Contact

Case Study

GenAI pilot demonstrates practical AI adoption in public health operations

InfoPeople designed and deployed a proof-of-concept GenAI-powered knowledge assistant and an RPA automation layer for a large municipal public health agency, demonstrating the real-world applicability of AI in a regulated public sector environment and establishing a foundation for broader adoption.

Share:

Industry: Public sector / Public health

Challenge

Demonstrate the practical applicability of GenAI tools in a public sector environment while solving a concrete operational problem: making agency knowledge accessible through natural language and eliminating manual data workflows that limited AI effectiveness

Solution

RAG-based departmental knowledge assistant built on GenAI and Power Apps, paired with a Power Automate RPA layer enabling real-time data availability and automated workflow execution

Success Highlights

  • Delivered a functioning domain-trained knowledge assistant capable of dynamically incorporating newly uploaded agency documentation into its knowledge base
  • Achieved accurate, relevant responses using agency-specific knowledge across multiple divisions
  • Automated file ingestion workflows, eliminating manual data movement and enabling real-time AI knowledge access
  • Freed staff from repetitive manual tasks, redirecting capacity toward higher-complexity work
  • Established a replicable model for GenAI deployment within regulated government environments

The Challenge

The agency operates across multiple divisions, each producing its own documentation (policies, user manuals, support guides, regulatory updates) in continuous volume. Staff needing information faced the friction of locating, navigating, and cross-referencing documents manually, with no unified interface for querying departmental knowledge.

Compounding this, keeping any AI-powered tool current required manual file transfers and data updates, a repetitive workflow that burdened staff and introduced latency into the knowledge available to automated systems. Leadership sought to test whether GenAI could solve this problem in practice: making agency knowledge immediately accessible through natural language while eliminating the manual overhead required to keep that knowledge current.

Our Approach

AI Knowledge Assistant

We implemented a domain-trained knowledge assistant drawing on source documentation provided by multiple agency divisions. Using Retrieval Augmented Generation (RAG), we enhanced large language model capabilities with the agency's internal knowledge base, enabling the system to respond to staff queries with accurate, domain-specific answers grounded in actual agency documentation rather than general training data.

The assistant was designed to incorporate newly uploaded or updated documents spontaneously into its knowledge base, meaning policy changes, new user manuals, and updated support materials become available without manual retraining cycles. The system operates through Power Apps, integrating AI capabilities within the existing Microsoft environment familiar to agency staff.

Robotic Process Automation: Real-Time Data Pipeline

The effectiveness of the knowledge assistant depended on real-time data availability. To close this gap, we implemented Power Automate to govern file ingestion and data movement, running as scheduled batches or triggering dynamically when targeted files or data met defined conditions. This removed the manual burden of moving files to keep the AI system current, enabling real-time knowledge access while removing dull and repetitive workflows from staff workloads partially or entirely, and freeing up time for more cognitively-involved tasks.

Business Outcomes

  • Unified knowledge access

    Unified knowledge access across agency divisions through a natural language interface grounded in actual agency documentation

  • Dynamic knowledge base

    Dynamic knowledge base that updates as new documents are uploaded, without manual retraining

  • Less manual work

    Eliminated manual file-transfer workflows, freeing staff time for higher-complexity work

  • Real-time AI readiness

    Real-time AI readiness through automated data pipelines ensuring the assistant operates on current information

  • Validated GenAI viability

    Validated public sector GenAI viability, producing a working proof of concept and a replicable model for AI deployment within regulated government environments

Conclusion

This pilot demonstrates what responsible AI adoption looks like in a public sector context: a targeted, well-governed proof of concept that solves concrete operational problems (accessible knowledge, automated data movement, freed staff capacity) while building the institutional evidence and confidence needed to take AI further. For organizations evaluating AI adoption in regulated environments, this engagement offers a proven starting point.