Zest AI Agent Brilliantly Books Restaurant Tables Over Text

The Zest AI agent turns one of life’s small chores — deciding where to eat — into a text-message conversation. The restaurant-discovery startup just launched a personal dining concierge that lives in your SMS thread, suggests places matched to your taste, and books the table without ever forcing you to open an app.

Launched on Thursday, the feature puts Zest alongside a growing wave of text-native assistants that promise to run real errands inside ordinary messaging. This article covers how the Zest AI agent works, how the company knows what you like, what powers each confirmed reservation behind the scenes, and why its founder believes texting — not apps — is where dining decisions now live.

Table Of Contents

What the Zest AI Agent Does Over Text
How the Zest AI Agent Knows Where You Actually Eat
Why Text-Native Assistants Are Replacing Apps
The Reservation Platforms Powering Every Reply
Zest AI Agent vs Every Other Way to Book a Table
The Real Fight Is Over Who Owns the Customer
How Zest Plans to Make Money from Dinner
What Comes Next for Text-Based Dining
References

What the Zest AI Agent Does Over Text

A Concierge That Answers Like a Friend

To reach it, you text the number (212) 470-6067. The Zest AI agent replies like a well-connected friend: you describe the evening — cuisine, craving, occasion, dietary restrictions, or neighborhood — and it returns a shortlist of restaurants that fit. You can ask follow-up questions about any pick, including how busy a place tends to get, whether it has Wi-Fi, or even what’s on the menu.

From Suggestion to Confirmed Booking

Unlike a recommendation engine that stops at a list, this concierge checks live availability and makes the reservation itself through connected services such as OpenTable, Resy, and SevenRooms. TechCrunch reports the startup began testing the concierge last month but held the launch until the booking step was ready, because a suggestion without a table is only half the job.

Group Chats Get a Tiebreaker

The Zest AI agent can also be dropped into a group chat, where it blends everyone’s tastes before proposing places the whole group will like. When friends stall, an in-thread voting feature settles the argument, and the agent books the winning table for everyone without anyone leaving the conversation. A launch page for the service lives at textzest.com.

The Limits of an SMS Concierge

Texting has trade-offs the demo videos skip. A message thread cannot show a map pin, a photo strip, or a star rating inline — rich context has to arrive as links. Browsing thirty options is faster by scrolling than by reading, and the concierge is only as bookable as the reservation platforms a given restaurant uses.

Zest is betting that conversation still beats those limits for the plan-to-book stretch of the journey, where most apps lose people between shortlist and confirmed table. The Zest AI agent wins that stretch only if every reply lands fast enough to feel like texting a person.

How the Zest AI Agent Knows Where You Actually Eat

Credit-Card Signals, Filtered by Plaid

Zest was founded in November 2024 with $1.8 million in pre-seed backing from Alexis Ohanian at 776 and Steve Jang at Kindred Ventures. Its original twist: the app links to your credit cards through the financial-data company Plaid to see where you actually dine out, not just where you claim you want to. The company says it looks only at transactions in food and drink categories to build your personalized map and lists.

Skipping the Cold-Start Problem

That transaction history is why the Zest AI agent doesn’t need a long onboarding quiz or weeks of chatting to be useful. “You text it, and it answers from your own history and your friends’. It doesn’t need weeks of conversation to learn you because it starts from where you’ve actually been,” founder and CEO Mario Gomez-Hall says. Suggestions work in your home city and while you travel, where a tourist list and a local’s shortlist part ways.

Privacy Is the Obvious Question

Letting an app read your restaurant transactions is a bigger ask than letting it read your searches — the agent learns date nights, business dinners, and who you were with that night. Zest’s category filter limits the scope, but diners weighing the trade should read our breakdown of whether AI agent privacy promises are ever kept. The convenience is real; so is the data trail.

Building Your Personalized Map

The raw card transactions land as dots first. Zest clusters them into venues, weights the neighborhoods you actually return to, and merges the overlaps between friends into something like a shared taste graph.

Only then does the Zest AI agent use that map as grounding, so a free-text question like “somewhere lively for four on Friday” is answered against your habits before it parses a single word of your prompt.

Travel Suggestions Without Tourist Traps

The history pays off hardest away from home. Because the company tracks where you dine out regularly, it can match the pattern of venues you love — cuisine, price band, density, how late the kitchen keeps going — against suggestions in a city you have never eaten in.

That is the pitch for travelers: skip the top-clicked list everyone else sees, and get picks that read like a well-traveled friend shortlisted them overnight.

Why Text-Native Assistants Are Replacing Apps

The App Is Quietly Becoming the Backend

Zest’s move reflects a broader shift: conversational agents are absorbing tasks that used to demand stand-alone apps. TechCrunch counted two dozen agents that now live inside text messages, from Instinct to Poke, and we assembled our own ranked list of text-message AI agents earlier this month. The interface that won is the one everyone already has open.

