Zest moves the group dinner decision into the message thread

A shared place to decide

Zest’s restaurant concierge offers dining suggestions and table booking through text messages. The company describes an assistant that can join a group conversation and use participants’ dining histories to help choose a place. Its product page says people can use the service without installing the app.

Our analysis is that this makes coordination part of the product. Finding an appealing restaurant answers one question; getting several people to agree on a specific plan answers another. Keeping that discussion in the existing conversation could reduce the work of carrying options between people and applications.

Dining history and tonight’s plan are different inputs

Zest’s app describes building a restaurant map from linked-card visits and using personal habits and friends’ repeat visits in recommendations. The company says card access is read-only and linking a card is optional. These are product descriptions, rather than findings from an independent test.

A history can suggest what someone often enjoys, but our interpretation is that it cannot settle every detail of a particular evening. Consider a hypothetical group in which one person usually chooses elaborate dinners but has only forty minutes available tonight. Another wants to try a new neighborhood, despite regularly visiting places near home. Their previous choices and present constraints would point in different directions.

A useful assistant would let the group state those exceptions clearly. The important distinction is between an inferred preference and a requirement someone has just supplied. That gives a recommendation an explanation people can inspect and revise.

An appealing option still needs a complete plan

The text service also describes location sharing as optional: a person can name a location instead, or share it for a limited period. That offers different ways to provide context for a request.

In our view, a conversation becomes more useful when the eventual decision is precise. A venue name alone leaves open the date, time and party size. A proposed slot, an attempted reservation and a confirmed table should be distinguishable to everyone making the plan.

For example, a hypothetical group might agree on a restaurant and then discover that its preferred time is unavailable. A helpful next step would preserve the group’s other constraints while offering a changed time or another venue. Restarting the entire discussion would lose much of the convenience the assistant is meant to provide.

Judge the complete outing

Our proposed evaluation follows the whole decision: whether everyone can understand the options, correct an assumption and find the final details. A recommendation that sounds personal is only one part of that experience.

The broader product opportunity is to make a small shared task easier to finish. For a dining assistant, the strongest evidence would connect the conversation to an accurate, understood plan. That is a way to assess the proposition as people use it, rather than a claim about Zest’s current booking success or customer satisfaction.