Nonprofits: Get Ready for AI!
By Sarah Di Troia
I am a 5’2” person. In all my years of clothes shopping and wearing, I have never once found myself within the range of “one-size-fits-all.” It may fit most, but it doesn’t work for me.
Retailers, of course, are well aware of the benefits of customization — and not just regarding sizing. If you own store-branded credit cards or shop online, these companies have a tremendous amount of data about your interests and shopping habits. AI tools like predictive analytics, recommendation engines, and virtual assistants have ensured that no retailer has interacted with you in a one-size-fits-all way for more than a decade.
Compare that to the way most community members show up to a nonprofit program. Here, the one-size-fits-all model still reigns. That needs to change.
Nonprofits Need a New Approach
A lot of nonprofits are afraid of AI, either because of its newness or due to justified concerns about bias, something I wrote about in an earlier post. But AI has already transformed the way we interact with most companies — it’s time to start putting these same machine learning applications to work in the nonprofit sector.
However, before they can design an AI pilot, nonprofits need to first build a strong foundation for AI. Towards that end, I have partnered with Project Evident to create a beta AI Readiness Diagnostic based on our AI for equitable outcomes work and research over the past two years.
You can go online, complete the diagnostic, and receive a customized report interpreting your scores and sharing resources to help you strengthen your foundation.
The report will delve into four key aspects of your organization's readiness for AI, including:
Design for Justice and Equity
Strategic Purpose
Knowledge & Skills
Data & Systems
#1. Design for Justice and Equity
Human bias reflected in our culture infuses all data and models, including those that form the basis of AI. The development of AI tools has primarily catered to the for-profit market, optimizing for scalability and profitability and omitting equity as a core design criterion.
It will be challenging to address concerns about equity in an AI design process if your organization does not have a culture and practice of addressing equity and bias in your current operations.
#2. Strategic Purpose
Creating an AI pilot to enhance equitable outcomes is a strategic endeavor. Like any such process, AI pilot plans are best developed when prioritized by leadership and co-created with diverse teams of stakeholders centering the voices of those most affected by the outcome.
Internally, AI pilot development should be supported by technology and measurement, evaluation, and learning staff. But it should be led by the program team because of their lived expertise in program implementation, data collection, and data use in front-line decision-making.
An organization's program logic model is the algorithm at the heart of an AI pilot plan. A detailed program logic model enables staff to understand where their program hypotheses are stronger or weaker, surfacing areas where different types of machine learning applications (recommendation engines, chatbots, natural language processing, precision analytics, etc.) could be applied to achieve an equity-driven mission goal.
#3. Knowledge & Skills
AI can analyze large and diverse datasets to create new insights and decrease the complexity of human decision-making. Importantly, humans are still central to ensuring enhanced equitable outcomes.
Fortunately, AI tools are becoming less expensive and require less data science expertise. However, to prepare for an AI pilot, you still need access to some data skills and a culture of using data to inform innovation and decision-making.
#4. Data & Systems
Organizations need to map, categorize, and store their data to understand their options for a future AI pilot. AI needs longitudinal data for training to recognize patterns; this is achieved by collecting complete data stories across your program logic model, from inputs to outcomes.
By doing this work now, you will not be in a holding pattern needing to collect more data before you can implement a future AI pilot plan to enhance equitable outcomes.
Final Thoughts
AI is neither good nor bad. It’s a tool, and like any tool, how it is used determines the degree to which it poses a threat or adds value.
It is already being employed extensively in the for-profit world. To ensure that their stakeholders reap benefits as well, nonprofit organizations need to understand, embrace, and put this tool to use in ways that benefit their own goals and objectives.