A Governed, Citizen-Facing AI Grants Agent

Applied & Generative AI for the Public Sector: Building on its modernization of the agency's grant pre-application process, Triverus delivered a governed, citizen-facing AI agent that gives applicants real-time program-fit guidance, with auditability and human validation built in.

The Challenge

The agency awards hundreds of grants each year through a consolidated, multi-program grant portal. Its pre-application process required applicants to log in, work through a complex form, and then wait up to 15 weeks for hand-reviewed feedback.

Triverus first helped modernize the process through automation, improving operational efficiency and reducing manual effort. The agency then set out to reimagine the constituent experience by moving to a conversational AI agent.

Why it matters: Applicants now get clear, real-time guidance on which programs fit their needs instead of waiting on staff review. The work shows how a state agency can put generative AI in front of the public responsibly, and it arrived as follow-on to a multi-year modernization, with each phase of delivery earning the next.

How We Helped

1. Modernized the pre-application platform first. Replaced the login-gated form with open, accessible intake and automated feedback, cutting turnaround from 15 weeks to 2 days.
2. Matured it as a multi-year partner. Refined the platform each grant cycle, building the knowledge that later showed where AI could help.
3. Added a governed AI agent. Built a conversational agent on AWS Bedrock that recommends programs, summarizes eligibility, and submits a pre-application on the applicant's behalf.
4. Built in auditability and human validation. A persistent audit log captures each conversation and the basis for the agent's determination, with human-validation points defined before action.
5. Held to state standards. Designed on the state's digital design system and aligned to enterprise accessibility and security standards.

Results

• 15 weeks to 2 days for pre-application feedback after the initial platform modernization
• Real-time program-fit guidance for applicants, replacing a multi-day review process
• Governed AI agent on AWS Bedrock with a persistent conversation audit log
• Human-validation points defined before release for higher-consequence outputs
• Built to state standards for design, accessibility, and security
• AI delivered as follow-on to a multi-year modernization partnership

Capabilities & Technology

• Capabilities: Applied & generative AI for the public sector · Responsible AI governance (auditability and human-in-the-loop validation) · AI built into low-code modernization and integration · Accessible, citizen-facing digital services · Cloud & serverless architecture · Multi-year managed delivery
• Technology: Microsoft Power Platform (Power Pages, Power Apps, Power Automate, Dataverse) · AWS Bedrock AgentCore & Strands Agents · Anthropic Claude Sonnet · AWS Lambda, API Gateway, DynamoDB, S3 · React · TypeScript

‍