Latest / Beyond the Slop: AI in the Non-Profit Sector / Episode 5: The first AI donation, AI & corporate partnerships, and the 'right to speak to a human?'
Transcript
- Craig: Hello and welcome to this week's Beyond the Slop. I'm Craig Linton. Anna: I'm Anna Hessenbrook and this week we are talking about the very first all AI completely automated donation has been reported by the Trussell Trust. So we'll be unpacking what that means for the sector. And on the flip side, we've got Citizens Advice talking about the right to reach a human in customer service and what that might mean for the future of supported care. Craig, before we talk about that, what have you been building this week? Craig: First of all, think we've got a celebration on our hands. We had over 100th listener this week. know, big thank you to everyone who's Anna: Woohoo! Craig: listened. next stop, 1,000. And then, you know, keep adding and all, as my old boss used to say about fundraising. So yeah, big thank Anna: you Craig: you to everyone who's been listening. That really made our day and we've had some lovely comments. So thank you. We really enjoyed seeing things on LinkedIn and a couple of emails we've had. So please do send stuff in. Yeah building this week there's been a lot really. I'm quite excited because there's another new model. We seem to say this every episode at the minute but there's this new Anna: Hmm? Craig: anthropic of being cooking to use the parlance of the tech bros and a released Opus and Sonnet 5.5 which I think what's quite interesting and exciting is these are faster and cheaper and so I was using this fable that we've talked about before that was great but you just burn through your credits and your tokens and every week I was getting up to 80 90 percent usage rate this week I looked earlier my usage resets tomorrow as we speak and I'm on like 23 percent for the week which is crazy and I've done probably more this week than other things and one of the things I've been building is a demo platform around stewardship and support the care for challenge events. I talked about this at a conference last week, the data and insight conference. I've got this ambition of creating these AI coaches for people doing events where you can choose like a personality. So you might have like a Sergeant Major coaching you, you know, for your marathon or half marathon, or you might have like, you know, your really energetic best friend who's really encouraging, or you might have like that sort of like grandma, granddad figure. who's always there with a cup of tea or you might have that sarky friend who is hot of gold, or any personality really. And I think this could be a really good, fun use of AI, all that stewardship that we'd love to do, we don't have time for, especially on things like virtual challenges and marathons and things like that. I think that could be really exciting and a little fun way to engage people, but it's clearly AI. but has an authenticity and real to it and you can ask questions and you can get motivation and tips and things like that and yeah I'm quite quite excited by that and looking for a couple of charities who are brave enough to test it and give it a go really so that that's been fun doing that this this week what about yourself Anna? Anna: Yeah, I've been really going at it with the newest models. So I've been working with Opus 5.5, which has been incredible as a new model. And also because I'm doing a lot more trainings also with ChatGBT, I'm learning more about the Astra model and how they are similar, how they're different and how they use different context. But what I have found is that with the Opus 5.5, number one, you can actually make really good slides, which has been very difficult, I've found But the quality of the sides is great. It's suddenly able to retain the sort of look and feel without making it look all sort of sloppy. Historically, had AIs sort of inventing logos or inventing shapes that didn't make sense. And now the models have gotten really, really good where you could just give them a template and they can then spit out the sides that look exactly as they are supposed to with excellent content. within them. that, although it's like a really small, not very sexy, agentic AI sort of thing, it's quite revolutionary because you could be really, really productive. And I would also say that one thing that I have noticed from having to learn more about ChatGBT as opposed to Claude, which is where I do most of my personal work, is they have something called template creator. which is like a skill in chat gbt where you can feed it like three or four slide decks and based on that it will create the actual template where the background things are locked in place that then mean you could use that template to build better looking slides etc etc so that's really useful i would also say from what i've actually been working on the past week number one i was working with the charity to help them look at how they could build a better corporate partnerships. offer online, so like a landing page to attract their corporate partners and get them to convert and, express interest formally and all this. And as part of that process, we were using Opus 5.5 and we found that it starts to create like a brand book for you, where all of a sudden Opus will pop out and say like, okay, these are all your brand colors. This is your font. the spacing that you use on your website, these are the this, this, this, this, these are all the places where that is not aligned, right? So for example, you might, it might suddenly pull out, you know, 15 brand fonts based on your website where you're like, wait a minute, we're supposed to only have five, right? Or these are all the various brand colors that you're using across your website. You could then go, wait a minute, that's not right. or these are the types of images that you have. So