Unleashing AI’s Potential in Campaigns and Organizing: Lessons from the Front Lines

By Kate Gage and Oluwakemi Oso, Cooperative Impact Lab

Artificial intelligence is rapidly transforming many industries, and advocacy, campaigns, & organizing are no exception. Since 2018, we have guided over a hundred organizations through experimenting with emerging technologies at Cooperative Impact Lab. Over the last year, we’ve focused on helping non-profit organizations and campaigns experiment with generative AI to enhance their work and further their missions while maintaining a clear understanding of the ethics and risks inherent to the technologies and specific to our communities.

Using a cohort model, CIL worked with thirteen orgs last year ahead of the 2023 election, and we are currently running a program with twenty-one movement and organizing non-profits from across the social impact space.

Over the next few weeks, we will release a series of case studies and resources from the 2023–2024 program, with a focus on lifting up insights crucial for any organization considering leveraging AI.

Read our first AI Case Studies here: Analyzing Canvassing Conversations with Fair Count in Mississippi & AI Case Study: MoveOn’s AI Approval Process

Key Themes and Learnings from our 2023 Cohort:

1. The Importance of Organizational AI Readiness

Successful AI adoption heavily depends on organizational readiness factors. The groups in our first cohort that made the most progress had dedicated staff focused on experimentation, an existing culture of innovation, buy-in from leadership, and technical sophistication among staff. Interestingly, alignment with specific programmatic goals or strategies that might seem more ripe for AI involvement, like social media generation or a need for text generation, was less predictive of success than these organizational capacity elements. If an organization was interested, invested, and willing to experiment, we could usually find a way for AI to support its strategic goals, often with free, cheap, or widely available tools.

Many organizations are driven by an interest in avoiding being “left behind”—they know that if they don’t keep up with the changing technology landscape, not just their organization but the communities they serve will not be considered or centered in this new technological era.

2. AI Unlocks New Capabilities, not Just Efficiency

The mainstream narrative surrounding the impact of generative AI focuses on efficiency — specifically metrics of time saved and increased productivity. However, the most impactful AI applications observed throughout our cohort went beyond efficiency gains to enable fundamentally new capabilities aligned with strategic objectives. For example, Fair Count, a CIL Cohort Member, used AI to analyze unstructured data from canvasser voice memos, extracting insights about voter interests and barriers to voting. This innovative approach allowed them to gather richer field data, identify emerging trends among potential voters, and even identify counties that were going to have lower voter turnout than expected.

When it came to applications of generative AI, we saw a wide range of experimentation. Many organizations started using AI for a singular purpose, such as content generation for drafting social media posts in different audience voices or quickly summarizing lengthy legislation for advocacy purposes. As cohort members went through CIL’s coaching & training program, they increasingly understood the strengths & weaknesses of the technologies. This enabled them to move onto more advanced use cases including data analysis or using AI to glean insights from unstructured data sources such as public comments, YouTube videos, and canvasser voice memos. One example is Red Wine and Blue’s work to scrape YouTube comments for keywords to improve their efforts to combat misinformation. Other cohort members used AI to extract nuanced stories from public submissions, act as an always-on thought partner for content creation, summarize verbose legislation into digestible talking points, or even build chatbots to collect public comments or provide information to constituents. Across the board, the most impactful applications were those that didn’t just increase efficiency but unlocked fundamentally new capabilities.

3. Experimentation Drives Adoption

Another key learning was the catalytic role of hands-on experimentation in driving adoption. While many organizations are AI-curious, actually rolling up their sleeves and testing the models and tools is critical for identifying the highest-value opportunities. Training and support, whether through a formal program like ours or driven by an internal champion, were crucial for adoption. Once trained, we saw tremendous creativity and ideas flourish.

In our experience, even just 30 minutes of showing examples of how AI has been used by other orgs, answering questions about data privacy & ethics, or showing brief tool demos & prompts, leaves leaders with new ideas and a curiosity to experiment.

