Agentic AI: Empowering Policy Innovation

Session Submission Type: Workshop

Abstract

As the complexity of policy challenges deepens—from climate adaptation to equitable technology access—emerging technologies like agentic AI offer new pathways to enhance resilience, inclusivity, and collaboration in policymaking. Agentic AI, defined by its ability to autonomously plan, communicate, and coordinate via multiple intelligent agents, represents a shift from traditional AI tools toward systems that simulate human-like reasoning and cooperative behavior (e.g., Acharya et al, 2025).

This workshop introduces participants to the core principles and practical implementation of agentic AI, with a focus on how these systems can support policy research and cross-sector collaboration. Through live coding, scenario-based design, and interactive exercises, attendees will explore how to build AI systems that dynamically process complex policy information, assist in multi-stakeholder coordination, and adapt to changing contexts—aligning directly with the conference theme of forging collaborative and transformative policy solutions.

Workshop Objectives:
The workshop will be divided into three interconnected modules, emphasizing the foundational concepts, technical skills, and policy applications of agentic AI:

• Module 1: Foundations of Agentic AI for Policy Innovation
  Participants will explore what makes AI “agentic,” including autonomy, goal-orientation, and multi-agent orchestration. We will cover frameworks such as LangChain, CrewAI, and semantic memory structures, introducing practical examples in policy contexts.


• Module 2: Designing Multi-Agent Workflows for Policy Problems
  Participants will walk through how to design and implement AI agents for tasks such as regulatory mapping, stakeholder coordination, and synthesizing large-scale policy datasets. Real-world use cases will illustrate the benefits of agent collaboration for resilient policymaking.


• Module 3: Hands-on Agentic AI Prototyping with a Policy Lens
  Using Python and open-source tools, participants will prototype a small-scale agentic system tailored to a policy case study (e.g., AI regulation, public health coordination, or climate resilience). We will emphasize transparency, traceability, and stakeholder alignment in system design.

Format and Activities:

1. Live Coding Demonstrations:
   Each module includes guided coding walkthroughs where facilitators build example multi-agent systems from scratch. Participants are encouraged to code along, modify workflows, and test different configurations.


2. Hands-On Exercises:
   Each participant will engage in exercises such as designing agent prompts, planning multi-agent tasks, and interpreting AI outputs in policy scenarios.


3. Collaborative Design Challenges:
   Teams will work together to outline agentic workflows addressing a shared policy challenge. This activity fosters peer learning and reinforces the collaborative nature of both agentic systems and policy solutions.


4. Interactive Q&A and Reflection:
   Embedded checkpoints throughout the workshop will prompt reflection and discussion on practical implementation, ethical considerations, and collaborative potential in applying agentic AI to public problems.

Policy Area

• Workshops

Moderator

• Burhan Ogut, American Institutes for Research

Organizer

• RUHAN CIRCI, American Institutes for Research

Speakers

• Bhashithe Abeysinghe, American Institutes for Research

• Michelle Yin, American Institutes for Research