that's really useful because then you can create a brand book within Claude or within ChatGBT that then shows you exactly how to use your brand font, colors, spacing, photos, everything, so that everything that is built thereafter follows those brand rules. And that I think is gonna be really, really useful for people who are really using AI to the fullest in the future. Craig: Yeah, good stuff. The design within Cloud Code, now we're not having to go to the design website, think is a real sort of, keeps it a game changer, but it's really fantastic. And as someone who's kryptonite is design over the years, these tools are just incredible, I think. And I've done some really good stuff using those design skills and downloading sort of libraries and templates and animations. It's really good at HTML animation. this new model and again you can do some inspiration you can do some really fun inventive stuff around you know totalizers or you know how well people are doing in terms of their fundraising and stuff so yeah these design skills I think are brilliant and I'm going to start making some videos for me and we've been doing some meta ads using some almost like slideshow videos so there's no images created by your AI they're all real images from the the client but the actual slideshow and video that it creates around it is using AI and we've seen some really good results from that so it's yeah it's really exciting and interesting what you're saying about corporate what was the end outcome of that what what what's changed because of that piece of work that you've done Anna: so the corporate partnerships, it was really interesting because they have had various different offers and you know, sort of gold, silver, bronze, any different packages and whatnot. And then sort of create a landing page that focuses on the value to the corporate partners and sort of did a back of the bag packet calculation. to say like this is the equivalent of this number of hours on the, you know, again, beneficiary time without sort of going to specifics about the client. but it had already gone and done some of these calculations to then help them promote a really clear funnel for them to say like, this level of corporate partnerships means that you can reach X number of children So it was just a really clear way of structuring the sort of shopping list that made sense for corporate partners. And then there was a separate section that was very much about understanding who they are and what they're looking for. And so it was packaging those same offers, but in a like, if you're an organization looking to da da da da da da, then you might want to consider this, this. So it was just very much more interactive and it was... intuitively trying to think about what the corporate partners themselves are hoping to achieve when they land on that page. So again, being very sort of customer centric, going in and thinking what are the KPIs they're probably looking to find or what is the outcome they're looking to achieve, and then structuring the information on that landing page in such a way that might be a lot more attractive to them than what they currently had, which was basically just a sort of get in touch form. So yeah, super interesting stuff. Craig: Good stuff and I guess that leads us nicely on to what we're going to talk about today. This story that was on UK civil society where the first donation made entirely through an AI agent received by the trust fund. What's this saying to you, Anna? What do you think is the interesting bit from this? Because I've got a few thoughts, but I'd be interested to know what you're thinking. Anna: so I am struggling to not say Trussell Trust, because I am old and I'm used to calling them that, but I will do my best. So the anti-poverty charity, Trussell, has said that they received the first donation entirely through the AI agent. And I think... It's interesting because I went in and had a conversation with this AI assistant. It was sort of a test environment where you could chat with this chat bot. It didn't seem too radical initially because it did just seem like a regular chatbot. You ask questions, it comes with answers. But what was interesting was that you weren't sensing any sort of if then question structure. It was very clear that you could ask any sort of question and it would have an answer for you. So, you know, I was asking questions like, So what is this setup exactly? And how does it work? And, you know, trying to understand the sort of AI premise behind it. And it was quite good at sort of giving me a good enough answer, but constantly trying to point me back to Trussell and the cause and what it was trying to achieve for them. So I thought that was really, really well built and really well structured. Is it a sign that we're going to see a lot more of this type of giving? I'm not sure. To be honest, I think it's probably a lot more likely that people are gonna be chatting with their own personal Claude or chat GBT about which charities should they donate to. And I think that might lead sort of the unintended consequences of that set up may be that essentially the charities who are better at reporting and measuring evidence and impact and you know pennies to the pound and all this sort of thing they might see a better conversion because those are likely to be the sorts of charities that an AI is more likely to recommend. And I think what it's doing is a lot of like logic reasoning in the background to say what packs the biggest punch for your money and what's likely to save the most lives. And again, this probably goes back to sort of effective altruism, which is, you know, a lot of the theory behind some of these AI companies and thought experiments. But at the end of the day, I think What this may lead to is charities having to think a lot more about how they document the evidence of their impact. They