‍

Part of Bikky's AI Summer School — a series where restaurant operators share how they're using AI day-to-day, what tools they're using, and what it's done for their brand.

‍

In this AI Summer School session, Bikky CEO Abhinav Kapur sat down with Phil Smith, Director of Marketing at Upstream Hospitality, who shared how he used Claude to plan and run the marketing for a new store opening in a new market.

Defining the success of new store openings

‍

New store openings at Upstream are measured by specific numbers: orders on day one, how many of those guests come back within 30 or 60 days, Google reviews in that same window, and the size of an email list built to warm up the area before the doors even open. It's an approach that's helped Upstream scale to roughly 20 locations, though two parts of the process have always taken significant time to prepare.

‍

The first is getting the opening plan itself organized. Details tend to live scattered across notebooks, meeting notes, and lessons from past openings. On top of that, the timeline shifts that hit almost every new store opening ultimately pull the plan out of shape.

‍

The second is building brand awareness fast enough to hit their day-one, 30-day, and email list targets, which are especially if a store is opening in a new market.

Phil saw an opportunity to use AI  to help with both problems, one for organizing the plan and one for running the paid marketing to warm up a new market before opening day. He shared both during his AI Summer School session.

Phil’s new store opening planning playbook

Here’s exactly how Phil created a step-by-step launch plan for new store openings with AI, built from his  notes and meetings.

‍

Step 1: Brain dump. Feed AI everything scattered across your notes and meetings, unedited: target open date, goals modeled on a past opening, budget concerns, and worries specific to the new market, like limited brand recognition or heavier local competition. It doesn't need to be organized, it just needs to be comprehensive.

‍

Step 2: Have the model interview you first. Instead of asking for a finished plan right away, instruct the model to ask the most important questions it needs answered, one at a time, and to respond like an operator would. Answer by talking out loud rather than trying to write a polished response, since a model doesn't need clean sentences, just a lot of real information.

‍

Step 3: Let AI build the plan. From there, have it build a full launch plan, working backwards from opening day in a format that can survive a date slip, since openings rarely land on the date first set.

‍

Step 4: Push the plan into your actual tools. Connect AI to your project management software so it can break the plan into individual tasks, assign owners based on what it knows about each person's role, and phase everything into pre-launch, launch week, and post-launch, .saving significant time by skipping the manual entry entirely.

Phil’s ad campaign playbook

Here's how Phil built a paid ad campaign to grow an email and phone list ahead of a new store's opening day, using his own guest data.

‍

Step 1: Identify your best existing guests. Mine your first-party ordering and loyalty data for guests who order frequently, for example, at least twice in the past six months.

‍

Step 2: Build a lookalike audience. Hand that group's real behavior to your ad platform, such as Meta, which uses it to find people across the country who resemble them. That becomes the target audience for the new market.

‍

Step 3: Connect your CRM. Connect AI directly to your CRM like and have the model build a live segment of high-frequency, high-value guests most likely to respond in the new market. Sync that segment straight into whatever ad platform you use.

‍

Step 4: Let AI build the campaign. Connected to your ad platform's account, such as Meta's Ads Manager, AI can create the actual campaign: ad sets, targeting, an exclusion list for people who've already ordered nearby, and budget setup, with little to no manual adjustment beyond settings the platform doesn't allow to be changed through the API.

‍

Step 5: Review any AI-generated creative closely before it runs. When a model is connected to tools that can generate or publish content on their own, like a connected design tool, it may go ahead and act even when that wasn't the ask. In Phil's case, it generated ad creative unprompted, and the result wasn't usable, it had to be redone by his team. Watch closely anything connected that can generate, publish, or spend money without a review step.

Conclusion

Phil’s warm-up campaign brought in 963 signups against a goal of 500 to 1,000, at roughly a dollar per signup for their Allentown opening. Thirty days after opening, the location was already sitting in the middle of the pack on loyalty revenue, holding its own against locations with years of established local presence, a strong result for a market where the brand started at zero.

‍

Using AI to build his new store opening plan has helped Phil move faster and with more confidence into unfamiliar markets, while putting what he's learned from past openings to use automatically. His advice to other operators: look at how AI can fit into the tools and processes you already have, and get started.

‍