Building A Restaurant AI Startup From Zero To 1,000 Customers In A Year
with Christian Wiens, Loman.ai
Building A Restaurant AI Startup From Zero To 1,000 Customers In A Year
Show Notes
Friday night at a busy pizza place: the line is out the door, the kitchen is yelling, and the phone won't stop ringing. Forty percent of those calls will go unanswered. Not because the restaurant doesn't want the business - but because every employee is already doing three things at once, and answering a phone is a task that requires full attention for as long as the caller needs.
Christian Wiens grew up working in restaurants. He saw the problem firsthand - including the trick of unplugging the phone on Friday nights at peak summer beach season just to survive the shift. When he saw the first voice AI demos in early 2024, the problem and the solution snapped together immediately. Loman.ai was live with its first customer by April 2024. By the time this episode was recorded, it had crossed 1,000 restaurant customers, raised $3.5 million, and was growing entirely on inbound demand.
This episode is about what it looks like to time a market perfectly, build a vertical-specific AI product that customers switch POS systems to use, and create a GTM engine that runs on customer stories and SEO - not a sales army.
What Loman Actually Does
Loman is a generative AI voice agent that answers every inbound call for a restaurant. Not a phone tree. Not an IVR. A fully conversational AI that greets the caller by name, knows the menu, knows what is 86'd in real time, knows your order history, and can take a payment over the phone - PCI compliant - while the kitchen prints the ticket simultaneously.
The product integrates directly with the major POS systems (Toast, Square, Clover, Spot On, OpenTable) and pulls live data: prep times, stock status, menu pricing, reservation availability. Sub-500ms latency means the turn-switching feels human. Forty percent of callers, according to Christian's data, do not know they are speaking to AI. The ones who do know often ask more questions - and that is also a feature.
The inference engine is the differentiator. Every call generates a diner profile: name, order history, contextual signals picked up during the call (background noise, dietary patterns, family cues). That profile informs upselling on every future call and eventually enables personalized outreach. Today the product is 100% focused on inbound call handling. The profile data sets up what comes next.
Frameworks from This Episode
These frameworks have been added to the AI for Founders Frameworks Library. Filter by Christian Wiens (Loman.ai) to find them.
The Restaurant Revenue Triangle
Three levers, two revenue and one cost: answer 100% of calls instead of 60% (direct revenue recovery), upsell on every single order instead of 5-8% of them (average order value lift), and redirect the labor hours previously absorbed by phone handling back to hospitality or eliminate them. The math stacks: recovered calls plus higher ticket size plus labor savings puts Loman in the class of products that improve the restaurant's P&L on multiple lines simultaneously.
Vertical AI GTM Playbook
Know exactly where your ICP lives - restaurant owners are on Facebook, not LinkedIn. Start SEO from day one even though it produces nothing for months. Build customer video stories as your primary sales asset and let customers do the talking. Create a referral engine through product quality alone before you build a referral program. Match your content channels to your buyer, not to what feels sophisticated.
Timing as Product Strategy
The founders who tried to build restaurant voice AI before GPT-3.5 had to train their own models - expensive, slow, and bad. The founders who launched two months after GPT-3.5 got the same capability for the cost of an API call. Christian's principle: if an LLM cannot yet do the thing you need, waiting is sometimes the right move. The technology timeline is your roadmap as much as your feature list is.
Tools from This Episode
Voice AI agent for restaurants. Answers every call, takes orders, makes reservations, upsells, and integrates directly with Toast, Square, Clover, Spot On, and OpenTable.
This Week's Experiment
Run a Phone Revenue Audit: Count Your Missed Calls for One Week
For any service business that takes inbound calls, spend one week logging every incoming call: answered immediately, put on hold, or missed entirely. For missed calls, estimate the average order or ticket value and calculate the weekly revenue leak. For answered calls, note how often your team attempted an upsell. Compare your actual upsell rate to what 100% AI upselling would produce. The gap between those two numbers is your Loman case study.
The Three-Revenue-Line Math
Christian breaks Loman's impact into three categories. First: revenue recovery. About 30% of restaurant revenue still comes through the phone, and 40-50% of those calls currently go unanswered or are put on hold. Answering all of them recaptures that revenue directly. Second: average order value. Staff upsells 5-8% of the time. Loman upsells 100% of the time. Even if only 30% of customers accept the upsell, that is a significant lift on every order taken by phone. Third: labor. The pizza restaurant Christian references saves $2,700 per month per location in labor by using Loman. Other customers keep the headcount and redeploy those hours to floor hospitality.
