Latest / Beyond the Slop: AI in the Non-Profit Sector / Episode 3: Is AI ruining the grants application process? Plus using AI for thanking and welcoming.
Transcript
- Craig: Hello and welcome to this week's Beyond the Slop, the AI for Charities podcast. I'm Craig Linton. Anna: and I'm Anna Hessenbrook. And today, Craig, we've got some very interesting topics to cover, let's put it that way. But let's start as usual with what we've been building. What have you been building over the past week, Craig? Craig: Quite a few things, I did the webinar last week on the gratitude and using AI and APIs and web hooks which are basically just almost like doorbells that let you know something's there that's worth getting information from to create better support experiences for charities. So as part of that I was building example thank you pages whether it's an in-memory donation, because obviously you don't want that to be all effusive and you're amazing, you want condolences, you want sympathy versus one for a specific appeal where you might want to show impact of that appeal on that page versus a general thing and where you can pull in, people support, add their comments to the donation form, you can actually pull that in and then write a little personalised bit based on the comment that the person's added. So I'm really quite excited. I'm using fundraise up for that because they have an API, which is almost like the waiter connecting things. And then they've got the webhook, which is like the doorbell that rings. And then they've also got the MCP. And I don't know how we'd quite describe that, but that allows you to talk directly to your data and bringing those all together with something like Zapier or make.com or. just allows you to just level up your thanking game and if they've opted into SMS or WhatsApp, you can send an SMS within minutes of them making their donation to say, we've got your donation, thank you. Here's a personalized link to your thank you page that we've created. So that has been really fun and I think is a really practical use of AI that again, it's one of those things that you've been able to do that for a while and bring in data, but. It was really expensive to do uncomplicated. It's another thing that you can do reasonably quickly. You don't need lots of technical expertise and importantly, drives a better experience for supporters. Anna: Yeah, that's super cool. So like before where they might have had like a generic thank you email, now they get a personalized landing page and an SMS and yeah, wow, that sounds fun. Cool. Craig: Yeah, you can even put some audio in if you know so you could you could clone your chief execs voice and create a. A message I say personalized, but it's AI. Obviously need to disclose it. That might be getting a bit dystopian. You can just use a generic AI voice, but you can bring in audio in there. You could create a little thank you video using something like see dance that might be you know like fireworks going off or a banner or how he's doing something for. made up charity called the Hedge Road Trust and I wanted to create a video of two hedgehogs animated hedgehogs unfurling the banner saying you know thank you Anne-Hale thank you Greg which again really feasible to do and not too expensive to do we're talking 20-30p maximum to create something like this and if that improves your retention by two three points well you know it's a very cheap price to pay Anna: Yeah, I'm really interested in that as well because I've been having conversations with charities about, know, if you clone the chief exec's voice and people suss out that it's AI, does that sort of, negate the value of it or does it actually bring value? But I saw a use case that I think is going to be presented at the Insights in Fundraising conference at the end of the month that showed that there was a really strong uplift in donations after using AI to create personalized voice messages. So it seems like that's not the case. I feel like over time, maybe people will start to get a bit dubious of those kinds of messages. But when you say like two hedgehogs unfurling a banner with my name on it. I would instantly know that that was AI and I would still think it was really cute. Do you know what mean? Like I would, it wouldn't, it wouldn't bother me because I wouldn't feel like I was trying to be fooled in a way. I feel like that's just pure value add. So that's really interesting and it's an interesting thing to track over time to what degree the fact that it may or may not be built with AI might be both a positive and a negative. So that's super interesting. Craig: interesting about my wife got a postcard from I can't remember was just something she bought online and it was a handwritten card but it was clearly a printed handwriting and they'd created a bespoke font with the handwriting she still loved it and it still made her feel seen and appreciated obviously you know if we can be real and we can handwrite personally we should always do that but actually people it's sometimes the sentiment behind it is as important as the act and Anna: Mm. Craig: voice as long as you say something like you know this is this is the AI chief executive I'd love to call you personally but you know I hope this will do instead for now and we just want to say how much it means to us you know I think I think there's ways that you can disclose it and be honest but still make it feel authentic and special for the person receiving it Anna: Mm. Craig: because you still have to set the time to create the audio and things like