Artificial Intelligence (AI) has recently been gathering a lot of attention as the technology to watch. Andrew Ng, a leading expert and researcher in the AI space whose career includes stints at Google and Baidu, has recently proclaimed AI is ‘the new electricity’. The improvement in the accuracy of modern techniques, especially deep learning, as well as the ease of access to AI computing infrastructure via platforms such as Amazon Web Services and Microsoft Azure, means that more and more organisations are able to access AI technology.
However, one of the biggest challenges is understanding AI technology, where it can be used and how to use it. For non profits, its all well and good that the latest AI solutions can recognise objects from video, create artwork and predict traffic patterns, but understanding where to use it in their organisation is a far cry from a lot of what is happening in the market. AI is not remotely at a level where you can explain a problem to it and it will figure out the solution – the technology is still quite specific in its application and requires human intelligence to decide which specific model (or combination of models) to apply.
Based on our experience and some of the common challenges we see with our clients, here are some of our suggestions on where AI could benefit NFPs, ranked from easiest to implement to more exploratory (but potentially higher impact).
Automating customer service with chatbots
Chat bots have been around for over a decade, however in recent years the improvement and ubiquity of Natural Language Processing (NLP) algorithms have made it much simpler to create a useful chat bot. There are a range of chatbot platforms such as Botsify, Google Dialogflow and Amazon Lex, however most are relatively similar in that they let you create a conversational tree with a series of intents, with the ability to extend the platform with custom APIs if you need. Basically, its like creating a script, and AI will try and match the best answer in your script to the question.
Whilst large-scale case studies are relatively rare, IBM is predicting 85% of all customer interactions will be handled without a human agent, which means streamlining of service interactions across many different industries and a general expectation to be able to interact with an organisation in this way.
For NFPs, chat bots offer a way to streamline common interactions and answers to enquiries, as well as some specialised interactions (e.g. donations, adoptions, bookings). Instead of requiring staff to answer basic questions such as services, availability, locations etc, the chatbot is capable of answering these questions in a person’s stead. They can also be extended to answer more unique questions (e.g. animals available for adoption) with a bit of additional technical work. Whilst chat bots will not replace a human interaction for a more complex query or interaction, they can save valuable staff and volunteer time in answering basic and repetitive questions, and can be extended to perform medium-complexity tasks.
Whilst donation analysis has traditionally been primarily expert opinion and some statistics, there is certainly space to solve some of the more intricate challenges with AI. There are a range of techniques that can be applied for clustering and classification, such as SVMs, kNNs and random forests, and is certainly an area open for further explanation.
There is certainly some initial research and interest in this area. Salesforce has published an article on AI extensions on their platform, and there has been some scholarly research into this area on the use of machine learning techniques for higher education donation. However, from our research these are part of a set of limited examples that are either broad suggestions and very-specific cases – large scale rollouts and industry best practice is still being explored and refined.
Based on some of the problems we have seen, we certainly believe there are a few specific problems in the area of donations and fundraising AI techniques can help with:
- Review the best candidates to become donors from a cold or warm data set (e.g. past Alumni for a university or school)
- Predicting candidates for regular donors based on previous interactions (e.g. one time donors)
- Classifying regular donors into groups and maximising their engagement and retention (
- Classifying and re-engaging lapsed donors
These examples all depend on the size of the dataset, however for most medium to large NFPs the dataset should be of a sufficient size to be able to apply some machine learning techniques.
Allocating resources in service delivery and volunteering
In NFPs that deliver service interactions and need to prioritise specific resources, or engage specific supporters, AI can offer a faster, automated way to prioritise resources.
Two examples from other industries and their potential applications are:
- Predictive policing – where do you best put your resources from the data you have available? Obviously, with policing you are using arrest and crime data, but if you are delivering a support services such as homelessness or social support, where do you best put your resources in response to capacity and time-based problems.
- Finding blood donors – in this example an NGO used public data to predict which people were likely to match different blood groups and be willing to donate. An example of how this might apply more broadly is matching existing supporters to volunteer roles that they may do well at and enjoy – cross-propagating supporters into different areas of the organisation.
As the work that non profit organisations is quite broad the problems and techniques used to solve them are quite different depending on the use case. However, one thing that is common is that finding and allocating resources can be significantly streamlined with the use of AI.
Getting started with AI in your organisation
Just as electricity did not solve all the world’s problems, neither will AI come in and solve every problem there is for NFPs. However, with AI now being a lot more accessible and comparatively easy-to-use than in the past, there are many opportunities for AI to streamline non profit operations and provide additional insight and structure that was not possible before.
We expect that over the next few years non-profits that wish to be leaders will find effective ways to streamline a few key areas of their organisations through digital technology and AI, giving them more time and scope to focus on strategic work and better service delivery. As the sector is still relatively early in its use of AI, our recommendation is to start with a relatively small MVP with a well defined and specific problem (e.g. frontline customer service), and then expand into other areas of the organisation over the next few years as experience and internal support is built.
* Photo by Alex Knight.