Latest / Beyond the Slop: AI in the Non-Profit Sector / Episode 2: AI fraud, dashboards, supporter care bots, and agents.
Transcript
- Craig: Hello and welcome to the latest episode of Beyond the Slop. I'm Craig Linton. Anna: And I'm Anna Hessenbrick. And this is our podcast where we talk about AI use cases in the charity sector, some of the cool stuff people are building and doing with AI, as well as some of the pitfalls, ethical concerns, and other challenges that we have in the sector. Craig, what are we talking about today? Craig: Well today we've got a very interesting I would say and controversial oops AI did it again which I think really shows some of the fears come to life I think of lots of charities and lots of fundraisers but we're going to start as always with what we've been building this week and Anna anything interesting that you've been testing and trying? Anna: Yep. So one of the use cases that I'm seeing quite a lot with charities is a lot of charities are keen to build dashboards. And so one of the things that I've actually set up a Claude skill, because I'm doing it so frequently now that it it sort of needs its own skill, is scraping the internet for publicly available data. So for example, if you're looking at a charity that has beneficiaries or programs that they deliver in various councils across the country. What you then do is you go and scrape the data on like what's the population of each county, what's the poverty level or whatever the causes that you're looking at. And then you pull the data from either their internal source data on programs that are being delivered or the annual reports and accounts or something like that. And then essentially what you can do is build these super cool dashboards where you've got an overview of how are your objectives being delivered in each of the various councils, what is the ratio of like potential beneficiaries that need support and how many are actually being reached. You could have a little interactive map. So there's lots of super, super cool dashboards that you could build. that really help visualize and strengthen the charity's idea of of how they are delivering to their aims. Craig: In the UK the government have this incredible API catalog which is completely free of charge you just register for a count and it has like all sorts of things that you can do. You can have the British Geological Survey you know maybe a dashboard of polar bears spotted in the air that they've done but like every council has it the Department of Work and Pensions you can do Citizen APIs you can get to drive a D.Va.L.A. I can never say authentication. I don't know why I even try to say it every time I say it. The Office for National Statistics, have the geography. There's so much stuff on there. It's insane what you can do all free of charge. Like I say, AI is brilliant for connecting that together and then doing the data visualizations. Anna: Yeah, absolutely. And one of the things that I have been really surprised is that I've done this for three different charities now. They've all reached out to get some training on how they could be utilizing AI to build stuff like this. And each time the cause has been really different. So I've sort of started out thinking, does does data even exist for this? And I've been positively surprised every time because no matter So far with the three causes I've worked with, no matter what it is they're trying to report on, there is a surprising wealth of information out there to draw from. So it's super cool what you can what you you can build based on just publicly available information and the data that you hold within your own organization. So that's Craig: Yeah. Anna: that's use case one for me. What about you, Craig? Craig: yeah well i've just spotted actually just because i'm on there the there's the national fish population database where you can look up fish length fish weight and all sorts so yeah if you if you're into that then you know find out the average weight of a perch or a pike and yeah it's incredible like you Anna: Ha. Craig: say what's there this week for me it's been all Anna: Like shooting fish in a barrel. no, sorry, terrible. Terrible. Craig: boo, we need some sort of air horn or like noise when a terrible puns made but hopefully that's not killed the vibe too much apologies for Anna: Yeah. Craig: Apologies on Hannah's behalf for that. I'm all about the bots this week and been building Almost like support care bot for specific use cases. So I've created my first email response bot. So what this does, attaches to a mailbox. It's a mailbox specifically for a. virtual challenge with one of my clients and you get the same questions over and over again. Where's my t-shirt? You know what? I can't. I've got problem with my just giving page etc. And so you know the team will literally like copy and paste you know same the same thing 50 times a day which I think anything where you've got like that I think is a great use case for a bot and so what it does is it goes to the inbox it drafts the reply at the minute because it's still testing. human review before sending, but hopefully once we've got the confidence in it, it will just reply and that frees up the team for all the lovely sort of support, care, support, experience stuff for the people who are doing really well, which is great. And the second bot, the tool and platform that I use has just got the new voice-to-voice translation that Google and OpenEye have released. So traditionally, if you're doing like voice AI, it sort of basically listens to you. almost like translated into data, into text and then back and that's why you always got that pause sometimes that used to be really frustrating. Now the model is of audio to audio in like 90 languages, which is incredible. So you can just switch. I mean, I can't cause I'm very much, I struggle with English, nevermind a second language, maybe a tiny bit of German, but I think you could, know, instantly switch from English to Danish and back again. And it just wouldn't pause. just naturally picked that up and translate. So we're doing that for a potential helpline