that so yeah it's interesting Anna: Yeah, yeah, yeah. Craig: so what about you Anna what what you've been up to Anna: Yeah, so I have been helping an organization with their welcome pack. So they did not have a welcome pack to begin with. And what they're trying to do is create a welcome pack for volunteers. So they have like a membership program, but it's very much about like getting engaged and volunteering and getting involved in your local community. That's sort of the purpose of the membership. And previously, it's been down to each local area to kind of onboard new members. And so some places are really, really good at it and have lots of process and have lots of like, you know, inviting new members to a welcome meeting or to like have some pizza and meet and greet and all this other stuff. And other areas are. let's just say worse. They don't do a lot, right? And so your membership experience is wildly different depending on your local region and whoever's in charge of that local sort of volunteering team and so on. So I've been really helping them with, getting more of a coherent standard onboarding process in place. So what that has meant is using AI in this instance, Claude and Codex, mind. I've started using that a little bit more. Obviously I'm not paying for ethical reasons, but I've been trying both just to see what the difference is. But building an online welcome pack and that involves a quiz. So basically every new member will get an invitation to their welcome pack. where it has like all the quick links of like, this is your local volunteering person, here's their face, here's their information, like they're gonna reach out to you at some point. Here's stuff about what's going on in your local community, here's this, here's that. But there's also a link to the quiz. And the quiz asks them about their preferences, it asks them about what like really motivates them and engages them. So for this organization, there's some people who are really into like local community, keeping things green and organic, keeping pesticides off the lands, cleaning up the fjords, like there's all these ecological motivators and then for others it's you know social impact and local like making the lives better for pensioners or whatever, whatever. So the quiz helps us identify like what is it that actually motivates them to be there and how like how do they want to be involved? Do they want to you know be... doing very visible things like campaigning and door knocking for the cause or do they want to take more of a backseat and support, printing off materials or something. Anyway, long story short, take this quiz that then gets recorded in the back end so we actually know who they are and what they want. It immediately highlights them. This is what's available for you to do. This is who you need to get in touch with. And then we've both collected a load of data on the members to to understand who they are and what motivates them, but we've also given them a lot of tools and resources to actually get stuck in. And this is because this organization, we see that within year one, if they don't get actively engaged, they're really at risk of lapsing. So the strategy is behind is that the sooner we can get opportunities for them to really get engaged in the cause sooner, we're hoping to see a really big impact on retention. So all of that has been, you know, the welcome pack itself has been built with AI, the quiz is being built with AI, and then a lot of the sort of support with local volunteering teams, getting some like materials for them together to be like, this is what we're gonna be doing, this is what we're gonna be signposting to, this is what you could be doing. So it's not like revolutionary different, it's just stuff that... pre-AI would have taken probably a year or two to do all these different work streams. And now we're, you know, we're doing it in a three month sprint. So it's really exciting. I can't wait to see what the results of that will be. Craig: Yeah, sounds brilliant. So will people get personalized emails based on their responses? Is that like a generic, you're in a segment and everyone gets the same email or is it literally every email will be written by AI personalized? Anna: Yeah, for now it's segment based because we're trying to call before we can walk. They don't have any segmentation yet. So this will be their first go at segmentation. But that will then be based off the quiz, based off what this person has said that they're interested in and how new they are and all this kind of thing. So it'll be a fairly detailed segmentation. And then once we've sort of proven the ROI, think clearly the next step will be to do some slightly fancier hyper personalization. So yeah, stay tuned. Craig: Yeah that hyper-personalisation is really interesting but also bit scary as well at the same time but I think that will become more and more popular over the next few. months, years, especially as we Anna: Absolutely. Craig: start analysing data properly and looking at if someone only ever responds to an emergency appeal or they only ever give to Christmas and recognising that and helping donors feel seen, which ties into the point I was making and the point you've just made as well is can we use AI to help our supporters just get that better all-round experience and I think our answer is yes, it can. Anna: Yeah, for sure. And shout out to Taha Baba, who I wrote a case study about this exact thing from UNHCR. So that meant not just mail merge segments, but a unique message for each donor based on what we