For a restaurant running on 3-5% profit margins, these three lines can be the difference between a profitable location and a struggling one. A $2,700 monthly labor saving on a restaurant doing $100,000 in monthly revenue represents most of a typical month's profit. The combination of recovered revenue, higher ticket sizes, and labor efficiency is why 15% of Loman's customers switch their POS systems specifically to enable the integration.
GTM Built for Restaurant Owners, Not Investors
Christian's GTM stack: SEO from the first day the website was live (now driving about 25% of leads); customer video stories produced on every notable account and distributed across Meta, YouTube, TikTok, and LinkedIn; and a 20% referral rate from customers - without a formal referral program - driven purely by product quality. A single long-form customer interview becomes YouTube content, shorts across every platform, social cards, and blog posts. One shoot, maximum distribution.
The ICP insight: restaurant owners are on Facebook. They are not on LinkedIn. Christian's early distribution was on small business Facebook groups - his first three restaurant owners responded within 30 minutes of his first post. The lesson: match your content distribution to where your buyer actually lives, not to what looks credible to other founders.
The team reflects the market. Sales is staffed with people who built their careers at DoorDash and Toast - people who have already sold to restaurant owners and know how they think. Engineering is staffed with people from Deepgram and ElevenLabs who built voice infrastructure. The founding insight: you need domain credibility on both sides of a vertical AI company.
Q&A
What exactly does Loman do and how is it different from a phone tree?
Loman is a fully generative AI voice agent - not a phone tree or IVR. It answers the phone as if it were a human employee of the restaurant, knows the full menu in real time (including what is out of stock), takes orders that print directly in the kitchen, makes reservations, answers questions about parking and specials, and can process PCI-compliant payments over the phone. The conversation is free-form: the caller speaks naturally and the AI responds contextually. Sub-500ms latency makes the experience feel human. About 40% of callers do not realize they are speaking to AI.
How does Loman make restaurants more money?
Three mechanisms. First, revenue recovery: 40-50% of restaurant calls currently go unanswered or on hold, and about 30% of restaurant revenue still comes through the phone - Loman answers every call, every time, capturing that revenue. Second, average order value: human staff upsell 5-8% of the time; Loman attempts an upsell on every single order, which even at a 30% acceptance rate represents a significant ticket lift. Third, labor efficiency: customers typically save around $2,700 per month per location in labor, or redeploy those hours to floor hospitality rather than eliminating them.
What is the inference engine and why does it matter?
After every call, Loman saves a profile of the diner: their name, order history, and contextual signals picked up during the call (background noise, dietary patterns, family references). That profile is used on every future call to personalize the experience - the AI knows your usual order, knows to suggest the spicy version if that's your pattern, knows your name. It also feeds upselling: the system learns what you are likely to order and surfaces it. The long-term value is personalized outreach and remarketing using data no restaurant currently captures systematically.
What is the right time to start SEO for an early-stage startup?
Day one. Christian started SEO the day he built the Loman website, even though it produces no results immediately. A quarter of Loman's inbound leads now come from organic search. The logic: SEO compounds over time, and the cost of starting late is permanent. For founders in new AI categories, there is often very little competition for relevant keywords - it is a greenfield grab. The tool set for doing SEO has also become largely automated: AI writing tools, backlink automation, and keyword tracking mean a solo founder can set up a functioning SEO engine without a dedicated marketing hire.
How did Loman grow to 1,000 customers with a 14-person team?
The growth engine has three parts: SEO (25% of leads), customer video stories distributed across Facebook, Instagram, YouTube Shorts, TikTok, and LinkedIn (converted to clips, blog posts, and social cards from single shoots), and word-of-mouth referrals (20% of leads, with no formal referral program). The product quality drives the referrals: customers who generate $2,700 in monthly labor savings per location and see their phone revenue stop leaking become advocates. The team matched the GTM to the buyer's media diet - restaurant owners are on Facebook, not LinkedIn.
What does Loman's founding team look like and why does it matter?
The sales side is staffed with restaurant industry veterans: the first DoorDash sales person, Toast President's Club winners, the first Olo sales rep. The technical side comes from the voice AI world: former Deepgram and ElevenLabs engineers. Christian's view: vertical AI companies need domain credibility on both sides. A great voice AI product built by engineers who have never sold to restaurant owners will struggle to close deals. A great sales team that cannot explain why the AI actually works will lose to someone who can. The combination of restaurant domain expertise and voice AI engineering expertise is the competitive moat.