What the Zest AI Agent Adds to the Thread

Here is how the text concierge compares with Zest’s own app, the product it grew from:

Factor	Zest app	Zest AI agent (text)
Interface	Maps, lists, search	An SMS thread
Group planning	Limited	Native group chats with voting
Booking tables	Hands off to other apps	Checks availability and books in-thread
Restaurant Q&A	Static listings	Ask about Wi-Fi, crowds, menus

TechCrunch’s Hands-On Verdict

In TechCrunch’s own tests, the agent worked fairly well, matching suggestions to taste while respecting conditions like time of day, party size, and location. Finding an open table is not new — apps have done that for years — but having the Zest AI agent quote available times back to you in plain sentences is the convenience that may pull texting past app-hopping for good.

Why SMS Beats a Chat Tab

Zest chose the oldest protocol on the phone, and the choice is deliberate. SMS needs no install, no account, and no new icon on the home screen — you text a number and the concierge exists. Read receipts, group threads, and contact cards already work the way millions of people expect.

That also reaches users a standalone chat app never would: people who plan dinners exclusively in the group message and would never download another reservations app. Distribution, in other words, is the real reason the launch lives in your inbox.

The Reservation Platforms Powering Every Reply

The Plumbing Behind the Punchline

When the Zest AI agent books, it is not calling the host directly. It connects through connected services — OpenTable, Resy, and SevenRooms — the reservation platforms where most US restaurants already publish their tables. Zest reads live availability from them, holds your slot, and returns a confirmation inside the message thread.

What Each Platform Contributes

OpenTable brings the broadest inventory, Resy anchors many of the country’s most in-demand rooms, and SevenRooms powers the guest-management systems behind independent restaurants. Together with Tock for time-slot dining, they define what any agent can promise. The agent is only as bookable as the networks it plugs into — which is why launching the booking step shaped Zest’s timeline.

Why Booking Is the Hard Part

Recommending a restaurant is a solved problem; committing a table is not. Availability moves while you chat, parties change size, and restaurants release tables in bursts. The Zest AI agent has to re-check, hold, and confirm inside the latency of a normal text exchange — the unglamorous engineering that separates a concierge from a chatbot.

What Restaurants See

From the dining room’s side, an agent booking looks much like any other platform request: a party, a time, a headcount arriving through a connected system. Restaurants keep control of their table maps, cancellation policies, and covers, and none of that changes because the intent started as a text.

The shift is in who shows the restaurant to the customer first. If diners stop opening reservation apps, venues depend more heavily on whichever agents and platforms sit between them and the table — which is exactly why platforms watch agent traffic so closely.

Could This Undermine the Platforms?

There is a quieter tension: OpenTable, Resy, and SevenRooms power the booking, but the Zest AI agent owns the relationship, the taste memory, and next time’s shortlist. Platforms tolerate intermediaries until the intermediaries become the front door.

For now the incentives align — more booked tables means more fees for everyone — but the balance of who charges whom could look very different once texting is how a meaningful share of diners start the evening.

Zest AI Agent vs Every Other Way to Book a Table

Four Ways to Dinner, Compared

Buyers effectively choose between the classic app, a general-purpose assistant, the phone call, and now the Zest AI agent. Each trades convenience for control differently:

Capability	Restaurant app	General AI assistant	Zest AI agent (text)
Learns your taste	From your ratings	Only from chat	From real card history
Group voting	No	No	Native in group chats
Books in-thread	No	Rarely	Yes, via OpenTable/Resy/SevenRooms
Answers restaurant questions	Static listings	Guesswork	Menus, crowds, Wi-Fi

Where the Zest AI Agent Wins

The agent’s edge is context plus continuity. It already knows you liked the ramen on Pike Street and that your friend never eats red meat, so the first shortlist is better than a blank search page. And unlike a general assistant, it can actually close the loop: TechCrunch’s reviewers watched it quote real times and complete the booking.

Where It Can Still Lose

Texting gives up visual browsing — photos, maps, and menu grids are harder to scan in a thread than in an app. SMS also has no universal identity, so spam filters and carrier rules can interfere. And if your favorite bistro only takes reservations by phone, no agent can help.

Friction, Measured

The takeaway: for a group dinner, texting removes most of the steps between craving and confirmation, even if each step still happens behind the scenes.

Approx. steps from craving to confirmed table (illustrative)

Phone the restaurant 6

Restaurant app 5

General AI assistant 4

Zest AI agent text 2

Honest Scores on Both Sides

The specialist wins on taste memory and booking depth, but general assistants keep real advantages. They already know your calendar, your messages, and your location; they cost nothing extra beyond what you pay; and they can reschedule the dinner, text the RSVP, and book the babysitter in one place. Specialised agents ask for their own data grant, their own thread, and their own trust.

Expect overlap fast. The moment a general assistant can wire into reservation platforms with the same reliability, the difference narrows to who learned your taste first — which is precisely the ground the Zest AI agent is trying to claim now, while it can.