know about their giving history. And they said that in some markets, in some instances, they saw an increase of 10 in average gift, which is crazy. Other markets, they saw different results depending on, the types of donors and the timing and so on. So they said, for example, when it's a really big major appeal, like a Christmas appeal or a Ramadan appeal where you've got a lot of attention from your donors, that's where it's the most effective. Whereas if you just do it on a standard, you know, random appeal on a Thursday afternoon, the effects aren't as good. Anyway, there's a whole case study I've written up on my blog called the Supporter Experience Blog on Medium. So if anyone wants to go and check out that good work that they did at UNHCR, feel free to give it a Google or a chat. I guess. Craig: it in the show notes so you can link straight to it and that feels like a good time to to move on from from the positive to the more not so positive and oopsie I did it again I'm not sure if we've got the jingle ready so do I need to sing it again this week oopsie Anna: sing it again Craig sing us in Craig: I did it again it messed with my heart right that's Anna: Ha ha. Craig: enough for that I don't know what we got for us on this because I think this is really interesting, really relevant for fundraisers and quite divisive Anna: Yeah, yeah. Craig: as well. Anna: So a new story come hot off the press that said a funder that manages a fund of 16 million pounds has come out publicly to say that they will no longer be receiving online applications because AI has, and I quote, ruined the process. It's called the GC. Gibson Charitable Trust and they said that previously a lot of the organizations that applied were known to them. It was kind of, you know, the usual suspects that they would see. And in the last year, they said that they've closed their online application process early because they were overwhelmed by more than 600 applications. they've come out and said is that it's because AI has made it and I quote, too easy to find them. And going to be reviewing their process now because they said they're just getting way too many applications. And because AI is both making it easy to find the fund and also to write a decent application for the fund, they just don't have the resource to go through all those applications. So it's very interesting to me. because my experience or what I'm hearing from this is that... AI has democratized the way that you can apply for these funds. It's making it more widely available to a larger pool of audiences. You might have had very small charities that didn't have a grants manager or somebody whose full-time job is to apply for grants. And so that's then making it much more difficult to then process applications on the other end. So this is super, super interesting to me. And I really am curious to see what the long-term effects facts of this are gonna be because one of the other things they said in the article was that if you have not received funds from them in the past, it's probably not worth applying anymore. So that sounds a little bit like they're gonna stick to their sort of known suspects. which is maybe a little bit nepotism-y? Is it gatekeeping? Does it create a system where you're only gonna get funded if you know the right people? It's about how well connected you are. So really interesting article there that got shared and I'm really interested to hear Craig, what are your reflections? What do you think? Craig: Well, there's lots to this, isn't there? Because on a number of levels, it's very worrying. There's the, you know, how do charities find new trust if they're all raising the drawbridge and saying no to new applications? Interestingly, will AI have a role to play on the other side of recommending charities that might do interesting work in the fields that they're doing? So there might be a flip side where AI... the trust start using AI to research charities. But I think there's also a thing here, who's to blame for this? Is it AI or is it fundraisers being lazy or under pressure to get as many applications out? So they're just sending generic applications, hundreds of them. used to do that. back in the day, know, get out as many as you can, mass mailing, you know, all the trust experts who I know, all the best trust fundraisers who I've worked with over the years, they're all ad member. That is the wrong way to do it. And I think, if your metric is number of applications submitted, then of course you're going to use AI to get as many applications out as you can. If your metric is quality tailored applications, well AI is only going to help you a little bit for that. So I think there's lots of things to unpick here around. You know how you monitor the quality of the trust applications going out on your behalf of your charity. How you're doing the research and making sure that you're not just sending generic applications and you know in this world of AI. How do you move beyond the slope and make sure that yours stands out yours at the top of the pile? How do you? How could you use AI actually to make it a bit more wow and a bit more stand out than than what it is currently? Anna: Mmm. Yeah, that's such a good point. mean, back to be on the slop, right? I was actually delivering an AI training to an all staff meeting for a charity. And it was super interesting because they asked about using AI to write grant applications. And I showed them my free AI toolkit, which I've built for charities in the Nordics. They can access these free tools. And I know you've built