make sure that that is searchable and that it's readily available information online for them to be able to use. Do I think we're likely to see a lot of chat bots leading to a huge tsunami of donations to all charities? No, probably not. If I were to look in my crystal ball, I don't think so. What about you, Craig? Craig: Yeah, I was going to ask you about the parent mechanism because obviously they talk about go cardless in there. So I'm guessing at some point it makes an ask of you and then send you a secure link to make it or do you give you details in the chat? I was interested in that part of things. Did you get to that point as well? Anna: No, I didn't actually get to the payment point, but maybe that'll be my action from today, is to go and make a donation to Trussell, because it's a great cause anyway, isn't it? And just see how does the actual structure of this work. I'll go and have a go of it. Maybe we could catch up next week on how it went, or anyone else listening, feel free to go and make a donation to Trussell and we'll see how it goes down and if it's scalable. Craig: Yeah, it's really interesting. know FundraiseUp, which I'm a fan of as a platform, they've launched this agentic giving and having a really big push on it. But from what I can understand on theirs, this is about making your charity findable and searchable within Claude or OpenAI, GPT, when you asked that question you just said about how do I, which charity should I give to? Which homeless charity at Christmas? be best etc. So that that would be interesting and they're using a piece of code on people's home pages to do that and then the other thing is you can see in the commercial world this is this is coming and it's coming fast Google and Amazon even though their partners have just been having a big row about Google launched like a personal shopping assistant that would purchase things so it could go on Amazon and buy your best, the best priced, I don't know, whatever it might be, toilet rolls there and order that and do it all for you. But then Amazon has blocked these agents because it doesn't really want agents on its websites or it doesn't want agents that aren't Amazon approved on its websites. So I suspect that, you know, you can see payments coming into. into OpenAI and chat GPT. You can see Google are moving big in this space. So you can give your agents, you can give your AI tools permission to pay for things and go off and research things and book things for you. So there is a shift in that direction. Even I as a big AI fan, I'm not sure I'm quite ready to give my credit card details to. to AI to go off and do my Christmas shopping for me. But I do think it's quite an interesting direction that it's moving in and know, payments is always a big area and that idea of reducing friction, making it easy to give and you know, just saying, I've just seen this, you know, advert on the telly, say it to AI, you know, I really want to support this. make a £20 donation or make a £10 a month gift to it because you've seen something really emotionally compelling. Just being able to do that without having to get your card out, without having to go online, etc. That could be quite good and I'd be really interested in things like children in need or comic relief. If they go down this route and say, know, chat GPT, make a £20 donation to comic relief and it just goes off and does it for you. I think that's where Anna: Mmm. Craig: it gets quite interesting. But I think we're probably a little way off that as yet. Anna: Yeah, it's really, it is really interesting. And I think we'll definitely be on top of it on the podcast. So stay tuned for future episodes. But I think this, this sort of push and pull of, of agents is a really interesting one that you've mentioned, because I also read about how Amazon have now blocked these shopping agents from Meta and elsewhere. And I'm really interested in their justification because I would guess having not read in depth any of their sort of press releases that they also worry about just being overrun with AI agents to the point that they might not actually be able to, have that much traffic on their site safely. But it will be interesting to watch essentially because I think that, you know, they are essentially blocking a route to income. And so the friction must be quite big to want, to have them want to not have these agents in there spending money, because they're spending real money, right? So, you know, back to this point about what do we let agents go and do for us as humans and what are the risks? I think probably there's a lot more to this story that will be coming out. Craig: Which leads us next onto the, I guess the oopsie I did it again bit and I've got a of a soft throat so apologies my singing will be to its normal standard but it's sort of, I've gone Beastie Boys today for this story and you've gotta fight for your right to speak to a human. Anna: Hahaha Craig: And Anna: you Craig: this is this is citizens advice and I was going to say bureau but again that they've dropped the words It's just pure old citizens advice these days and they said say that providers of essential services should have this guarantee of the right to talk to a human because Chat bots and another essential so, you know, they're taking over the customer service world But you know should you have that right to speak to a human especially for vulnerable people? But all the people I talked last time about the voice AI that I'm testing with a client and it's been generally great but there's a small percentage of people who really struggle with it and I completely get that and I can absolutely see why they would want need, not just want but need to speak to a human to be able to get what they need from it. And they talk in this report that was in the Guardian about this endless loop of being trapped and we've all been there