📰 Key Takeaways

Zest, a restaurant discovery app founded last November, has launched a new feature: a personal AI “restaurant concierge” that lives in your text messages, similar in concept to AI agents like Instinct or Poke. Users just text Zest to get recommendations based on their own and their friends’ dining history, and can book a table directly. Founder and CEO Mario Gomez-Hall says “the agent is the product,” because the system starts from places users have actually been and doesn’t need weeks of conversation to get to know them.

What makes Zest different is that it connects to users’ credit cards through the financial services company Plaid, tracking real dining-out spending to figure out their tastes and interests. It only pulls dining-related transactions, and uses them to generate personalized maps and lists, with suggestions both in your home city and when you travel. The company was founded in November 2024 and raised $1.8 million in pre-seed funding from Alexis Ohanian of 776 and Steve Jang of Kindred Ventures.

The new feature can check availability and complete bookings through services like OpenTable, Resy, and SevenRooms. Zest can also join group chats, combine everyone’s tastes, settle indecision with an in-thread voting feature, and then book for the whole group. Users can ask by cuisine, what they’re craving right now, the occasion, dietary restrictions, or area. They can also ask how busy a restaurant is, whether it has Wi-Fi, or have it send over the menu. In TechCrunch’s hands-on test, the AI did a pretty good job listing new restaurants that fit their tastes based on time slot, party size, and location. For the rest of the test details, see the original article linked below.

💬 JudyAI Lab Take

Zest puts an AI agent right inside text messages, so you send one message and get a recommendation and a confirmed reservation. What’s worth noticing is that it doesn’t start from a blank conversation. It starts from what users have actually spent on dining with their credit cards.

Founder Mario Gomez-Hall says “the agent is the product,” which points to a design mindset: an agent’s value isn’t in how clever the chat interface is, it’s in having the right data from day one. Zest connects to credit cards through Plaid and only pulls dining transactions, which skips the weeks of conversation the system would otherwise need to get to know you. Another thing worth noting is the choice of entry point. Text messages and group chats are where people already are. Once Zest joins a group chat, it blends everyone’s tastes, settles the “where do we eat?” debate with in-thread voting, and books for the whole group. Going from recommendations to checking availability on OpenTable, Resy, and SevenRooms and actually completing the booking also shows that an agent’s endpoint is getting the thing done, not just giving advice. For AI builders, the data source, the entry point, and whether you can close the loop often matter more to a product’s weight than the model itself.

Next time you design an agent, ask yourself first: what behavioral data do users already have that you can legitimately access, so the agent is useful from the very first interaction?

📅 Original Article Info

🔗 Further Reading