again, where the team are just overwhelmed with the volume of calls and again same things keep coming up time and time again so almost like a triage still very clear escalation I want to speak to a human bang away it goes but it's just that front line that's just saving so much time because basically the team haven't got the capacity to answer all the calls at the minute for this this one project so that's quite quite exciting and it's interesting the the AI voice stuff particularly, how much is coming on, but there's still some nuances and it still can sometimes struggle with accents a little bit and talks over you and a bit stilted at times. But it's definitely, I think the benefits outweigh those occasion times that that happens. Anna: Yeah, that's super interesting. And, you know, on the EU Act that we were talking about recently with AI, I think, you know, it's really clear when you're using an AI bot that you have to disclose that at the very top and say you're speaking with a with a bot. and I think you always worry as a charity that that's gonna put people off or that they're gonna sort of can consider that to be disingenuous. But I think it's also because we've had so much experience historically with bots being completely bloody useless. And you ring up and it's like, tell me who you want to talk to, and then you say something like, and they've completely misinterpreted it, and you're just like, just give me a bloody human already. Craig: Yeah. Anna: but I think AI has come so far in the last, I wanna say like three months even, really recently, to the point where that's not the frustration anymore. So I think as long as you disclose at the top of the call and you're bot is actually genuinely really useful. I think that sort of brings the friction level down quite a bit, doesn't it? Craig: Yeah, and I think it's finding use case by use case and finding, you know, where you can try this with minimum risk. Though funny enough on the the voice box, there was a news story this week about I think canceling Yorkshire. It did introduce voice AI and just like inundated with complaints that the voice AI couldn't understand the Yorkshire accents or like certain sayings which I'll dig that out and maybe share it in the show notes. But yeah, it may be so because I I've had that sometimes with AI, it really struggles with my sort of North East accents so it's still not perfect but definitely as you say much much better than it was previously. Anna: So funny. I've also got another use case that I am playing around with at the moment, but it's these AI assistants, and there's a a few different companies that are having a lot of success with this. but I actually got an email today from Make, which is a very popular tool that people use with their AI to interconnect various systems and automate. processes within it and they've just launched their own bot assistant called Maya, where you essentially can go and tag Maya in Teams or in Slack or in various different channels and a assign tasks to them and Maya can go into your CRM and pull the data out and analyze it and do all these jobs. and there's, you know, make's not the only provider. There's loads of different companies that are sort of making this a thing, but it goes from a sort of chat GBT or chat style function to where you can actually allocate and delegate tasks to this AI person. So I'm having a little play around with that to see I I hear exceptionally positive things. Craig: Great, and just to say, was just gonna say make.com have a really generous nonprofit program. It's really well worth applying for. If you use Zapier, which is the one that I think lots of charities do, make.com I find much easier to use and cheaper, and they have this nonprofit program as well. But I think that's interesting because I love those use cases, but. Again, when I've talked to this with IT team, they're very nervous about having agents and bots and things. I think talking about how you put guardrails in place or how you test it and things, I think before going on LiveData, I think is really interesting. But I love that. That whole agent space, I think is really interesting. And again, something we may talk about in the future, because, you know, there's like this agentic future that people talk about where basically we're all sat having a cup of tea while we've got all these agents doing all our tasks for us is quite scary but also quite inspiring if we can make it work as it should work. Anna: Yeah. I mean the the the make dot com one just launched, I think it was today, I got the email. but there's another company that makes a an agent called Victor. There's another agent you know, there's there's lots of these different products. And I sort of went in out of curiosity and you know, went on trustpilot and read it and was looking at all the sort of is this actually very useful? And there's just glowing Reviews of people being like, my God, these agents are actually just so useful. I will go and say, like, write me an agenda for this meeting based on the conversation we've had on Teams. And da-da, you've just got an agenda ready, or like, you know, send out the minutes, i whatever. Like all this stuff where instead of just writing a document or writing a piece of text, it's actually doing all the actions that you've asked it to do. The thing that You know, from an ethical perspective, that also is a big question. Not only is how much data do you want to give these agents access to, but also this will genuinely have a real effect on the jobs market. You know, there's there's been so much toing and froing and discussion about, you know, are jobs really at risk? Is this really a a thing? essentially the the the agents, you know, really are doing a lot of the work that a a sort of junior employee would be doing, right? A lot of the sort of administrative donkey work. that you have a lot of sort of new graduates doing that kind of work. i there's there's a real concern there, right? What what will this actually do to the jobs market? Because quite a lot of busy work and and and administrative work could quite easily be handled by these