case study // what i'd build
A small taproom loses hours every week answering the same five questions and chasing event inquiries in DMs. Here is the system I'd build to take that off the owner's plate — a fast website, a live tap list, and an AI agent that handles the front desk on the web and over text.
Foothold Brewing — a 7-bbl neighborhood taproom outside Asheville. Two owners, four part-time pourers, twelve taps that rotate weekly, a rotating food-truck calendar, and a back room they rent for private events. No marketing hire. The website is a one-page site from 2019 with a PDF menu that's three months stale.
None of this is a beer problem. It's a front-of-house and information problem — exactly the kind of repetitive, rules-based work an agent handles well, and a brewer rarely has time to systematize.
A clean, mobile-first site that loads instantly. Hours, location, the rotating food-truck calendar, and a live tap list that updates itself — never stale again.
A chat agent on the site and over SMS that answers FAQs in the taproom's voice, drafts private-event quotes, and captures every lead with a name and email.
The agent drafts a weekly "what's on tap" email and social post from the same live data, ready for the owner to approve in one tap.
Plain architecture — nothing the owner has to babysit. The agent reads from one source of truth (the tap list and event rules), so the site, the chat, the texts, and the weekly email never disagree with each other.
Built on a static front end (fast, cheap to host), an LLM grounded in the taproom's data, and lightweight hooks into the tools they already use — point-of-sale for taps, a calendar for events, a phone number for SMS. Guardrails keep the agent inside what it actually knows; anything it can't answer routes straight to the owner.
Why these are credible: faster lead response is one of the most consistent findings in sales research, and the admin hours come from tasks the agent removes entirely. On a live account I'd instrument all three and report the real numbers — measured, not asserted.
I'm a fermentation scientist who ships software. I worked as a Head Distiller, spent seven years as a Yeast Specialist at New Belgium, and ran yeast and microbiology at the ASBC (Microbiology committee, via White Labs) — then I taught myself to build agentic AI, software, and websites, and now I ship them. So I speak both languages: I understand a brewer's actual day, and I can build the system without a translation layer or a five-person agency markup. fermentation × intelligence × code.
Pricing shared after the Diagnostic, scoped to the actual build — no surprise retainers.
live // let's talk
I'll do the half-day Diagnostic first so you see the plan before you spend on the build. If it's not worth it, I'll tell you.
Book a free 30-min call →