AI in the dining room by California Burrito Founder
We sell burritos. No consumer app, small data team, thousands of staff who spend the shift on their feet. Here's what AI has actually been useful to us.
California Burrito runs 140+ restaurants across six Indian cities. Our operation is stoves, trucks, cold chain and 2700+ people who don't work at desks.
We're not going to walk you through our roadmap. We're going to give you four categories of thing that used to be impossible at our size — expertise we couldn't hire, output we couldn't commission, software we couldn't justify, and data we couldn't read — with what we built in each, what it cost, and what failed.
Each one comes with a tell: how to spot the same opportunity in your own operation.
The four buckets
1. Expertise we couldn't hire
A food technologist to optimize our corn processing and to optimize our bulk cooking recipe economics. Never a full-time role at our size, never worth a consulting engagement for one question.
> The tell: a recurring problem nobody owns — because the person who would own it isn't on your payroll.
2. Output we couldn't commission
90+ blockbuster training films, 60–180 seconds, scripted and dubbed into Hindi for staff who don't read English marketing copy. We didn't replace an agency. The agency was never hired. This content simply did not exist.
> The tell: things that are absent, not slow. Write down what you'd make if making were free — then notice that list has never existed.
3. Software we couldn't justify
Our kitchen production screen. A boutique recruitment engine designed to our hiring process. A customized ops ticketing/maintenance and audit app. An engineering team was never going to be a line item here - but now may become one.
> The tell: processes running on WhatsApp groups, spreadsheets, and one person's memory. Every offline business is full of these and has stopped seeing them.
4. Data we couldn't read
Cameras have recorded our dine-in floor for years, to a disk nobody ever opened. Now: service speed, table hygiene, throughput. Delivery has been high-resolution for a decade because it instruments itself. Dine-in has been dark since the invention of restaurants.
Also here, and honestly: voice. We tested Storefox for service analysis and it didn't hold up — two people talking at a counter is data we're capturing and still can't read. Sarvam may crack the multi-speaker piece. Same bucket as the cameras, opposite outcome. It's where the hard edge of this category is.
> The tell: exhaust. Anything you already capture and never look at — CCTV, call recordings, compliance photos, handwritten forms.
And then: who actually does this?
With thanks to Vartika and Harsh from Elevation Capital, our internal hackathons. How we got people who aren't engineers building real tools, and what came out of them. This is the answer to "fine, but who in my company does any of this?"
Who is this relevant for?
Founders and operators in restaurants, retail, QSR, logistics, healthcare delivery and multi-site services
Ops and supply chain leaders
Product and engineering people building for non-desk workforces
Anyone weighing buy vs. build in a physical-world business