What Happens When Restaurant Operators Spend a Day Building With AI
Eight restaurant operators spent a day building with AI at Marqii's MCP workshop in NYC. See what they built and the prompts you can try yourself
Every operator I know has a wishlist. It's the report they wish ran itself, the dashboard that pulls everything into one place, the question they wish they could just ask their data. Until very recently, that wishlist has lived on someone else's roadmap.
On October 21, I watched eight of our restaurant partners start crossing things off their lists. They came in from all over the tri-state area to New York for a full day of hands-on building with the Marqii MCP. There was no sitting through slides. Laptops were open, sleeves were rolled up, and everyone was building.
I want to share what I saw, what our partners built, and the prompts you can use to try it yourself. If you've been curious about AI for restaurant operators but didn't know where to start, this is a good place.
What's the Marqii MCP?
Quick primer, because I get this question a lot. MCP (Model Context Protocol) connects AI assistants like Claude or ChatGPT to the tools you already use. The Marqii MCP connects your AI to your Marqii data: reviews, listings, hours, menus, search appearances, social posts, and website widgets.
In practice, that means you can ask a question about your business in plain English and get an answer built from your real data. And with your approval, it can make the update for you, too.
Two things stuck with me
First, you can build almost anything right now. Ideas that used to sit on a wishlist got built in a single day, inside an AI assistant, by people who run restaurants, not engineering teams. That's the new normal.
Second, your tools don't have to live in silos anymore. The room got loud when partners started pulling their Marqii data into the same conversation as their other systems (POS exports, labor schedules, marketing calendars) and asking questions across all of it.
That's the shift. This isn't about another platform to log into. It's about the tools you already use finally talking to each other.
Greg LaFauci on getting your website found by AI
One of my favorite sessions of the day didn't come from our team. It came from a fellow operator. Greg LaFauci of Counter Service walked the room through how to index your website for AI search with an llms.txt page.
Here's why it matters. More guests are asking AI assistants where to eat, not just typing into a search bar. An llms.txt file is a simple, plain-text page at the root of your site (yourrestaurant.com/llms.txt) that gives AI models a clean summary of who you are: your locations, hours, menu links, what you're known for, and where to book or order.
The way Greg framed it stuck with me. Your website isn't just a brochure anymore. It's a source AI pulls from when a guest asks, "Where should we get dinner tonight?"
What I took away from Greg's session:
- Keep it simple. Plain text and clear headings. Write it like you're briefing a new host on day one.
- Point to your best pages. Link to each location page, your menus, and your ordering and reservation links.
- Keep it accurate. If your hours or menu change, your llms.txt should too. The same "one source of truth" rule that applies to your listings applies here.
It's a small lift with real AEO upside, and a few partners had a first version drafted before lunch.
What our partners built: the prompts
We split the afternoon into three hands-on blocks: asking questions, making updates, and building something new. Partners started from a set of prompt cards, then made them their own. Here are a few of my favorites you can try with the Marqii MCP.
Ask your business anything
The goal for this one was simple: walk away with one finding about your business you didn't know before.
- Find your weak spot: "Using my Marqii data, which of my locations had the lowest average rating in the last 30 days? For that location, what are the common themes in negative (3★ or less) reviews?" Then follow up: "Show me the actual reviews behind those themes. Is this new, or has it been building over the last 60 days?"
- Found, but not chosen: "Using Marqii, compare search appearances, customer actions (calls, directions, website clicks), and review rating by location for the last 3 months. Which locations get found but don't convert? What do their reviews say that might explain it? Give me a one-page summary I could send to my ops team."
Doing: real updates, no dashboard login
We had one rule for this session: end every prompt with "Show me the change first and wait for my approval before updating." These are live updates to Google and other sites, so you stay in the driver's seat.
- Holiday hours: "Using Marqii, set [location] to closed on Thanksgiving, Thursday, November 26. Show me the change first and wait for my approval before updating."
- Hours audit: "I've attached a spreadsheet with our hours for every location. Using Marqii, compare it to the hours currently listed and show me a table of what needs to be updated. Wait for my approval before updating anything."
- Photos: "Using Marqii, find which of my locations have the fewest photos. Help me write a description and alt text for each new image. Show me the plan before uploading."
Building: content, websites, and dashboards
I gave everyone an hour to build. Partners picked a track and went deep.
- Content: "Using Marqii, find the dish guests rave about most in reviews at [location]. Write a Google post about it and save it as a draft."
- Website: "Using Marqii, create a reviews widget showing our best Google reviews and a menu widget for [location], both styled to match our brand. Give me the embed code for each."
- Data: "I've attached our sales export from [POS]. Using Marqii, combine it with review ratings and search data by location for the same period. Did locations with dropping ratings also drop in sales? Build a dashboard and a short summary I can share with leadership."
- Staffing vs. reviews: "I've attached our staffing levels by location and day. Using Marqii, compare them to reviews mentioning slow service or wait times. Are complaints higher on understaffed days? Show me the pattern in a chart."
That last kind of question used to take an analyst and a week. In the room, it took a prompt and a few minutes. Watching that happen never gets old.
Tips from the room
A few things I'd pass along if you're trying this yourself:
- Start small. Try one location before rolling a change out everywhere.
- AI not pulling your data? Add "using Marqii" to your prompt.
- Getting "no data"? Widen the time range.
- Too much output? Ask for a set number of bullets.
- Posts save as drafts, so nothing publishes until you say so.
A first look at Marq, our new in-app agent
We also gave the room a first look at Marq, our new in-app agent, which is now live. Just ask, and it can update your hours or menus, help you draft social posts, and surface insights from your Marqii data, all without leaving the platform.
Where we're headed next
Thank you to every partner who gave us a full day, and to the Marqii team who made it happen. We're taking this show on the road next, so let me know if you're ready for us to come to your city.
Marqii restaurant partners: if you want in on the next one, let us know or reach out to your account manager.
Not working with Marqii yet? Schedule a demo to see how Marqii helps your restaurant get discovered and get chosen.
More resources:
Introducing Marqii MCP and How to Set It Up.
10 Things You Can Do in Marqii Without Leaving Your AI Chat
How Restaurants Are Using The Marqii MCP To Work Smarter
Why We Built the Marqii MCP: Giving Operators Back Control of Their Data