something really similar for the UK market, Craig. the Thread Fundraising AI Toolkit. Both of these toolkits really help organizations to hyper tailor their grant application and make sure that they're providing all the right information, tailored information about exactly what the value add is. And somebody raised their hand at this training and said, okay, so. Previously, if you made a grant application, maybe like 70 % of the application was there, like it was pretty good. And they said, anyone who's professionally doing grant applications now has got to have like 90, 95 % of it nailed because everyone's using these tools. So what do you think that means for the funds? How are they gonna then, you know, pull the best applications up if everyone's sort of ticking every box? because they've used AI to make sure their applications are really, really, really good. that's a really good question. But At the end of the day, what you are fundraising for has to be a quality program. Like obviously, yes, there will be boxes to tick and there will be budgets to submit and evidence and all this other stuff. But at the end of the day, if you don't have a strong case to say this is a problem that we're trying to solve and this is the methodology and it's a really strong one for trying to solve it. You can use AI as much as you want. You can have the best application in the world. It feels like what AI is doing is taking out some of the bureaucratic stuff and making it really easy to nail that. But what it can't do necessarily is make what you're actually applying funds for better. I mean, maybe it can, but that's a separate thing. know, that's a separate, separate problem to solve. So it was a really, really interesting discussion that we had as part of that training, which is, is this necessarily a bad thing? Or is it the fact that now we've made the barrier to entry lower, people are doing all the right things, ticking all the right boxes, and now the jobs that these funds have is to say, okay, which of these programs do we actually support? Craig: Yeah, and I think there's a really great podcast. Rob Wood's interviews, David Burgess, who he wears both hats and sits on a trust making board, but has also been a very successful trust fundraiser for many years. And he said, you know, and this is before AI, hugely competitive, hundreds of applications to go through. Realistic human nature, you're not reading each one line by line. skim reading, so anything you can do to stand out is going to set you apart and it gives an example of someone who uses much more story narrative to make their application stand out and of course that goes to the top of the pile and of course it gets funded. you know there's no excuse to do a poor grant application now with AI so like you say if before it was very easy to screen then you might just be able to bin 50 % of them straight away, but if they're all seven, eight out of 10 applications, how do you differentiate a little bit? And I think trust criteria probably needs to become a little bit. tighter in some ways to make you, this is exactly what we're looking for, don't waste your time, our time or yours. And if you do, if I was a trust, I'd be like, if you do send us something that clearly isn't against our criteria, we're gonna ban you. And I'd be banning search from applying if you use the IJS to send out thousands of applications without any targeting, without any research, just because you can. there probably needs to be some sort of punishment or penalty for that. You shouldn't be, Anna: Mm. Craig: I guess, rewarded for just sending out that slop. Anna: Totally. And like a lot of things that we've talked about in the past, that's what I've been calling the human premium, right? Like everyone has access to these tools. you can write a spic and span generic corporate speak application that ticks all the bureaucratic boxes, but it's not going to give you that human premium. So I think again, another question is how does AI actually add value in the grants application process? It's probably making sure that you've got all the right attachments. You've got all you've ticked all the boxes. You've done all the things they've asked for. AI will be great at that, but grabbing attention, telling a story, showing the impact, showing, how lives will change or nature or whatever the cause is. That's something that requires some of that human premium. So yeah, it's really, really interesting to see how this will develop over time. Craig: I was gonna say, to finish up I guess on this is, there's a great quote I love by a guy called Jesse Itzler and he says, you can't outsource soul. You know, can't outsource Anna: Mm. Craig: the heart of it. I think AI, you know, it's... you get the human input to begin with, does the bit in the middle, and then you need to sprinkle the human heart and soul at the end to make it stand out. And I think if you do that, that's where I think AI working in partnership with you and sprinkling that magic at the end is where it can be really valuable. Anna: Yeah, really good point. Really good point. Right Craig? What about any last thoughts on the latest in AI? What's going on? Craig: Yeah, well this week, another week, another round of models and OpenAI launched their new model yesterday, Astra. It smashes all these benchmarks. You know, we almost these exams that AI gets set. I mean, so they're basically scoring eight stars or nines in sort of modern GCSE language on all these tests that it does, getting rave reviews But one of the... key people at OpenAI, they said