Why nonprofits should deploy artificial intelligence to achieve fundraising success
The International Society for Third Sector Research published a recent paper that summarizes why nonprofit leaders must use technology to connect with stakeholders in the emerging digital, civil society. According to Huiquan Zhou and Shihua Ye, “The rapid development of information technology has boosted nonprofit organizations’ ability to reach out to a larger and more diverse audience… in a fast and affordable way.”
Utilizing the internet, email and social media has substantially expanded the visibility of nonprofit organizations worldwide. In analog communication, individuals generally reach what is called the Dunbar Number, a limit of about 150 stable social connections. In digital communication, an individual can reach a much broader audience. For example, Barack Obama reaches over 100 million people with each tweet. The social network and the increasingly ubiquitous use of these technologies means the transformation of connecting with donors in the future will likely have to be done in some digital form during their development cycle. The 2018 Global Trends in Giving Report shows that, regardless of generation, online is now the preferred channel for donations. Outreach to donors online requires an organization, no matter the size, to utilize some form of technology.
Based on current trends, most nonprofit leaders of mid- to large-sized organizations are coming around to the idea of utilizing technology in their development cycle. For example, Gretchen Buhlig, CEO of the ASU Foundation, made the following comment in a recent interview with me related to technology deployment at the foundation, “We have 1.2 million alumni. There’s absolutely no way that we can connect with that big of a base of alumni to do fundraising without technology.” When an organization already has a large base of constituents or potential constituents, the only way to remain connected is through technology. If Dunbar’s number applied to the ASU Foundation, then for the foundation to maintain a personal connection with each donor, it would need roughly 8,000 development officers; even if all 8,000 were volunteers working from home (excluding costs like salary and overhead), then the foundation would still need a small army of staff to recruit, train and manage this volunteer development team.
AI analytics is a description of a type of machine learning that mines insights within a dataset. AI analytics automates much of what a data analyst would do. Since nonprofit organizations usually have more data about donors than they know what to do with, AI analytics can be an efficient way to discover meaningful patterns within large datasets. According to the AI in Advancement Advisory Council, “Not only do (nonprofits) see a donor's profession and where they live, but also have giving data, understand affinities, know when they like to give, and more.” AI can help nonprofit leaders sift and segment this data into meaningful and actionable content such as customized messaging for individual donors or program updates for a specific population. Buhlig shared how the ASU Foundation is using AI deep learning, “For us to be more efficient and effective at our work, we want to understand what the interests of our donor population... what can we learn from their habits, their consumer patterns…then we can start targeting messaging that’s going to engage them much more.”
Natural-language processing (NLP) refers to the capability of AI to understand the text and the spoken word. NLP is useful for an AI model to identify the intent of a person based on what they say or what they type. Since much of the work in which nonprofits engage results in subjective measures, NLP can be a method to encode the subjectivity qualitatively. For example, the Children’s Society in the UK used an NLP chatbot to deliver fundraising information to potential donors. The Children’s Society is using the generated conversational data to automate “authentic and personalized donor communication.”
Sentiment analysis refers to the capability of AI to analyze and determine the emotion of written text or spoken word. Since nonprofits are in the business of building relationships, the ability to analyze the sentiment of stakeholders at a large scale would inform organizations about how effective their programs, services, or fundraising efforts are perceived. For example, a group interested in understanding the sentiment towards diabetes on social media completed an analysis that discovered some immediate results that may “be relevant for developing better public health strategies and for promoting a positive and constructive attitude” about diabetes online.
The AAAC reports that innovations like AI address a major fundraising concern about “donor fatigue, (and) reaching deeper into the donor pyramid with less staff.” As organizations have transitioned from a gift officer to a digital gift officer, they have had to reinvent their fundraising model to be a digital-first approach. According to Zhou, “Organizations now need to update their stakeholders, preferably in real-time with multimedia, about their performance…to ger more donations an organization needs to disclose more information (and) network with the public on social media.”
