4 Areas Where AI Could Add Value for Nonprofit Finance Departments

Recent research underlines the challenges that artificial intelligence (AI) must overcome if it is to be seen as trustworthy by finance professionals in the nonprofit sector. The study, published by the Journal of Zoology, correlates bigfoot sightings with the size of the black bear population in North America. The study found that for every 5,000 black bears seen, one bigfoot was spotted. For every 1,000 more bears that were spotted, the probability of a bigfoot sighting increased by 4%.

Now we know bigfoots do not exist, but an analogy can be drawn here with current machine learning and large language models (LLMs), which must be trained to recognize objects. How do you train an AI model to tell the difference between a black bear and bigfoot when you have no empirical data to explain what a bigfoot is?

There are several factors to consider when implementing AI tools.

1. Planning and Data Governance

The priority is to understand what problem the AI is trying to solve. Once that is agreed upon, putting the right data governance strategy in place is crucial. If information is siloed or duplicated, it cannot be collated into a single view. 

Consequently, the finance team must take the lead in ensuring data integrity and privacy. Failure to do so could compromise sensitive information, and finance must work with IT to put the right measures in place for reporting and compliance. This is particularly important if a nonprofit is considering using an external LLM (such as ChatGPT) to prevent exposing confidential internal data.

2. Automation Versus AI

Another key consideration is understanding exactly which business processes could benefit from AI. There are many occasions when AI Is positioned as the solution, but robotic process automation (RPA) can already fulfill demand in areas such as donor management. For example,  it can process high volumes of transactions, sales orders or month-end reporting. Where AI can add value is in the analysis of transactions, recognizing patterns and anomalies. One obvious example is identifying inconsistencies, which could help to uncover fraud.

3. Importance of Institutional Memory

The value of AI lies in its ability to visualize and interpret information, but it must be trained to understand patterns. This requires an appreciation of the institutional memory of experienced employees to inform the AI model and avoid bias, as well as understand anomalies. 

If a transaction occurs once a year it might look anomalous to an AI tool, but an experienced finance expert might recognize that, although it is not the norm, it might be a perfectly valid transaction. Losing the institutional memory of seasoned employees could be just as damaging as not developing the new skills to work with AI systems.

4. Curating the Right Datasets

There is no universal standard dataset for nonprofits and, due to sensitivity, these kinds of organizations cannot share data to create a single source of information to train AI models. Often the service or product being delivered is incredibly personalized to the needs of the individual constituent. 

As a result, the datasets could be smaller and more targeted, which will require careful curation. The finance team will need to ensure the AI model understands how valuable the organization’s work is to someone who may be vulnerable or to an animal that is endangered, so that it delivers accurate insights.

While there are particular challenges to adopting AI in nonprofit settings, it is important not to lose sight of its potential to drive innovation and value to an organization. 

Imagine another scenario: A nonprofit focused on the conservation of black bears. Under that overarching goal are various projects to feed the bears, protect their habitats and prevent poaching. Traditionally, the finance team would use financial information to create insights, but this would not provide the complete picture. 

If the nonprofit is to be proactive, it must blend internal and external data. Bears migrate, so it is difficult to know where to buy land or allocate resources to fulfill project requirements. Adding geospatial data, information about weather patterns and tracking data from rangers enriches the picture. Layering AI over all these sources will enable the nonprofit to know more accurately where to invest.

AI will not be required for every process and scenario, but, when deployed effectively, it will become a significant advisory tool in helping the executive team to fortify business strategies. By being able to analyze a broad set of data sources, AI will deliver greater confidence around the impact of decisions and reduce the cost of predicting how best to target resources. Sadly, though, it is very unlikely to ever spot a bigfoot.

The preceding post was provided by an individual unaffiliated with NonProfit PRO. The views expressed within do not directly reflect the thoughts or opinions of NonProfit PRO.