HarliBot

01 The Problem

The City of Harlingen receives hundreds of resident inquiries every week — questions about permit status, utility billing, trash pickup schedules, permit applications, and more. The city's call center was handling these requests manually, creating bottlenecks, after-hours gaps, and inconsistent answers.

A significant portion of Harlingen's population is Spanish-speaking, and existing self-service tools were English-only. The city needed a solution that could handle routine inquiries around the clock, in both languages, without requiring staff time for every interaction.

02 The Approach

We designed HarliBot as a bilingual NLP-powered chatbot integrated with the city's existing service data. The core architecture used Dialogflow CX for intent classification and conversation management, with custom Python handlers on AWS Lambda for dynamic data lookups and business logic.

Language detection was handled automatically — the bot responds in whichever language the resident initiates the conversation in. We trained intent models on real resident inquiry data to ensure the classifier performed on the actual vocabulary and phrasing patterns of Harlingen residents, not generic training examples.

The React frontend was embedded directly into the city's existing website, requiring no separate app download or account creation for residents.

03 The Stack

• Dialogflow CX — intent recognition, bilingual conversation flows, entity extraction
• AWS Lambda + API Gateway — serverless fulfillment layer for dynamic lookups
• Python — NLP preprocessing, custom intent logic, data integration
• React — embeddable chat widget deployed on the city's website

04 The Outcome

HarliBot successfully handles routine resident inquiries in both English and Spanish around the clock, reducing the volume of repetitive calls reaching the city's call center. Residents get immediate, consistent answers without navigating phone trees or waiting for business hours.

The bilingual capability was a meaningful equity improvement — Spanish-speaking residents gained the same self-service access as English speakers for the first time.