Scaling Social Impact with AI

The social impact sector is quickly moving past basic AI adoption efforts to more robustly embracing of AI. Nonprofits and government agencies who have seen early success are considering what their next steps should be. Funders and investors are giving increasing thought to scaling AI social innovations.  Indeed, one of the major philanthropic trends over the last 20 years is an emphasis on nonprofits "going to scale." Funders and ambitious nonprofit leaders focus on demonstrating what works within one locality in order to build the case for an organization to deliver the program in multiple regions. AI is opening new opportunities of how social impact organizations can think about scale (or in many cases nor consider doing so). I posit two frames from which to think about scaling social impact with AI:

1. Only a limited number of social impact organizations should consider scale efforts, even with new opportunities that AI presents. For the majority of social impact work, collaboration and learning across organizations is going to be more important for effective and ethical AI than scaling AI-driven programs.
2. AI presents significantly different models for scale than the typical organizations scaling models of the past. Those models require new forms of investment, new capabilities, and new operating models for the organizations that choose to pursue them. 

Scale Ain't for Everyone

Not every nonprofit is suited for scaling, and that’s okay. At times over the last two decades, the philanthropic sector has pressured organizations to scale that were either not ready or not really up to the task of scale. I used to joke that nonprofit executive directors with ambitions to scale would present at the Harvard Business School Social Enterprise conference. There would be a metaphorical laying of hands on the executive director and the nonprofit organization would receive the necessary funds to scale up. Joke aside, the philanthropic impulse to value scale was clearly on trend for much of the first two decades of this century. 

The pressure to scale has often overshadowed the importance of deep, localized impact. A countervailing philanthropic trend to focus on deeper community engagement gained more steam with national reckoning following the murder of George Floyd put more emphasis on organizations led by people of color who had developed meaningful, deeper ties within communities. These organizations of and representing marginalized communities have vital roles to play in ethical AI development that are just as important if not more so than any scaling efforts. I will return to this role momentarily

Across all nonprofit organizations, the majority will not be in a position to scale. In my recent post, AI Adoption In Social Impact Sector: What Do We Mean?, I noted that we are seeing 3 different levels of AI adoption in social impact organizations:

1. Consumer Use (40-60% of Organizations): Use of consumer AI tools to help with personal and team productivity engaging generative AI vendors, AI within established applications, and increasingly pre-built AI agents.   
2. Advanced Use (25-40% of Organizations): Use of advanced AI skills such as prompt engineering with fine tuning and RAG to build out models for specific use cases on top of existing Large (and Small) Language Models and increasingly build out Agents on top of existing AI tools and applications.
3. Builders (10-15% of Organizations): Purpose-built models and productization of AI tools that will be used by other social impact organizations.

The consumer users of AI will not scale since they are not building models or AI tools, but rather using the tools others build. Some advanced users organizations will scale through new methods of sharing their AI tools and models. Many of the builders can and should scale, but they will need to develop new capabilities around digital product management and marketing that most are organizations are not yet position to do.

AI Models for Scaling Impact

AI introduces new pathways for scaling impact that diverge from traditional methods.  The traditional notion of scale—replicating a successful model in one region across multiple geographies—has its limitations. AI doesn’t necessarily make this approach obsolete, but it does open doors to alternatives that prioritize collaboration, innovation, and leveraging existing resources. These pathways redefine what it means to scale, often reducing the need for physical expansion while increasing reach and efficiency. Here are three primary models:

1. Traditional Organizational Scaling – Least Effective for AI

This model involves replicating a successful program or initiative across multiple locations. While this approach can work in certain contexts, it often requires substantial resources and infrastructure, which can strain smaller organizations.

AI tools can assist traditional scaling efforts by automating processes, enhancing data collection, and improving decision-making. However, the limitations of this approach remain significant, particularly when it comes to adapting programs to new contexts and ensuring equitable access to technology. 

2. Model Sharing – Most Collaborative

AI Models and well developed prompting is the new program AI makes it easier than ever to share tools, data, and methodologies across organizations. The field needs platforms and infrastructure to deliver on this approach by enabling nonprofits to upload, share, and refine AI models collaboratively.

Instead of every organization reinventing the wheel, model sharing allows for the co-creation and dissemination of solutions. For example, an AI model designed to predict food insecurity trends in one region can be adapted and applied in another. The local organization with deep ties to marginalized communities can further refine models based on the feedback of community members.  For example, the food insecurity model may perform different with rural and urban communities or across different cultures and national contexts.  The collective efforts engaging locally can improve overall models. By fostering collaboration, this approach amplifies collective impact without requiring individual organizations to scale their operations.

3. Productizing – Most Scalable

Productizing involves turning a program or service into a standardized, scalable digital product or platform. AI can transform a localized intervention into a digital tool that reaches thousands—or even millions—of people. 

For instance, an AI-powered mental health chatbot developed for a specific community can be refined into a widely available app. This model offers the highest potential for scaling impact, but it also demands significant investment in technology development, user experience, and ongoing maintenance. Social impact organizations without product development, management, and marketing capabilities will need to develop those capabilities to deliver on this kind of scale.

Indeed, productizing requires organizations to adopt new operating models that emphasize agility, cross-functional teams, and continuous improvement. Product management is not the same as program management and required a whole different organizational design and significant shifts to org culture to operate with agile product development practices.  While this approach can unlock unparalleled scale, other challenges should be considered such as pricing the digital AI products at prices that not only support sustainability but allow for consumer AI use and less resourced organizations to take advantage of the AI social impact product.  

Collaboration as the Cornerstone of AI-Driven Scale

The common thread across the AI-driven models is the importance of collaboration. AI’s potential to drive social impact isn’t limited to individual organizations; it lies in fostering networks, sharing resources, and building collective intelligence. In model sharing, organizations are working together.  In the product model, the suppliers of AI driven products are continuously listening and incorporating the feedback of their customer base (other social impact organizations).

For funders, this means supporting initiatives and platforms that prioritize open collaboration for model sharing. It means investing in a new capabilities that will allow for digital AI product driven scale. For nonprofits, it means rethinking their role within broader ecosystems and embracing partnerships and new ways of being that take advantage of the AI opportunity

The Future of Scaling with AI

As AI continues to evolve, the way we think about scaling social impact must also change. Traditional models focused on geographic replication are giving way to approaches that leverage AI’s ability to process data, automate tasks, and create digital solutions. Instead of every organization aiming to scale, AI encourages a shift toward shared platforms, interoperability, and ecosystem-level change. This shift is particularly important for ethical and effective use of AI in the social impact space.

The future of scale is about ecosystems, not silos. It’s about sharing knowledge, building platforms, and creating scalable tools that transcend organizational boundaries. By embracing these new paradigms, we can ensure that AI serves as a force for good, driving meaningful and sustainable change on a global scale.