daily transaction records
- Setting up a weather and events data feed so the AI can factor in external conditions
- Building the hourly prep schedule generator tailored to your menu and kitchen workflow
- Configuring daily reports delivered via WhatsApp or email each morning and evening
Here's our process:
- We map your current workflow -- We sit down with you and understand how your kitchen preps, what your conveyor cycle looks like, and how you currently decide quantities.
- We build the connections -- Our developers write a custom program (an API connector) that lets the AI pull data from your POS and external sources. No manual data entry, no spreadsheets.
- We test end-to-end -- We run the AI's predictions alongside your normal prep for two weeks, comparing its recommendations to what actually sells. Nothing goes live until the accuracy is proven.
- We maintain it -- When you change your menu, add seasonal items, or open longer hours, we update the model to match.
You don't need to be technical. We handle all the development -- you just tell us how your kitchen runs, and we make the AI fit into that.
The Result
- 30-50% less food waste -- prepare what you'll sell, not what you hope to sell
- NZD 1,500-3,000/month saved -- less wasted premium ingredients going in the bin
- Fewer stockouts -- AI flags when you're about to run low on popular items mid-service
- Hourly precision -- prep schedules broken down by time of day, not just daily totals
- Continuous improvement -- the model gets more accurate every week as it learns your patterns
What AI Can't Do Here
- AI can't control your suppliers -- if your fish delivery is late, you still need to adjust manually
- AI can't account for kitchen mistakes -- over-seasoned rice or poorly cut fish still gets wasted
- AI won't eliminate all waste -- some waste is unavoidable in sushi (trimmings, end cuts, presentation standards)
- Predictions need 3-6 months of POS data to become reliable -- it's not instant
- AI can't handle one-off events it's never seen before -- a surprise bus tour of 40 people will still catch you off guard
Who This Is For
- Sushi train and conveyor belt restaurants where unsold food expires on the belt
- Sushi bars and takeaway shops that pre-prepare display cases each morning
- Any sushi restaurant losing NZD 100+ per day to unsold food waste
- Owners who want data-driven prep schedules instead of guessing quantities each morning