This month we’re very grateful to Dan Sutch from CAST for talking to us about how funders are experimenting with Artificial Intelligence (AI), and how they’re responding to other people using it. 

  

  

Dan: We’re a ten-year-old charity set up to build the digital, data and design capability of the VCSE sector. We have a particular focus of working at the intersection of grantmakers, charities and digital experts. Even ten years ago we realised that if charities were to use technologies in new ways, funders would have to work in new ways too. These days about three quarters of our work is about AI.

  

  

Dan: The history of CAST is partly a history of helping funders respond to new technology. In 2023 we saw the huge uptake of ChatGPT and saw that nobody was responding to this in a coherent way within the sector. We’d already run sessions for funders on other types of technology, so invited various funders to an online call that is now called the ‘AI for Grantmakers Peer Group’. It started as 30 people in the first video call, and it’s now 430 people! It’s not slowing down.

  

  

We mapped all the major tasks that grantmakers routinely carry out, like “monitoring grants”. We then used this list of tasks to cluster together similar AI experiments going on inside different funders. For example under “monitoring grants” we have identified a bunch of different funders all trying to use AI to generate more useful findings from monitoring reports.

  

There are now so many AI experiments going on that you can probably name almost any aspect of grantmaking work and there will be someone who has been experimenting with the role of AI. 

  

However, that’s not a very strategic way of looking at things. We encourage funders in our community to realise that AI might impact on them in very divergent different ways. For example as well as AI making it quicker or easier to make grants, it might also exacerbate a social problem that a funder is seeking to resolve. In short AI is both an operational challenge for funders, and also a strategic one.

  

  

Some of the examples I can share might not seem that novel because they are AI uses that people in all sorts of organisations are carrying out, not just funders. For example there’s loads of people using AI notetakers to take notes of calls with grantees or applicants, or exploring a new topic using deep research.

  

But then we’re also seeing some activities that are definitely grantmaker-specific. So one funder ran an experiment to see if they could replace traditional written grant reports by instead asking grantees for a learning conversation. These calls were then recorded and AI used to transcribe what was said. They’ve not made this permanent because of concerns about data protection and privacy, but this is typical of a thoughtful experiment in our community.

  

Another funder had a 100 grant application long list, which they manually triaged to produce a 20 application short list. That process didn’t use AI at all, but when it was over they asked an AI bot to try to carry out the same shortlisting process - it picked 19 of the 20 same proposals as the humans.

  

And in one funder board, they loaded a set of previous board minutes into an AI, and used it as a way of quickly accessing what had been said and decided at previous meetings as a way of maintaining institutional knowledge.

  

Most of the experiments we see are totally internal, and grantseekers wouldn’t know about them unless the funder chose to publish something. This isn’t about being secretive - just that most experiments are small scale and tackling internal challenges. However, in one rare public outing for AI at a grantmaker, we saw one grantmaking organisation trial a tool on their website that allowed applicants to determine whether or not they might be eligible for funding in a conversation with an AI bot. 

  

  

A really huge number of people who work for funding organisations are having a go, kicking the tires or running experiments. The speed of growth of the AI for Grantmakers peer group points to that.

  

However what we are seeing are large numbers of small experiments which are normally championed by one or two motivated individuals within funders - what we are not yet seeing is major changes of structure or process that are led or signed off by leadership. This is something in charities as well as funders - lots of individual experiments, but far less aligned organisational approaches. This seems to be one of the biggest challenges for the year ahead - aligning activities across organisations. 

  

It’s also time for funders to start looking closely for signs that AI is starting to change the parts of the wider world that they try to influence. If you fund youth skills you should be on top of the data about AI and employment. If you fund the rights of women and girls, you need to know how AI enables new forms of exploitation and abuse. 

  

And, of course, you need a strategic response to help you cope if AI means the number of funding proposals you get starts doubling every 12 months. The new head of Innovate UK, Tom Adeyoola, has already announced plans to rebuild that organisation’s entire grantmaking system to cope with the AI era of applications.

  

  

We often hear funders in our community say “We don’t use AI to make decisions”, as a clear declaration of taking a cautious and responsible stance. And that certainly seems a reasonable precaution at this still very early stage. 

  

But I have started to realise that when funders tell us they aren’t using AI to make grant decisions, they are mainly talking about just the final ‘Fund or don’t fund’ decision. But real grantmaking processes are full of smaller and more subtle decisions, and we’re starting to see some funders using AI to make these without necessarily being aware that they are basically letting AI make some sorts of decisions.

  

For example if I use an AI to summarise a long grant monitoring report into a shorter summary, which I pass on to my board of decision makers, I am actually letting the AI decide what to include and exclude, unless I carefully review it myself. So this is a key place to be cautious - where AI might be influencing the reputation of a grantee in ways that we’d be wise to study carefully.

  

  

Well, there’s quite a few things! CAST offers a whole series of courses and peer groups you can join, which are available here. Ultimately though, right now it’s really all about working out how to conduct your own learning experiments which are safe and educational. I think this is easiest if you form a peer group, perhaps of funders similar to you. And do check out our AI experiments library, detailed what various funders have been making public about their own AI experiments.  

 

 

Latest Reading, Watching & Listening - Modern Grantmaking recommends   

  

"I know," says the AI. "But I wanted to demonstrate impact."