Slang AI Proved Restaurants Need Voice AI. Now It Needs to Prove They Need Slang.

Sydney Grims had a greeter problem. She runs business development for Fearless Restaurants, fourteen locations plus two hotels, about 1,500 people. She kept watching the same scene: front-of-house staff trying to welcome a walk-in while the phone rang behind them. One location alone took 250 calls in a week. Each unanswered ring represented a reservation that went to a competitor, a catering inquiry that died in voicemail. Private dining bookings never materialized. Event leads vanished. By her math, that single restaurant was hemorrhaging roughly a thousand covers.

"I want my greeters at the front door talking to guests, not on the phone," Grims told Forbes. "The biggest thing isn't just answering questions. It's the customers potentially being lost from multiple calls at once."

That pull, phone ringing versus dining room needing you, is exactly what Slang AI was designed to fix. Out of New York, Slang just pulled in $36 million in Series B money. US Venture Partners led. That brings the total to $68 million. The money, $28 million in equity and $8 million in debt, goes toward expanding beyond voice into text messaging, post-visit follow-up, and hotels.

The central question is not whether restaurants need AI phone agents. They do. The question is whether Slang can maintain its lead in a market that went from empty to overcrowded in eighteen months.

Two data scientists, one overlooked phone line

Alex Sambvani and Gabe Duncan were data scientists at Spotify when they started talking about phones in 2019. Not the streaming kind. The ringing kind. Sambvani, who holds a Stanford engineering degree and a Harvard MBA, had spent years building products for hundreds of millions of users. Duncan brought similar technical depth. Both noticed something that Silicon Valley had largely ignored: the experience of calling a business was terrible, and nobody was fixing it.

"Imagine if you could train Amazon Alexa to answer your business phone calls," Sambvani told Hospitality Technology. "That's what we've created. There's no menu or button pressing."

They picked restaurants deliberately. The industry runs on inbound demand. A busy full-service spot might take hundreds of calls a day. Reservations, hours, directions, allergy questions, private events, catering. Miss those calls and the money walks. Answer them, and you pull a staff member away from the dining room floor.

Slang's founding insight was that the problem is narrow enough to solve well. A restaurant phone call has a limited vocabulary. Questions repeat. Reservation systems have APIs. The challenge was making the AI voice warm enough that callers didn't hang up.

The model

Slang charges restaurants $450 to $600 per month per location for its "Superhost" platform. Every call gets picked up. Midnight on a Tuesday, noon rush on Saturday, doesn't matter. Reservations go through OpenTable, SevenRooms, Tripleseat, or Yelp, depending on what the restaurant already uses. Hours? Dress code? Nut allergy? Superhost fields it. It recognizes returning guests and VIPs. When a caller needs a human, Slang routes the call to the right staff member. When someone asks about takeout, it texts a link to the online ordering platform.

Slang now operates across 2,000+ restaurant locations globally. Texas de Brazil uses it. So do Carmine's, Founding Farmers, Riot Hospitality Group, and Dineamic. Do the rough math at midpoint pricing and you get annual recurring revenue north of $12 million. Slang won't confirm the number.

Unit economics work because the product replaces an expensive alternative: human attention. Grims pays roughly $450 per month per location for Slang at Fearless Restaurants. A dedicated phone employee at minimum wage costs multiples of that. Slang claims restaurants see up to 20x return on investment and a 50% increase in phone reservations. Staff save roughly 200 hours per month. Guest satisfaction scores sit at 96%, a number that USVP partner Rick Lewis called "almost unheard of in enterprise software."

The revenue quality matters, too. Slang's own analysis of its call data found that 71% of all inbound restaurant calls tie directly to revenue: reservations, orders, private dining, catering. Another 34% of calls arrive outside business hours. Every one of those after-hours calls represents money left on the table for restaurants without AI answering.

The inflection

The timing was not accidental.

COVID rewired how Americans interact with restaurants. Walk-in culture gave way to phone orders, pickup requests, and reservation calls. Even after dining rooms reopened, the phone volume stayed elevated. Restaurants that had operated on a "we'll get to it" approach to missed calls discovered that customers had lost patience for voicemail.

The labor shortage made it worse. Restaurants are still hundreds of thousands of workers short of where they stood before COVID. The people who stayed took on more work. Answering the phone became the task most likely to slip.

Slang was already in market when both trends hit. In late 2023, Homebrew wrote the $20 million Series A check. Celebrity chef Tom Colicchio came in too. By that point the question was settled. Restaurants would pay for AI that picked up the phone. The question shifted to how fast Slang could scale.

The proprietary dataset became the accelerant. Over five years of handling restaurant calls, Slang has amassed 25 million customer interactions from 10 million unique guests. That corpus trains models on the specific language, cadence, and intent of hospitality phone calls, a dataset that generic voice AI platforms lack.

The constraints

Slang's problem is no longer whether voice AI works for restaurants. It is that everyone else figured out the same thing.

Restaurant Business Online ran a story in August 2025 that would make any category leader nervous. "The restaurant voice AI market is getting crowded," the headline read. The article noted that startups were flooding the space as barriers to entry fell. Two companies, Vox and Loman, announced seed rounds in the same week.

That opening drew a crowd. Hostie AI started in San Francisco in 2024. It now operates in New York, Miami, and Atlanta. Revmo AI says it completes 82% of calls. Slang sits around 63%, though Revmo generated those benchmarks itself. Loman AI targets pizza chains and high-volume takeout operations, claiming 26% phone revenue increases for early adopters. Certus AI markets itself as a direct Slang alternative with deeper ordering capabilities. RestoHost runs 150 restaurants in Atlanta alone.

