Reflection on Using AI to Augment Systems Change Practices Training
Written by: Lihn Doan, The Convive Collective
Engaging with the session, Using AI to Augment Systems Change Practices, offered a pause and reflect opportunity to my day-to-day work as a researcher in philanthropy. Like many of us in this space, I enter this work driven by deep personal commitments to equity, justice, and human-centered sensing and want to tackle complex social problems at their root. Yet, this session made it glaringly clear to me that if I am not deeply intentional, my own uncritical adoption of artificial intelligence can silently erode those very principles, deepening structural inequalities and introducing serious ethical breaches into my research practices.
The Danger of Unintentional AI Use
Unintentional AI adoption often creeps in through a desire for efficiency. innocently enough: using generative AI tools to summarize meeting notes, draft reports faster, or process community feedback quickly. However, relying on AI without acute awareness introduces severe systemic risks:
The Inequality Multiplier: AI can easily widen the gap between well-resourced institutions and frontline communities. Educated researchers and large global funders use AI to become faster and smarter, while less-resourced grassroots actors risk having their lived experience replaced by automated proxies.
Distorting Systemic Realities: AI models default to agreeable, smooth jargon trained largely on dominant Western and institutional viewpoints. In my own experience, because most large language models are trained on internet-scale data, they inherently risk amplifying pre-existing stereotypes, systemic bias, and structural blind spots. When used passively, AI reinforces funder-centered success stories while completely missing the invisible structural drivers of change.
Ethical and Data Privacy Breaches: Feeding sensitive qualitative data or community transcripts into unsecure AI platforms creates significant privacy risks and violates community trust.
Moving Toward Intentional AI to Augment Systems Change
To avoid these traps, the session outlined a practical path for using AI as a constrained instrument rather than a trusted thought partner:
Augment, Do Not Replace: Focus AI strictly on tasks that exceed human working memory, such as synthesizing massive datasets across timeframes and complex system actors. Keep empathy, relationship-building, and final decision-making exclusively in human hands.
Sufficiency Over Efficiency: Shift from using AI merely to reduce short-term human effort toward applying it only where it adds genuine analytical value to complex systemic problems.
Guiding Reasoning with Frameworks: Active direction is required. Direct AI using specific analytical tools, such as pre-mortems, systems defenders, or alternative worldviews, to challenge cognitive biases, push back on groupthink, and expose blind spots.
Use “AI in the mix”: Reframe away from traditional "human-in-the-loop" structures (which center the AI) and keep human decision-makers firmly at the core. Establish clear consent protocols, ensure strict data privacy, and plan AI usage collaboratively with project partners so no one is caught off guard by how insights are generated.
Is AI like a dessert?
One of my favorite exercises from the session was selecting a metaphor to define our relationship with AI. For me, using AI in research feels like eating dessert.
The day humans invented dessert must have been exciting for everyone—adding a welcome touch of sweetness after a meal. It started as simple cookies and eventually evolved into elaborate, highly sophisticated dishes. But can dessert ever replace your actual meal? Absolutely not.
In systemic research, our "main meal" is the hard, grounded work of human listening, relationship building, empathy, and contextual sense-making. AI is merely the sweet extra—a specialized treat that adds rapid synthesis or temporary flair. Relying on it as our primary sustenance, or consuming too many convenient AI treats over time, will silently destroy our analytical health and ethical integrity.
Core Takeaway
This session reinforced that AI is an unreliable instrument that requires firm, explicit boundaries. By treating AI not as an automated decision-maker but as a tool to disrupt our own assumptions, we can leverage its analytical scale to elevate our research without sacrificing the human relationships, community trust, and ethical integrity at the heart of true systems change.
AI Disclosure Statement:
In keeping with the practice of transparency discussed in the session, I declare that an AI assistant was used to assist in editing and polishing the phrasing and structure of this written reflection.
Appendix: Key Session Tools & Frameworks Reference
1. Custom-Trained Chatbots: https://www.policysolve.com/ai-training
Indigenous Systems Thinking Chatbot: Guides reflective conversations through two Indigenous frames: the Mari Te Ruru approach and Melanie Goodchild’s First Nations Thinking.
Western Systems Thinking Chatbot: Trained on traditional Western systems change frameworks, specifically Donella Meadows’ leverage points and the FSG / Water of Systems Change framework.
2. Custom Strategic Protocols
Pre-Mortem Tool: Simulates potential strategic failure points so teams can plan backward to mitigate risks.
Devil's Advocate Tool: Evaluates strategies through opposing stakeholder lenses and opponent logic.
Systems Defender Tool: Analyzes how existing systems and institutions naturally defend against change, even without active opposition.