The Real Fight Is Over Who Owns the Customer

“The Agent Is the Product Now”

Gomez-Hall’s one-line summary of the launch is also the strategy: the agent is the product now. The map, the lists, the app — all of it becomes plumbing for a relationship conducted in the messaging app your users never close. That is why the launch matters beyond dinner: whoever books your table learns your calendar, your party, your budget, and your city.

Incumbents Are Watching

The text-agent field is crowded and well-funded — Instinct raised a $1 billion Series C at a $10 billion valuation, and our write-up asked whether even the buzziest agents can survive the platforms they live inside. Dining is a tempting vertical: high frequency, real money, and a natural group dynamic that generic assistants fumble.

The Trust Moat

Owning that conversation requires trust. Users already hesitate about letting assistants run pieces of their life, and Zest compounds the question by starting from credit-card history. Its bet is simple: diners will trade data for a concierge that never suggests the wrong room.

Why Dining Is the Perfect Agent Pilot

Dinner plans are the rare task that is frequent, social, paid, and forgiven. A wrong restaurant suggestion costs a bad evening, not a missed flight or a bad mortgage. High frequency means the Zest AI agent gets dozens of chances a month to prove itself, and every corrected pick sharpens the next answer.

The social layer matters just as much: every group chat where the agent books is marketing that costs nothing, because the four friends who watched it work are the next four users.

What the Incumbents Have to Lose

If a specialist concierge owns the dinner conversation, reservation platforms risk becoming wholesale infrastructure — fulfilling bookings set up by someone else’s interface, with someone else’s brand on the customer’s memory. That is the same displacement every software category fears, arriving in food first because the task is small, cheap, and endless.

How Zest Plans to Make Money from Dinner

Bookings First, Ordering Next

Zest has been honest about the business model. “Booking and ordering are the first revenue lines. Zest books your table through OpenTable or Resy today, with food ordering coming, and there are transactions there to participate in,” Gomez-Hall says. In practice that means participating in the booking fees and referral economics that reservation and ordering platforms already pay to partners who send diners.

The Scale Problem the Zest AI Agent Solves

The app reached its first thousands of users organically; texting is the distribution play to grow past them. A concierge that works inside a group chat spreads exactly when plans spread — every group is a funnel. And because the Zest AI agent learns faster than a general assistant, each completed reservation improves suggestions, which drives more bookings, which improves revenue.

The Longer Game

Over time, the startup is weighing expansion beyond restaurants into businesses, shopping, and more — recommendation transactions everywhere you spend locally. Here is the roadmap as Zest describes it:

Revenue line	How it earns	Status
Table bookings	Partner fees via OpenTable/Resy/SevenRooms	Live now
Food ordering	Transaction share on orders	Coming next
Beyond restaurants	Local business and shopping referrals	Under consideration

“Worth Something Well Beyond Dinner”

“The reason that gets interesting is that Zest learns your taste faster than a general assistant does, because it starts from where you’ve actually been instead of asking you questions. That’s worth something well beyond dinner,” Gomez-Hall adds. A taste profile built on real spending is an asset class — if the company can monetise it without becoming the creepy friend who knows too much.

The Unit-Economics Bet

Participating in booking transactions is modest per table — reservation fees are small, and restaurants have long resisted paying for covers. What makes it work at scale is frequency plus accuracy: a concierge that books the same household twice a month, every month, turns tiny referral slices into a real revenue line.

Ordering changes the math further. A food order carries a much larger ticket than a reservation, and the same taste model that picks your dinner can pick your Friday night takeout without a new introduction.

The Risk Side of the Roadmap

Revenue rides on goodwill. One bad booking experience, one data incident, or one visibly wrong suggestion in a group chat can undo weeks of trust-building, and the Zest AI agent has no human concierge to call and apologise. The company’s discipline with the card data will decide how far this model runs.

What Comes Next for Text-Based Dining

The Bet Against the App Store

Zest frames the launch as a wager on where interfaces go next: founder Gomez-Hall openly questions whether mobile apps remain the primary interface at all. If the Zest AI agent keeps gaining ground, the restaurant discovery category ends up as a feature inside messaging — with the concierge collecting the referral fees.

Is the Zest AI Agent Worth Texting?

If you already track spending in an app like YNAB or Mint, connecting Zest to Plaid is a familiar trade. Text the concierge once with a real plan — cuisine, city, party size — and judge whether the shortlist surprises you. For group dinners specifically, the voting-plus-booking combo is the clearest reason to switch, because it collapses three apps into one thread.

Keep expectations grounded: the Zest AI agent is strongest when your history is rich, and a sparse card trail gives it little to work with. Give it a few weeks of real use before deciding it is not worth the thread.

What to Watch

Three signals will tell us how this lands: whether general assistants bolt on in-thread booking of their own, whether OpenTable and Resy tighten or welcome agent access, and whether diners keep a second AI thread on speed dial. We will follow all three — and update you as the Zest AI agent grows beyond dinner.