and it's so frustrating. you mentioned that tree logic and branch logic machine learning way of doing it. Obviously they've improved now and they're a bit more able to pull things from different areas but there's still this problem and we still get frustrations when using these bots. So yeah, I don't know if you had any thoughts or take on that one, Anna. Anna: I absolutely support this. And again, you know, I am a big user of AI and I am helping charities at the moment automate processes and help with their support to care work. Nevertheless, I 100 % support this for multiple reasons. Number one, because as you say, vulnerable people really need this. I have a 92 year old grandmother who is still alive and she is absolutely one of these people who does not understand when she gets through to a support care machine. She doesn't hear so well so she kind of needs the other person on the other end to be shouting down the phone at her. It's a whole situation. She finds it very frustrating. She finds it very scary. You need that. Reason number two. Reason number two, I would say that essentially I, as a not old or vulnerable or technologically challenged or any of these things, I want to speak to a human when I've gotten through to a machine that doesn't get it, right? So there's two things there. Number one, if you are automating and using AI for some tasks, they have to be good at what they do. So big big fan of like A, do a lot of insight into what your callers are actually trying to achieve on these calls. Do a lot of monitoring of... Do a lot of monitoring of outcomes. as you had mentioned last episode, Craig, you're working with a charity and they will be giving feedback, either thumbs up, thumbs down or CSAT or something or other to make sure that your AI is actually giving the intended experience for your supporters. And give people a third option to speak to an actual human. And again, we've talked about that before. It should be treated as a triage and not the sort of end destination of your support care or your customer service because there is nothing more rage inducing than getting through to some recording or chat bot that does not know the answer to your query and you cannot actually get through to anyone who does. That is just the fuel to get me. I am a bull to a red flag when that situation rises. So that is my hot take. I 100 % support this. And number three actually, which I forgot. In all of this talk about AI and losing jobs and the jobs market being so affected, I would absolutely say, you know, this call to have the right to speak to a human, A, creates jobs or maintains jobs for humans to do that are really important. Of course, you're making efficiencies, but it is actually human connection is a really important thing. And I remember, you know, I've managed support care teams throughout the years. I remember that occasionally you would just have supporters who rang in, you know, once a month or so, pretending to have some problem and actually just wanting an atta, right? There are some of these people where your support care just sort of are, you know, mental health support or, or, you know, trusted people that they want to ring up and speak to. And of course you could say, well, we want to optimize or automate that away. But I absolutely disagree with that. I think there are people for whom you are performing an essential service by speaking to them. You are building a supporter for life, a customer for life in that situation. These are real meaningful human connection moments that should be maintained. So that's my very opinionated hot take. Craig, what about you? Craig: Yeah, no, think the article itself said that, you know, 49 % of people felt stress or frustration, 27 % felt powerless, 44 % experienced delays, you know, 14 % gave up trying to resolve it altogether. Now, if that's... 14 % of people trying to make a donation, for example, trying to give up, then, hey, we've got a problem, haven't we? So I think, yeah, you can use it as a first line and try and solve things, but absolutely make sure you've got it backed up so you can get through to a human quickly and effectively. Or just leave a message and have the system so people will reply quickly. That's definitely been my learning from my recent experience, both with voice and with chat AI. Yeah, I know. Anna: Yeah, so I also just wanted to say at the same time that this report came out from Citizens Advice, there was also a report from Gardner where they make a prediction that the EU will be likely to make this a requirement. So the right to be human may, and I don't know where their prediction is coming from if they have some insider knowledge that we don't, but that this may become a EU requirement in the near-ish future. And as we talk about quite a lot in the show, Craig, obviously I operate from Denmark which are a member of the EU but generally speaking as we see with many things the UK is usually quite quick to follow and so it'll be very interesting to see if that is what happens over the coming weeks and months. On a similar note as well, Craig, we both are aware that the Chartered Institute of Fundraising are creating some AI guidance at the moment. And there have been a few comments coming out in various news reports and charity news sites about the type of direction that they are sort of going in and the type of things that they want to cover in their advice. And it does very much seem as though they are also taking inspiration from the EU Act. more evidence that what the EU tends to do the UK will usually tend to consider following suit. Craig: Sobering stuff and I'm probably a good pointer enders for this week and I know we've got a few things in the pipeline coming up for future episodes that we're excited to share and Make sure you tune in and listen next time. I'll be on the slot until then. Bye for now Anna: See you next time.