The most valuable way to communicate AI’s potential is for staff and leaders to see concrete examples of how organizations like theirs are using AI. In addition to publishing case studies from our cohort members, CIL has developed an AI Prompts, Workflows, and Tool Tracker specific to campaigns and organizing so every team member can see how AI can help them in their daily work.

As the case study about Fair Count’s analysis of unstructured canvasser voice memos demonstrates, teams must be prepared to try a variety of prompts, tools, and workflows. This is especially important as the tool landscape, models, and technologies change almost daily.

4. Responsibility and Governance are Essential

Further practical resources, such as AI policy templates, model analysis, and usage guides, need to be developed to support organizations achieve organizational readiness. This will ensure that staff and leaders are informed enough to know how to responsibly leverage the technology. In our experience, organizational leaders want to “do it right”, but need support to understand what that means.

While there are a plethora of resources available for large-scale companies and daily work, the sensitive nature of data related to community organizing, and electoral campaigns requires specific guardrails and ethical guidelines. Developing these types of assets for the field is essential work that merits further investment. Opportunities exist to support organizations through leadership training, policy workshops, shared research, tool recommendations, and hands-on assistance.

You can read about how AI Cohort member MoveOn has built a system for review and approval of AI usage, and find their policy in our AI Case Study: MoveOn’s AI Approval Process.

5. Challenges and Opportunities

The 2023 CIL AI Cohort surfaced several other core challenges and opportunities for organizations interested in AI that we’ll be exploring over the coming weeks. These include the importance of human involvement and oversight — or “human-in-the-loop,” the overwhelm of the rapidly evolving tool landscape, the fear of violating platforms’ always-changing terms of services, and the need for political-specific or open-source tools for highly specific actions. We have seen a clear demand for training and coaching, resources like organizational policies and approval processes, landscape analysis and more.

At CIL, we’re responding in a few ways:

• Over the next few weeks, we’ll be releasing a series of case studies from our 2023–2024 program on topics like Organizational AI policy development, unstructured data analysis to improve public comment collection, AI for social media content creation, and using AI for translation. You can find our first AI Case Study here: Analyzing Canvassing Conversations with Fair Count in Mississippi
• We’ll also be releasing resources for campaigns and organizing in partnership with organizations like AI Impact Lab, Higher Ground Labs’ Progressive AI Lab, and Zinc Labs
• We have developed a beta version of an AI Prompts and Workflows Library specifically tailored for campaigns and organizing. We will be building this out further and are looking for partners to keep building it out. Have something you’ve used AI for you’d recommend to others? Submit it to the database here!
• We are running a second cohort of 21 non-profit organizations who we’re working with 1:1 to identify how AI can support their work and already finding new opportunities for each one. If you’re interested in this type of support for you or your grantees, let us know!
• We also offer one-time 1:1 coaching sessions to talk through potential ways for orgs to leverage AI. These are usually arranged by a funder for their grantees, and we’re always interested in expanding these opportunities.
• In January, we co-hosted an in-person AI Summit for Campaigns and Organizing with Zinc Labs and Higher Ground Labs— you can find the presentations here — and we’re planning more in-person opportunities now, to be announced soon.

As the AI landscape evolves at breakneck speed, ongoing dialogue and collaboration between organizations, campaigns, technologists, infrastructure orgs, and industry are key to staying ahead — any one organization is not going to be able to fill all the gaps. By learning together, sharing insights and resources, and supporting one another, we can make sure that our communities are not just passive bystanders of this groundbreaking technology, but are at the center of policy, impact, and new applications.

Through experimentation and collaboration, the advocacy community can collectively navigate the challenges and opportunities of this powerful technology. Together, we can shape a future where AI empowers organizations to make an even greater difference in the lives of those they serve.

Want to stay in touch with CIL? Send us a note or join our mailing list here to stay up to date with events, resources, and our work.

Thank you to all our cohort members and our partners: Zinc Labs, Shamash Global, Trestle Collaborative, AI Impact Lab, Higher Ground Labs, Aya Umoh, Tudor Mihailescu, Dr. Rayid Ghani, Dr. Phil Resnik, Andrew Gamino-Cheong, and especially Han Wang for their work and support on this project.