This community innovation has been accepted at the 2026 DHIS2 Annual Conference and will be in abstract track/lightning talk.
Embedding AI Agents in DHIS2: A multicountry pilot
DHIS2 supports routine health information systems, monitoring, evaluation, and learning (MEL), and program decision making across low and middle-income countries. While the platform offers strong configuration and analytics capabilities, key workflows such as metadata creation, paper-to-digital data entry, and data analysis often require specialized technical expertise. These demands can delay reporting, increase administrative burden, and limit engagement with data by non-technical users. To address these challenges, FHI 360 has developed an AI agent-enabled DHIS2 application, currently in an early prototype phase, to automate selected high-effort workflows while remaining fully embedded within the DHIS2 platform. Implemented as a native DHIS2 application using the DHIS2 Application Framework, the solution integrates Azure-hosted AI services and aligns with established DHIS2 standards for governance and security. The application supports task-specific, human-in-the-loop AI-assisted workflows across metadata management, data entry, and analytics. Rather than using a generic conversational interface, the solution employs constrained AI agents guided by DHIS2 metadata schemas, validation rules, and structured domain knowledge. These agents generate DHIS2-aligned metadata from MEL tools for human review and approval, enable rapid digitization of paper-based and electronic forms, and support natural language analytics for summaries, tables, and visualizations directly within DHIS2. The application will be piloted between February and April 2026 in selected countries in Africa and Asia. Evaluation will assess model performance and implementation outcomes, including workflow efficiency gains, uptake across user roles, interaction frequency, perceived trust, and usefulness. Findings will generate transferable design patterns, governance lessons, and evidence on where AI agents add value and where human oversight remains essential, supporting broader responsible AI adoption.
Primary Author: Abumere Ejakhegbe
Keywords:
DHIS2, AI Agents, Workflow Automation,Metadata Management, Natural Language Analytics, MCP
2 Likes