Case studies
What we've built.
Every engagement delivered under a signed statement of work. Each case below gives the problem, what we built, and what came of it.
Aug – Nov 2025
Conversational Voice Ordering System
Independent Restaurant Client
Problem
A busy small-business restaurant was losing 15-20% of call-in orders during peak lunch hours. Calls went unanswered, wait times stretched while staff juggled the phone, and the people who should have been cooking were taking orders instead. The restaurant needed a phone system that could handle an order in natural conversation, price it correctly, and send it to the kitchen.
Solution
An AI voice agent that answers the phone, takes the order in natural conversation, confirms each item and modifier back to the customer, calculates the total with tax and fees, gives an accurate pickup time, and logs the order for kitchen staff. It runs every hour of every day, so no call rings out and nobody sits on hold.
Flow
Call answered → order taken in natural conversation → items and price confirmed back to customer → pickup ETA provided → order logged and routed to kitchen.
Outcome
We built the system and validated it end to end in staging. Voice quality, order accuracy, and operational fit all met the bar for launch. Rollout to production was deferred because of an integration constraint with an outside platform on the client's side.
Sep 2025
Multi-Location SKU Reconciliation
Multi-Location Convenience Client (3 retail locations)
0
SKUs audited
0.0%
of catalog flagged
0
price mismatches surfaced
Problem
Three retail locations ran the same point-of-sale system but maintained their product data separately. Prices, categories, and item descriptions had drifted apart over the years. Nobody had a consolidated view of how big the problem was or where the money was going.
Solution
We collected raw SKU exports from each location, cleaned and normalized the data, then mapped every SKU into a master reconciliation sheet. Alongside it we built an Excel dashboard for fast investigation: single-cell scan-code lookup, side-by-side comparison of all three locations, and automatic flagging across description, department, pricing, and price-group fields.
Findings
Of 11,608 unique SKUs, 4,737 flagged with at least one mismatch, or 40.8% of the catalog. That included 1,890 retail price mismatches, 96 of them with a single-SKU spread above $5. The mean spread was $1.24 and the median $0.40.
Outcome
The dashboard put every mismatch in front of the client in a form they could investigate quickly. They used it to standardize pricebooks across all three locations without combing through 11,608 items by hand.
Jun 2026 – In progress
Restaurant Intelligence
Internal product, deploying soon
What it is
A question-and-answer assistant for a restaurant operator. You ask it a plain question about your own business, in English or Spanish, and it answers from your own sales, spreadsheets, kitchen notes and local conditions, and shows you where each answer came from.
Problem
An operator's numbers sit in a point-of-sale system, a few spreadsheets and a stack of handwritten kitchen notes, and the local calendar lives in their head. Nothing puts those together or tells them where an answer came from. A generic chatbot is not the fix: it fills gaps with confident guesses, which is the last thing you want behind a staffing or ordering call.
What we built
An assistant that answers questions and can also work on your documents and spreadsheets, with an undo for what it changes. A document and spreadsheet editor with imports and exports. A sales dashboard. A calendar of local events with a read on how each one may affect traffic. Weather and live web search as extra sources. Kitchen notes and kitchen processes tracked in one place. Sign-in for each user.
Where it stands
As of October 2026 the product is deployed and the owner has walked it from start to finish many times. It is still in development and is not in use by a client. Before it can be handed over, it needs an agreed definition of done with the client, a rule for who can see and change what, and a review of the redesigned interface.
How it works
It never guesses. Every answer is built only from sources it was handed, and each answer points to them. If it cannot back an answer with a real source, it says so and declines instead of filling in the blank. That is what makes it safe to put behind a real operating decision.
Jul 2026 – In progress
Marketing Operations Platform
Deploying to two independent restaurant clients soon
What it is
One platform that runs a local business's whole marketing operation: its website, social posts, email and texts, review replies, online ordering, and the reporting that ties it all together. Each business switches on only the parts it needs.
Problem
A small operator usually pays for a handful of separate services to keep any marketing going, and still ends up with a few generic posts a month for the money. Everything lives in a different tool, the content is thin, and more often than not the business doesn't even own its own website when it's done.
Where it stands
The first business is being built out now: the content, the owner's approval step, posting to social and email, and online ordering. Catering, text marketing, and a reporting dashboard come next. Every piece is optional, so no two client setups look the same.
How it works
At the center is a content engine that drafts as many posts as each channel needs, far more than the services it replaces, and checks its own work so the posts don't all start to sound alike. Nothing publishes on its own. The owner sees the day's queue on their phone and approves, edits, or skips in a couple of minutes.