Then come the bigger names. SoundHound, publicly traded, already operates across thousands of restaurant locations and recently launched in-car ordering with automakers. Yelp rolled out its own AI receptionist, Yelp Host, late last year. It ships bundled with Yelp's existing marketing tools. Popmenu, already embedded in thousands of restaurants for marketing, found that 83% of diners abandon a restaurant after reaching voicemail twice.

The barrier-to-entry problem is structural. The underlying large language models that power conversational AI are available to anyone. What once required years of proprietary speech recognition can now be assembled from commodity APIs. Give a decent engineer a good prompt and an OpenTable API key. In a few months you've got a passable restaurant phone bot.

Step into a mid-priced Italian place on a Friday night. Count the phones going off behind the host stand. That owner is Slang's ideal customer, but the pricing creates an additional constraint. At $450 to $600 per month, the product targets multi-location groups and upscale independents. The vast majority of America's 750,000+ restaurants are single-location operations with tight margins. Competitors like Hostie start at $199 per month. If voice AI becomes table stakes, the winner may be the one that gets the price low enough for the long tail.

The bet

Sambvani is betting that answering the phone is just the beginning. The Series B announcement made the strategy explicit: Slang will expand into "multi-modal experiences that extend beyond voice" and move into "new areas of the guest journey, from personalized recommendations to post-visit follow-up and reviews."

Sambvani wants the full guest relationship, not just the phone line. Text messages after a missed call. Personalized recommendations for returning diners. Automated review requests and event follow-up for private dining inquiries would round out the product. Hotels and venues, not just restaurants.

That ambition has a foundation. Slang already sits at the first point of contact between a guest and a business. Its dataset captures not just what callers ask, but when they ask, how often they return, and what they book. If you run a 10-location restaurant group, that behavioral data could power a communications product that competitors, focused narrowly on phone answering, cannot match.

But building a phone agent is one problem. Building a full guest lifecycle platform is several problems stacked on top of each other. Every new modality requires its own integrations and compliance burden, plus a customer success playbook that doesn't exist yet. Slang has 68 employees' worth of funding. SoundHound has a public market valuation and automotive partnerships. Yelp pulls 90 million monthly users.

Claire Hughes Johnson put money into this round. She ran operations at Stripe. She knows what scaling looks like up close. Her participation signals belief that Slang can make the leap from point solution to platform. But belief and execution remain different currencies.

What it signals

Slang AI's Series B tells a broader story about where vertical AI is heading.

AI's first wave targeted horizontal problems: write better emails, summarize meetings. The second wave picked industries. Slang picked restaurants. Harvey picked law. The bet across all of them is identical: domain expertise, not general intelligence, creates the defensible position.

The restaurant voice AI market now functions as a live test of that thesis. The underlying AI technology is commoditizing rapidly. What remains proprietary is the training data and the operational knowledge baked into every call flow. Slang's 25 million calls represent something a new entrant cannot buy off the shelf. Whether that corpus matters more than a competitor's lower price or a larger platform's distribution is the experiment running in real time.

The money flowing in reflects what everybody already sees. The market sits around $10 billion today. By 2029 it could hit $49 billion. That expansion creates room for multiple winners. But it also attracts the kind of attention, from Yelp and SoundHound to the next YC batch, that turns a blue ocean red.

The investor composition of this round reflects the tension. USVP leads with a growth-stage bet. Thayer Investment Partners brings hospitality sector expertise. Johnson brings operational credibility. The existing investors, Homebrew, Wing VC, Stage 2 Capital, all doubled down. None of them are running for the exits. But the round size tells you something. At $68 million total raised with no disclosed valuation, Slang has proven product-market fit. It has not yet proven it can dominate.

The test

The next twelve months will answer a specific question: Can Slang grow faster than the market commoditizes?

Slang needs to double its restaurant footprint while maintaining the 96% satisfaction scores that differentiate it from cheaper alternatives. A credible multi-modal product, text and email at minimum, has to ship before competitors expand from the other direction. And if you're Sambvani, the hotel expansion is the real proof point, the evidence that "Superhost" translates beyond the restaurant phone line.

The measurable milestone is straightforward. By early 2027, Slang needs to be handling communications across multiple channels for a material number of its 2,000+ locations. If the expansion works, the company becomes the guest communications platform it claims to be, and the 25 million call dataset becomes the foundation for something much larger. If it stalls, Slang remains a good phone answering product in a market where good phone answering products are no longer scarce.

Sambvani built Slang on a simple observation: restaurants lose money when they miss calls. Five years and $68 million later, that observation has been validated by an entire industry of imitators. The next chapter requires a harder proof. Not that the phone matters, but that Slang matters more than the phone.

SpaceX and xAI Enter Secret $100M Pentagon Contest for Autonomous Drone Swarms

SpaceX and its subsidiary xAI are competing in a secret Pentagon contest to build voice-controlled autonomous drone swarming technology, Bloomberg reported Sunday. The $100 million prize challenge, la

The Implicator

Nvidia's Six-Chip Gambit: How Jensen Huang Is Building a Computing Empire You Can't Escape

Jensen Huang walked onto the CES stage in Las Vegas on Monday wearing a black jacket with a crocodile-scale pattern—his signature aesthetic of controlled flamboyance. What followed was less a product

The Implicator

OpenAI merges audio teams, targets new voice architecture by March 2026

OpenAI just merged several engineering, product, and research teams to overhaul its audio models. The company plans to ship a new architecture by March 2026 that sounds more natural, handles interrupt

The Implicator