Case Study

Lenfest AI Collaborative and Fellowship Program: Culinary Compass

An overview of The Minnesota Star Tribune’s AI-powered restaurant guide + open source code

September 19, 2025

Dining out is one of the most consistent drivers of lifestyle traffic for The Minnesota Star Tribune. Readers in the Twin Cities want more than reviews — they want help navigating a sprawling food scene across dozens of neighborhoods, cuisines, and preferences.

Traditional coverage, such as roundups of “best patios” or “kid-friendly spots,” met some of these needs but fell short for readers who wanted answers to specific, everyday questions: Is this place in Longfellow or Uptown? Do they have gluten-free options? Can I bring my kids?

The newsroom set out to build Culinary Compass, a long-term, interactive restaurant guide that could make food and culture coverage more searchable, evergreen, and aligned with how Minnesotans actually choose where to eat.

Product design & development

Approach

Culinary Compass was designed as a cross-functional experiment to move beyond static lists toward an interactive, attribute-driven database.

Process

Discovery workshops: A brainstorming session with around 30 staff members identified the most important audience needs. One surprise: “kid-friendly” emerged as a top local request, reflecting the family-oriented dining habits across the metro area.
Editorial curation: The Food & Culture team identified around 150 restaurants, created new blurbs for each, and manually tagged dozens of attributes including patio, vegan, gluten-friendly, open late, and neighborhood.
AI integration:
- Extracted structured data (e.g., location, cuisine) from years of reviews.
- Embedded articles in a PostgreSQL/PG Vector database to enable personalized search.
- Built a guided “quiz” that matched user preferences to restaurants and generated AI-written explanations for why a place was recommended.
Technical Stack:
- Backend: Python, AWS Lambda, PostgreSQL with PG Vector
- Frontend: React-based interface with guided search
- Workflow: Google Sheets for newsroom input, later imported into the database

Outcome

Culinary Compass launched in mid-2025 with around 150 restaurants and an interactive quiz interface. The project is both a working product and a living prototype. It is currently in beta, but seeing promising engagement with limited promotion.

Key learnings

Local relevance matters: Minnesota audiences wanted guidance tied to neighborhoods (e.g., Longfellow vs. Uptown) and practical attributes (patio, kid-friendly) that aren’t captured in traditional reviews.
Guided search > open search: Like many publishers, the Star Tribune found structured filters and quizzes more approachable than “type anything” AI search boxes.
Balance between AI and manual curation: AI could extract some structured data, but fresh editorial curation was essential to ensure accuracy in a rapidly changing dining scene.

Impact and what’s next

Culinary Compass is both a prototype and a framework:

A technical base (embeddings, quiz matching, interactive UI) that could extend to other lifestyle verticals beyond food.
An editorial model that blends AI efficiency with human oversight, ensuring trust in a space where details change quickly.
Newsroom stakeholders continue to identify additional features.

Why this matters

Culinary Compass is more than a restaurant guide. It represents how local news organizations can use AI to transform lifestyle coverage into durable, audience-first products.

Strengthening local connection: Dining is a universal entry point into community life. By making restaurant coverage searchable by neighborhood and attributes that matter to Minnesotans (like kid-friendly, patio seating, or gluten-free options) the Star Tribune is deepening its relevance to everyday decisions.
Reimagining evergreen coverage: Traditional food roundups expire quickly. By blending AI-powered structuring of archival content with manual editorial updates, Culinary Compass shows a path toward evergreen products that grow more valuable over time.
Balancing innovation and trust: The project underscores an industry-wide truth: AI alone can’t replace editorial judgment. The Star Tribune’s choice to prioritize fresh blurbs and human curation demonstrates how innovation can coexist with the accuracy and trust audiences expect.

In short, Culinary Compass shows how AI-enabled guides can expand the business and civic value of local news while keeping editorial integrity at the center. This project is a piece of the broader Agate AI foundational framework that The Minnesota Star Tribune has and is building designed to surface content in more interactive and contextually relevant ways.