AI and the charity sector - where to start?

Thanks to Sabita Ravi, Philanthropy Manager at Quartet Community Foundation, for this guest blog on her exploration of utilising generative AI tools in the charity sector.

Most small, grassroots charities aren’t known to be at the forefront of adopting technological innovation. This is often because they don’t have the internal expertise or resources to understand and implement technology-based tools to complement their work.

The same feels true of the adoption of and engagement with AI tools. The AI Governance report 2022 found that 43% of charities/third sector organisations had no knowledge of AI at board level and an equal number didn’t consider it a board issue. These are worrying figures when the world around us seems to be on the verge of an AI revolution. 

It is in this spirit of enquiry and engagement with the AI debate that I set out to read commentary around AI, particularly generative AI, in relation to the voluntary sector and those who fund it. I also thought it useful to experiment a bit myself to see what some of the implications would be for our work at Quartet Community Foundation and that of the hundreds of local grassroots organisations we support.

With no previous experience of generative AI, I started by using Chat GPT for a standard work task – reading requests for support. I made up a fake project and within seconds, Chat GPT produced a simple letter for support that even included a quote from a beneficiary.

The results were certainly astounding but they also raised questions about the accuracy and authenticity of information. Bias and factual inaccuracies within AI tools are well documented and how we guard against these is a live issue.

Based on this experiment, one immediate issue for charities is to think about how generative AI can impact their funding bids and aid and provide evidence for their work. If inaccuracies are present, this will inevitably damage credibility and trust with funding organisations like Quartet that have a responsibility to channel resources wisely and with due regard to the evidence provided in support.

On the flipside though, there is consensus that Large Language Modules (LLM) like Chat GPT can help to streamline labour and language intensive operations like writing funding bids or even governance policies.

Whilst they need customisation, the bulk of the work will likely be made lighter through this tool. There are signs that transcription / transliteration tools will also evolve alongside, reducing barriers for people who currently cannot access or produce information in English (more here: Rhodri Davies: Would AI be good or bad for philanthropy? Will AI replace grant-makers? (thinknpc.org)).

This may lead to greater and better accessibility for our some of our communities than currently exists. It may also benefit volunteer-led community organisations, which struggle with time and skills to write funding requests, by levelling the playing field with their professional paid counterparts.

Another interesting issue to consider is the type of analysis that can be currently offered by LLMs. As an example, when I asked Chat GPT why there was an increase in knife crime in Bristol, it offered a generalised overview of the correlation between crime, poverty, education and opportunity.

When I compare that answer with what we’re hearing from charities working on the ground, it is clear that explaining hyper-local realities and challenges are far above the abilities that LLMs currently possess. This supports our belief that we should continue to prioritise building trusted relationships with local charities and leaders. These relationships enable us to make well-informed and well-timed interventions that are rooted in place and local knowledge. This is not possible (yet) by just reading words off a page. 

However, some AI tools such as this one I discovered (AI that spots inequality could monitor living conditions in cities | New Scientist) offer exciting opportunities for collaboration and triangulation. This in-development tool could enhance how we understand urban inequalities and offer a real-time alternative to that of local and national governments. 

With its use of street imagery, it can help to visually communicate the correlation between issues like localised air pollution and poverty, for example, and begin to attract resources for interventions in a targeted way. As funders, it would provide us with a deeper analysis that we could then act upon, either for targeting resources or engaging with communities via outreach.       

Just by dipping a toe in this area, I’ve found fascinating innovations taking place within AI and we, as part of the voluntary and community sector, are not immune to this. However we approach it, I have no doubt that AI tools will challenge our working assumptions and capabilities and we will need to adapt to the significant shifts that these will bring to our professional lives.

So, what next? For us at Quartet, the time feels right to explore this area further and understand what and how other funders are grappling with AI. For the voluntary sector, already stretched and challenged by growing community needs, the answer might be to join together with leading lights such as Tech4Good South West and start talking about this at senior leadership level.