AWS Public Sector Blog
City of Virginia Beach launches AI-powered search assistant to transform citizen access to information
The City of Virginia Beach, Virginia, serves over 450,000 residents year-round, and that number doubles during peak tourist season. The city’s public website is where most go first to pay property taxes, find parking near the oceanfront, or check event schedules. For years, finding answers meant navigating content scattered across 32 departments, third-party sites, and hundreds of PDFs. Traditional keyword search helped, but users often needed to know where to look before they could find what they needed.
In December 2025, Virginia Beach changed that. Working with Amazon Web Services (AWS) and AWS Partner Allwyn, the city launched a generative artificial intelligence (AI)-powered conversational search assistant on its public website that helps residents and visitors find answers in plain language, without having to hunt through search results. In its first month, the tool handled more than 1,300 queries, and adoption continues to grow, demonstrating how quickly AI-powered search can deliver value for public sector organizations.
Simplifying information access across a complex digital environment
When generative AI started gaining traction, the city saw interest from across the organization. But Tracy Lyles, enterprise architect at the city, and her team wanted an enterprise-grade solution that would make life easier for residents, not something flashy that wouldn’t hold up.
The search assistant also fit into a broader strategy. In late 2024, the city identified over 68 potential AI use cases across operations, including an initiative to modernize non-emergency citizen services with Amazon Connect. With that many possibilities on the table, the team had to be deliberate about where to start. The search assistant became a priority because it addressed a clear pain point while establishing a scalable foundation that could expand to additional departments and data sources over time.
Working with AWS and AWS Partner Allwyn
With the goal defined, the city needed collaborators to help turn the strategy into reality. Virginia Beach already had teams working in AWS, which made it a natural fit. The city was also drawn to the capabilities of Amazon Kendra and Amazon Bedrock for intelligent search and generative AI, which could work together to understand questions and return meaningful answers.
AWS led the architectural strategy and technical oversight, guiding AWS Partner Allwyn’s implementation and integration efforts. Weekly meetings brought together Virginia Beach’s IT staff, Allwyn’s implementation team, and AWS architects to address questions and maintain momentum. If a technical issue arose midweek, someone was always available to help.
For Lyles, that close collaboration made all the difference. “What I loved about the entire process was that we had the technical expertise there,” she said. “AWS acted as an advocate for us.”
Discipline mattered, too. Karthik Samala, a generative AI specialist at AWS who worked on the project, noted that many organizations struggle to complete AI projects because they’re constantly chasing new technologies. “Virginia Beach stayed the course,” said Samala. “That’s where most companies don’t complete these projects. They hop off that course and try to adopt new technologies but can’t finish what they started.”
For Anant Mittal, solutions architect at AWS, the weekly working sessions weren’t just about architecture, but impact. “An AI assistant becomes more than code,” Mittal said. “It tells the citizen, ‘You matter enough that we’ll respond, even at midnight.’ It quietly builds trust, one conversation at a time.”
Building on a flexible technical foundation
When a user types a question into the city website, the request flows through Amazon CloudFront, then to an AWS Lambda function that processes the query. Amazon Kendra retrieves relevant information from the city’s indexed knowledge base. Claude by Anthropic in Amazon Bedrock synthesizes a response grounded in that retrieved content, and Amazon DynamoDB stores conversation history to support follow-up questions.
The city chose Claude for its cost, the team’s familiarity with the model, and its strong reasoning capabilities. Amazon Bedrock’s flexibility proved valuable during testing, as the team could experiment with different models and switch between them to troubleshoot issues or compare performance. Amazon Bedrock Guardrails keep responses on topic and grounded in the city’s content. And because all interactions are logged, the team can review what users are asking, spot patterns, and refine the assistant over time.
Coordination, testing, and managing expectations
The project took about seven months from concept to launch, a notable milestone for any municipal AI initiative. Much of the work involved coordination across 30+ departments, each with its own content and way of organizing information. The team gathered requirements, identified points of contact, and conducted extensive testing to make sure the assistant could handle that variety.
Testing ultimately revealed something unexpected. Generative AI behaves differently from traditional software. “With the assistant, we had to look at the content it was providing to make sure it was accurate,” said Jonathan Bolden, systems analyst at the city. “You’re not going to get a word-for-word response that’s the same every time.” That meant educating testers across departments about what to expect and how to evaluate responses.
With that in mind, the city held back on major promotion at launch. The team communicated first through internal channels and social media, giving themselves time to monitor interactions and gather feedback before pushing the tool more widely.
Early assistant adoption shows promise
The assistant launched in early December 2025, right as residents headed into the holiday season. Still, the team saw steady growth: about 1,300 queries in its first month, with usage climbing. Feedback has been constructive, and the team is refining the assistant based on what they’re learning. Some city employees have become regular users themselves. “I’ve been using it since we deployed it, and it’s been great,” said Bolden. “I was surprised by how often I use it to find an answer.”
For residents, the benefit is time. Instead of digging around webpages or calling a city office and waiting on hold for a routine question, they can type it into the assistant and get an instant answer. Over time, the city expects this to reduce call volume and free staff for more complex requests.
Looking ahead, the city plans to expand the assistant’s capabilities in phases, starting with additional departments, then new data sources, such as PDFs stored in Amazon Simple Storage Service (Amazon S3). Translation features and deeper analytics could follow later. And with those 68 AI use cases on the roadmap, the search assistant is just the beginning.
Lessons for other organizations considering AI-powered search
These challenges aren’t unique to municipal government. Any organization with information scattered across departments, users who don’t know where to look, and staff overwhelmed by routine inquiries could benefit from a similar approach.
For those considering an AI-powered assistant, Lyles and Bolden offered these key takeaways:
- Start with clean, well-structured data. Recognize that response quality depends on source content quality.
- Invest in staff education. Help staff and users understand how generative AI differs from traditional search.
- Control scope carefully. Focus on delivering the initial solution before expanding.
- Prioritize technical collaboration. Encourage regular working sessions among internal teams, implementation partners, and AWS.
- Use guardrails from day one. Configure Amazon Bedrock Guardrails to keep responses appropriate and on topic for the audience, particularly government.
- Think beyond the first use case. Build a foundation that can expand to additional departments and data sources with minimal rework.
Virginia Beach followed that playbook, which paid off. What started as a way to help residents find parking or pay a bill has become a proof point for how the city can use AI to serve its community. With a deliberate approach, strong collaborations, and clear focus, Virginia Beach turned a seven-month project into a foundation for what comes next.
To learn how AWS can help your agency build AI-powered citizen services, contact the AWS Public Sector team.