A delivery restaurant doing 60+ orders a day usually pays for three roles it barely notices: someone assigning couriers, someone babysitting the back office, and someone who is supposed to run marketing but mostly reposts photos. That's $6,000–10,000 a month in salaries for work that mostly follows rules, and rules are what software handles well. Our three assistants, Delly, Backy, and Marcy, take on exactly that kind of work. What follows is what each one does through the day, and where the savings actually come from.
Delly assigns couriers better than a tired dispatcher at 8 p.m.
Courier assignment is a math problem that most people solve by habit. A dispatcher glances at the screen, picks whoever looks free, and moves on. Delly treats it as a real optimization problem. For every new order it weighs each courier's current position, the orders already in their bag, when the kitchen will have this one ready, and the delivery zone, then assigns the order so total lateness across everything in progress comes out lowest. A human picks whoever's closest. Delly picks whoever keeps the whole evening moving.
After the hand-off it keeps watching. When an order sits ready on the pass longer than it should, Delly flags it. When a courier's ETA slips past what you promised, it can reassign before the customer notices anything. This is the tedious side of logistics management, the part a person does unevenly and software does on every order.
The money side: a dispatcher costs $2,500–3,500 a month, and sloppy assignment adds 5–10 minutes to your average delivery time without anyone clocking it. Operators on our platform save $3,000+ per month on dispatch alone. That's the salary, plus fewer refunds for late orders, plus one to two extra deliveries per courier per shift once routes stop overlapping.
Backy does the back-office rounds nobody enjoys
Every restaurant has a list of small checks that keep things honest: is something on the menu sold out but still orderable, did an order get stuck between "accepted" and "cooking", are today's numbers drifting from a normal Tuesday, did a branch stop taking orders without telling anyone. A manager runs these rounds when there's time, and there's rarely any.
Backy runs them around the clock. It watches the order pipeline and pings the right person the moment something stalls. A branch whose acceptance rate suddenly drops gets caught. The daily summary a manager would otherwise stitch together from three screens arrives already done. Same process automation logic we use everywhere: define the normal state, spot the deviation, tell whoever can fix it, and hand them the context along with the alert.
Customers running Backy tell us the back-office work gets done with 2–3× fewer manager-hours, which for most operations works out to another $3,000+ per month: half in payroll, half in problems caught at minute 2 instead of hour 2. A stuck order you find late costs you a refund and a regular. Found early, it's a two-minute fix.
We'll show Delly, Backy, and Marcy working live — with your volumes plugged in.
Marcy runs the campaigns your team never gets around to
Where Delly and Backy save money, Marcy brings it in. The mechanism is segmentation and timing rather than clever copy. Marcy splits your customer base by recency, frequency, and spend, the classic RFM cut, then runs the plays any decent marketer would run given the hours: a nudge to customers who haven't ordered in 21 days, a cashback bump for big spenders who are drifting off, a push at 5:40 p.m. to people who usually order dinner at 6. Every campaign goes out through the marketing tools already wired into your app and site, so you measure results in orders, not opens.
The math on a modest base: 4,000 customers, 15% of them asleep (no order in 30+ days). A win-back push that wakes just 5% of that segment is 30 orders. At a $25 average check, one campaign brings in $750, and Marcy runs a campaign every week, on schedule, without anyone asking. Across our customers, Marcy generates $5,000+ per month in orders that wouldn't have happened otherwise. This isn't attribution magic. These are orders from customers who had gone quiet and came back through a tracked campaign link.
What the assistants don't do
The assistants won't invent your strategy, write your brand voice, or fix a bad menu or a slow kitchen. Delly can't make a courier ride faster; all it can do is stop wasting the couriers you already have. Marcy's campaigns only work because the ordering channel underneath them converts, so a push pointing at a clunky checkout earns you clicks and no orders. And below roughly 20–30 orders a day none of this helps much: at that volume you don't have a dispatch problem or a segmentation problem, you have a demand problem, and you fix that with marketing basics rather than automation.
They're also not chatbots, and no customer ever talks to Delly. They work on the operations side, on assignment, monitoring, and campaigns, where the inputs are structured data pulled from your own monitoring panel and where being right 99% of the time adds up instead of embarrassing you.
The number to check before you buy any restaurant AI
One test cuts through every AI pitch, ours included: ask what specific task the system takes off a specific person, and what that person costs you today. "Insights" is not a task. "Assigns every courier and reassigns when an ETA slips" is a task. Across the 3M+ orders processed on Dots, the pattern holds. The assistants pay for themselves when they take over hours someone is already paid for (Delly, Backy) or run the plays nobody had hours for (Marcy). Count your dispatcher-hours and your unsent campaigns before you count anything else. That's your ROI baseline, and you can work it out this afternoon.
Apps, ordering site, logistics, and the AI that runs them — built around your operation.