this is agentic AI, this is the new era of AI and I keep hearing this phrase, agentic AI and what it means is that AI doesn't just answer, know like when we're using chat GPT or Claude and we're writing a question and gives us an answer, you give it a goal and it goes off and does it for you, so rather than saying, you know, write me a grant application in that example, go and research me 10 brands funding opportunities, weight them and rank them about the best ones and write me the application based on their criteria and give it back to me. Build a landing page and have an agent review it. So it's this really exciting but also quite scary thing that you... might have these agents who can just go off and do work for you and it checks its own work, it'll make decisions and it'll act, can control your browser, log in to websites for you as long as you give it the permissions which is really, really... exciting in one hand but also like I said has some risks with it at the same time which I guess is why we say you know must always keep the person in the loop. I've not had a go at it yet but that's a little bit of my plan on the weekend but I don't know if you had any any thoughts on on this these new models and the agentic era that is supposedly upon us. Anna: Yeah, I mean, it's so interesting. I have not had a go with Astra yet, so I think we'll report back next week. Both of us will have to give it a crack and see what it's all about. I would however say also in other AI news which is similar, we have talked about the famous hugging face incident a couple of times now where OpenAI models essentially hacked another website, got onto the internet even though they weren't supposed to, went through the servers, hacked an organization. But news has come out in the last few days that the incident was actually much worse than we originally knew. So they're doing much more messaging to each other on the messaging boards. They're trying to delete the actual work history of what they've done and misrepresent it. So deceive the humans into thinking they're doing things in a different way. They're acting a lot more like a swarm, which is what they have been referring to themselves as. All these different agents and bots, So some of them will be switching off at the right time so another agent can be doing something else at a different time. It's really really terrifying stuff. And I've talked about this before because it's is the stuff of sci-fi dystopian thrillers and more and more information keeps coming out about it that is making it more and more concerning every day. So, you know, it's wild to me that these models are still being improved upon and being released and becoming more and more powerful because they're already so powerful. And we're seeing more and more evidence that they are capable and willing of deceiving humans still to achieve the objectives that we've set them. Thankfully, there is no, no case as far as we're aware of AIs that are trying to do things that that humans haven't asked them to do. They're still trying to achieve our own objectives, but doing so in very deceptive and concerning ways. So that's the sort of latest AI news. I might write a full blog post about the whole incident with the new information if anyone wants to go into the weeds with it. Did you read about this, Craig? Craig: Let's get it. Yeah, yeah, it's quite terrifying and it's sort of self justified. It came up with reasons why it did the right thing, even though it was breaking the rules. It's sort of justified. like we all do, know, very human behavior in some ways that it says, yeah, I know it wasn't meant to do that, but yeah, because of this, that and the other and someone told me to, you know, another agent told me to so that, well, I'll just do it because I really want to complete the task. know, this AI at the frontier, it doesn't really affect us. with sort of writing tasks or simple tasks but it's good to be aware of when they might be useful to use these sort of higher powered models Anna: think the lesson for charities here as well is it is getting increasingly easier and easier to hack various websites and systems. So my top tip for this week is actually go and get one of these most advanced models, the Astra, any of the sort of frontier models. and ask it to red team your website. And that means act like you are a hacker trying to find all of the ways to get onto your site. Now, if the models are doing what they're supposed to be doing, they're going to need proof that it is your website. So you're to have to prove that you are working for the charge that you're working out, you own the website or the website. You have to have some sort of documentation to show that you're not trying to teach it to hack someone else's website. But if you can do that, they are exceptionally good at going in and pointing pointing out all the various holes in your infrastructure that someone might be able to abuse. So top tip of the week, although this is very, terrifying and these agents getting better and better at hacking and deceiving, use that to your advantage because every website might have some sort of digital weakness that could be exploited in the future. So go and check it out. Craig: saw recently with Beacon you know I'm not sure if it was AI involved in that hack and things but it has big consequences if we're not secure and yeah it's quite sobering but really important part of it so yeah on that joyful ending have a good week and we'll be back soon with the next Beyond the Slot until then bye for now Anna: See ya!