Leadership and management recommendations
Nonprofit leaders have an opportunity to expand the use of artificial intelligence within their fundraising efforts. It is recommended that leaders consider building AI expertise in-house, use data analytics to segment donor groups with the express purpose of customizing content, and collect rich, labeled data using their donor conversations to train models for conversational AI in email, phone, text and online chatbots.
In-house AI expertise
The progression from a development officer to a digital one must continue to evolve to include an AI development officer. Organizations should build infrastructure and staffing around internal expertise with AI. The capability of AI is too valuable and important to the future success of connecting with donors in the digital civil society for organizations to outsource it. (Imagine if an organization never hired or developed an in-house officer that knew how to use email.) McKinsey’s data supports this idea. Organizations that can utilize AI to address real organizational needs and create real value “tend to have the ability to develop AI solutions in-house—as opposed to purchasing solutions—and they typically employ more AI-related talent, such as data engineers, data architects, and translators.” By building internal AI expertise, organizations can leverage that expertise to increase their organizational efficiency and expand their outreach efforts exponentially.
Donor segmentation to customize development content
AI gives organizations a powerful way to analyze data at scale. Using data analytics to segment donors is where machine learning is deployed to identify patterns in the data to determine operational decisions. Donors may be part of many kinds of segments. For example, a donor might be a female, age 36-40, residing in the state of New York. Each of these data points could be used individually or collectively to discover relationships to how that profile behaves. The goal of donor segmentation is to identify behavioral relationships to allow an organization to make operational decisions. The strategy that will be most beneficial to organizations is a strategy that will help organizations engage with donors in a more meaningful way. Strategic questions about why donors have engaged, are engaging, or may engage with the organization should drive the data analytics queries. A recent case study by Dataro.io with Parkinson’s UK shows that when an organization implements custom content for donor segments, it will translate to a much higher net return on the appeal when compared to no donor segmentation. Nonprofit leaders utilizing AI to segment donors must remember that donor engagement is king when determining the types of queries to run on their datasets. Leaders should think strategically about what kind of potential patterns may exist to train AI machine learning.
Collect rich, labeled data to train conversation models
The most effective method to create clean, domain-specific, labeled data is to collect the data from actual conversations between the organization and the donor. For example, to collect conversational data by analyzing email text with Email Tree AI to identify key donor intents and corresponding appropriate responses. Or, utilizing an application like ProtoCall AI to label phone, text or chatbot conversations to not only collect the data, but to also define through data analytics the data about the best response to increasing the probability of donations. When an organization leverages its existing interactions to generate actionable data for its AI models, it creates a perpetual optimization loop to become increasingly better at communicating with donors at scale in the digital space.
The digital civil society is continuously expanding and the methods of connecting with stakeholders in this digital space allow organizations to scale their outreach like never before. The omnichannel communication of email, text, social media and other online platforms continue to increase as the preferred methods of outreach, leaving nonprofits to go where the people are in order to remain relevant and valuable. Most organizations have already embraced some form of these outreach channels and have taken steps to institute AI within their organization. A challenge for nonprofits is that the expertise to further AI utilization within their organizations is scarce, but the benefits are multi-fold. An in-house expert will be better able to collaborate with other organizational entities, have a better understanding of the mission and challenges, and have the patience to iterate on the AI implementations. It is also recommended that as an organization does analysis on its donor base data, that the primary strategy should be to identify means to engage with donors in a more meaningful way. The demographic, psychographic, behavioral and geographic data available on donors is only helpful if it can tell an organization how to better connect more effectively with donors. Nonprofit leaders that adopt and deploy conversational AI will achieve future fundraising success by effectively reaching more individuals to reverse the alarming trend of few individual donors.
Nathan Bezzant is a 2021 graduate of the Master of Nonprofit Leadership and Management program at Arizona State University. For the past 20 years he has worked in the marketing and technology industries, and currently works for a software company that builds applications to help organizations in both the for-profit and nonprofit sectors to communicate with their stakeholders through technology.