How Small Spirits Brands Can Use AI to Outmaneuver Legacy Brands

Wyatt Carder: What we want to talk about today is how we’re using AI specifically—building one specific tool to create leverage for our sales team.

The idea here is that with new technology and what’s available to smaller brands at a lower cost, we can do a lot more in 2026 than we could even in 2024. Since AI has come out as a consumer product, along with all of these other technologies that we can use as harnesses or leverage, it really creates a lot of opportunity.

The use case that we’re going to talk about today centers on one of our key metrics in running the brand: how many accounts actually purchase the product per month, and how we avoid churn. That is a primary objective of ours.

So, where we start with this system is our own set of sales data, which most brands have access to in some format—whether it’s from their distributor, a portal, or whatever you have. The goal is to connect it to an LLM or AI capability (which could be Claude or ChatGPT), and then create a layer where you can interact with that data on your own.

Unlike using ChatGPT in the standard chat interface that everybody else uses, we want our own interface that connects directly to our internal data. For us, that interface is Slack. The result is that we can chat with our sales data and perform any analysis we need, provided we’re smart about how we structure it.

For building these workflows, the basic tenant is that you need an input. For us, that’s our actual sales data. That data needs to go into a database to hold it. Then we create the inference using AI, which is how we get smart about the data. As that loop goes around, the system gets smarter, and the output yields either time efficiency or more revenue made—which is the ultimate goal with all of these tools.

The tech stack that we use for this specific tool includes:

• Claude Code: For the initial build. You can use natural language to tell it what you’re looking to do, and it writes the code and helps you deploy it where needed.

• Trigger.dev: To set it up automatically. When you’re talking about agents and AI, you want execution to happen without your manual intervention. Once configured, you want it to run automatically, and Trigger.dev hosts and handles that automation.

• Supabase: Where we store our information. It serves as the database that the AI reads.

• Notion: The interface we use for actually viewing and working with the data. We set up our own dashboards there for our team to use.

• Slack: The primary communication method we use to tap into the data and chat with it in real time.

All of that costs about $100 a month. A lot of these tools are free—Trigger.dev, Supabase, and Slack all have free tiers that we utilize. Notion is seat-based, and Claude Code can run up to $200 a month depending on API usage, but you can get this entire workflow running on a $100-a-month budget with no problem.

Looking at the entire system architecture: all information comes into our data portal running on a Mac Mini. Think of the Mac Mini as an always-on local server. Unlike a laptop that shuts down when you close the lid at the end of the day, the Mac Mini stays running constantly so that Trigger.dev crons can execute automated tasks on a machine that is awake.

That data feeds into Supabase where it lives and is read. Supabase connects bi-directionally with Notion, where our team works visually. From there, Trigger.dev pushes automated alerts into Slack, making the Slack interface live and able to query our sales data through the LLM.

When Joey messages Slack from the field, Vercel (where everything is hosted) picks it up, Trigger.dev routes the information, Claude processes the request, reads the database, makes a decision, and sends the response right back to Slack. In the exact same way he would send me a text asking “What are we doing later?”, he is chatting directly with our sales data.

For this specific use case, our goal is to catch account drift. We’ve determined that our account base follows specific order patterns—some order every 30 days, some every 60 days, and some on a variation in between. When an account that historically ordered on a regular cycle changes its pattern, we flag that as account drift. The system calculates these metrics and reports back: “This set of accounts is expected to order this month based on their pattern; this set of accounts is drifting.” We then use that intelligence to ensure those accounts receive immediate touches and account maintenance.

That report comes directly into our Slack channel every Monday, Wednesday, and Friday. It summarizes: “Out of X amount of accounts, here is how many ordered this month,” along with a direct link to the Notion database where Joey can review the expected accounts.

Joey Angelo: As a traditional salesperson, seeing stuff like this initially can be intimidating. When Wyatt and I started this journey, I was a little apprehensive until I fully wrapped my head around it. You see tech like this and think it’s designed to replace people like me. That isn’t the case at all. What it actually does is give you extra time back—time you can put to good use sitting across from your customers. It is a very, very powerful tool.

We have several workflows set up, but this is one we use on a regular basis. That entire system Wyatt described literally feeds right to my mobile phone via Slack, just like a DM. Any question I want to ask about any market in the United States is at my fingertips. You think it, you text it, and the data comes right back to you.

When you work with different salespeople, distributor managers, VPs, and directors across the country, many don’t even know their own real-time numbers, or they tell you, “I have to go back to the office and get back to you tomorrow.” After working with a real-time setup like this for a week, you can never go back to the old way.

To give a real example of flipping drifting accounts: any good salesperson knows that you can open all the accounts in the world, but if you don’t retain them, churn will kill your business. We use this to eliminate churn.

While sitting in Miami, I can run a quick check to see which local accounts are slipping. Right on my phone, while sitting at a red light or in the back of an Uber, I text the query and it returns a list—for instance, 24 drifting accounts in that market.

Looking down the list, I might see an account like Moxy Low Roof. I know that venue closed briefly for rooftop construction and recently reopened. When managing thousands of accounts nationwide, you can’t manually keep track of every local nuance, but when the account pops up as drifting on your screen, you immediately know to go visit them. By simply contacting them or making an in-person drop-in, those accounts usually come right back.

Wyatt Carder: That’s where this creates real operational leverage. Previously, Joey would be on the road and have to email me or someone else asking, “What’s going on with this account?” or “Can you run a report on which accounts are ordering?” The fact that he can pull that up himself in real time empowers him in the field—and makes my inbox a lot less cluttered!

Being able to query the universe of accounts around him using natural language allows the LLM to compute trends, flag account velocity changes, or highlight which key accounts haven’t been visited recently.

Joey Angelo: You might think in your head that you visited an account just last week, or that they already placed an order. But when you pull up the black-and-white data in real time, you realize that wasn’t the case. Having exact order dates available via text is an incredible tool.

This workflow replaces old distributor data portals, Excel spreadsheets, endless email threads, and manual number-crunching that results in stale data. When you pull a traditional report, it’s only accurate the day you run it—by tomorrow, the numbers are already outdated. With an automated AI loop reading live databases, the data updates continuously.

In this industry, distributor data is often two to three weeks old by the time you receive it. Having real-time, constantly flowing data gives you a massive speed advantage.

People ask how a small craft brand can stay competitive against major industry strategics without getting wiped out. The answer is using these tools to stay extremely nimble in your day-to-day operations. Initially, our goal was just to survive, but now in several major markets, we’re beating large competitors on velocity and account retention metrics.

Wyatt Carder: The main takeaway for everyone is that neither of us are software engineers, yet we built this entire pipeline. It takes some initial effort, but the educational resources are freely available online. In fact, you can use AI to teach you how to build AI tools—you can take screenshots of error logs, ask the model for debugging guidance, or use browser tools to help configure the setup.

It enables non-technical founders with an idea or an operational problem to build their own custom solutions. Once that foundation is set up, the possibilities are endless—especially given how affordable these technologies are to run.