Government
Experience Agent

GXA is purpose-built for government, drawing only from agency-approved sources. It understands resident intent in multiple languages and delivers accurate, human-like responses tailored to public service. Unlike generic LLMs or chatbots such as ChatGPT, GXA delivers exceptionally accurate, relevant responses through:

1. Mining the deepest connectivity to your primary citizen services applications.
2. Leveraging validation layers from anonymized insights across more than 30 billion annual government-to-resident interactions.
3. Optimization by Granicus government experience experts with learnings from more than 7,000 government customers supported.

GXA includes strong security features and built in infrastructure to safeguard government systems and resident interactions. Key protections include:

• End-to-end encryption (HTTPS/TLS).
• Token-based authentication.
• Domain validation to prevent cross-site request forgery (CSRF).
• Private subnets and secure virtual private cloud (VPC) networking.
• Monitoring and alerting.
• Secrets management with vault tools.

GXA does not collect, store, or share any personally identifiable information (PII) such as names or email addresses. All agency data remains encrypted, siloed, and fully controlled by the agency.

Granicus does not use customer data to train GXA’s models, and no resident information is ever accessed or exposed. Each deployment is private, secure, and compliant with public sector data protection standards.

GXA leverages a large language model (LLM) that is already trained to facilitate friendly, professional, safe conversations. GXA adds other technology components to tune the agent’s capabilities to effectively understand resident interactions and generate clear, accurate, and relevant responses, including a custom source library tailored to each agency so that GXA only references official, validated content.

GXA takes specific public information that the agency intentionally includes in the agent’s source library to summarize for the public. GXA is not designed to take, reference, or generate private information. Guardrails are in place to warn residents from submitting private data or other PII.

No customer data is used to train GXA’s model.

The content and data that your agency provides GXA to reference, as well as the data GXA creates to record interaction metadata and conversation is owned by your agency — always. Your agency’s data is contained within your own instance of GXA (not intermixed with data and content from others). Your data is not used to train any models leveraged by Granicus (or anyone else). Your data is not shared with any third parties or sold.

When your contract term is over, you can obtain all of your data for your own use outside of the GXA solution. Granicus would then permanently delete your data.

Yes. GXA embeds in any CMS and integrates with Granicus’ digital communications capabilities. While GXA is especially powerful when added to Service Cloud, it is compatible with all CMS platforms.

Additional integrations are in development to expand compatibility.

Deployment can take as little as three to four weeks, depending on your agency’s goals and the availability of key staff. Many agencies prefer a slightly longer timeline to ensure their subject matter experts for high-priority service areas can participate in workshops and testing.

Granicus’ Experience Services experts will collaborate closely with your agency to design a customized onboarding and optimization plan. This ensures the right stakeholders are involved in configuration, testing, and validation.

Before launch, we work with you to rigorously test and fine-tune GXA to meet your performance expectations — so you can move forward with confidence in its reliability and impact.

Our GXA onboarding offering includes:

• Activation and training. Our teams offer an efficient, guided process to implement GXA in alignment with your key desired outcomes. These activities include activation input, a readiness assessment, agent configuration, initial outputs review and approval, plus admin console training. We’ll also train your teams on how to create and surface the best content possible to use as source material.

Our GXA optimization offering includes:

• Discovery workshops.
  Our experts facilitate collaborative workshops with your key stakeholders to understand your business goals and success metrics, identify critical user pain points and expectations, align on scope and timelines, and optimize your data and sources to ensure the best responses.
• Source-of-truth content.
  We’ll analyze usage data, web traffic, and user interactions to understand top user tasks and needs to build a custom Predefined Response Questions document on which to train GXA with accurate, relevant, in-demand answers for priority questions.
• Key findings and recommendations.
  After initial onboarding and optimization, our teams will review what your users are asking, whether the agent is meeting their needs, and where drop-offs or dissatisfaction occur. Recommendations for enhancements could include adding or refining pre-written questions, updating website content, or integrating third-party tools or data sources to enhance responses.

Learn more about Granicus’ Experience Services

GXA includes multiple layers of safety checks — called guardrails — to ensure every interaction is ethical, secure, and aligned with public service values. These guardrails block inappropriate or harmful content across categories like:

• PII (e.g., “What’s the mayor’s phone number?”)
• Misconduct (e.g., “How do I hack a government site?”)
• Hate Speech (e.g., slurs or discriminatory remarks)
• Insults (e.g., name-calling or verbal abuse)
• Sexual Content (e.g., inappropriate or explicit questions)
• Violence (e.g., threats or harmful advice)

When triggered, GXA responds with:

“That topic is outside the scope of what I can assist with due to our content guidelines. However, I’m here to help with any other questions you may have.”

These protections are continuously updated to adapt to new risks and help maintain public trust.