kinds of bots. So I think even though it's been a a conversation that's been sort of punted around over the last couple of years about is AI coming to take our jobs so far There hasn't been a huge impact as far as we could tell. But these quite competent, ef efficient bots I think are probably gonna start to tip that scale. Craig: Absolutely, absolutely and like I say, watch this space because it's still very early in the development of these. But yeah, excellent. Let me know how you get on. forward to hearing more in the coming weeks. we've got quite a big oopsie. I did it again this week and I know you wanted to talk about this whole Lily J foundation controversy. you want to tell us more? Anna: Yeah, yeah. So this is a controversy that's come up. it's from Australia and it's an a an online influencer who has set up a charitable foundation. It's called the Lily J Foundation and had talked about, you know, setting up orphanages in Uganda, doing aid work in Gaza and a couple of other things, and then posted some videos of these orphanages and and her in Uganda delivering help to children. And and if you sort of give it a Google, you'll see these stills from their videos to show that it's very it's very convincing AI footage But there are some really clear telltale signs that this is this is definitely created with AI. And so ABC, the news network in Australia, then went and tried to do some investigative journalism and they found that the foundation actually isn't a charitable organization at all. It's a private company that she has set up. And you know, when they speak to authorities in Uganda and Gaza, there's nobody who can recognize any of the services that she has claimed to be delivering. So it's a huge scandal. It's obviously fodder for the press. and it affects us in the charity sector massively because even though this isn't a legitimate charity, it is the kind of fear scenario that that has that real effect on whether people trust charities or not. And I think We've talked a lot about in the past that trust in charities has been falling and charitable giving has sort of been moving away from institutions and more towards these sort of grassroots fundraising initiatives online or through social media accounts. And this is the dark shadow of that trend, which is, you know, people feel more trusting of a person that they can see and that they can relate to and that they you know, is posting videos of the beneficiaries and where their money is going. So it ha you know, a lot of donors feel more connected to that. But the problem is when you don't have a real charity and you don't have people who are experts in charitable giving and delivering programs and services, you are much more at risk. Of precisely this, which I mean this is directly fraud, but you could have a lot of situations where people genuinely have good intentions. They just don't have the knowledge and expertise that the charity sector has to be able to do it responsibly. So whether or not this has an effect and is a trend that we're gonna see more of, whether that affects how people see charitable giving in a positive or negative sense for for charities remains to be seen, but it's definitely it's definitely a a worrying trend that may be emerging here. Craig: Yeah, and I think AI just brings those possibilities of fraud, I guess, making it easier than ever. And look, there's always been charitable fraud, very sadly, and people have always exploited people's goodwill and generosity, know, sadly, for hundreds of years. And it does damage reputations and trust. And I hope. hopefully this has been caught and reported but it goes back to that basic human need doesn't it? know what she's doing is creating those compelling content, the emotional stories, know clear ass you know she's actually doing take the AI bit out of it as she said she's actually doing like really strong fundraising with all the stuff we tell people to do But it's fake and it's not clear if this foundation, well it's clearly not real, but what's happening to the money? She's got these millions of views across Instagram and things, so no doubt generating quite a lot of income from that. And it does, we don't want to give people the excuse to say, they're all rotten, you get tarnished with the same brush, we? it's, all rotten, we're all terrible. you know you sometimes see on Facebook I know you're not on Facebook kind of but often you get these like ads for one-off appeals and they've got these horrific images of children that again clearly when you look into it it clearly scams but like crowd for crowd funders a platform that has been exploited a lot over the last sort of 10-15 years and it's very easy to sort of do in some ways but sadly there's a few bad apples spoiled it for everyone but I do think AI imagery and video is controversial. Anna: Yeah, and I think, you know, the the the difference with AI to to the fraud that has always existed is really that you can be really convincing. And I think had I, you know, you can see the videos online of of this Lily J Foundation and they're really convincing. it really looks like she's in Uganda, she's interacting with these children. It's only some really small details Once you see them, they seem obvious. But unt you know, if someone hadn't pointed out to me, for example, the there's a sign in the background that says Lily J Foundation that is in one place during one scene and then is suddenly not there, and a a ventilation or something is has moved slightly along the wall. Like it's things that very clearly signal that this is not a real place. But you know, the humans are like singing and dancing and moving around and and it's so convincing. And back to back to your point from last episode about Will Smith eating spaghetti. and for those who haven't listened, sorry for the reference, but but you know, AI has really got to this point where you can make some really, really convincing content that is really hard to spot, that it's AI, and and the capabilities are just getting better and better and better all the time. So unfortunately it means that you can use these tools for good or for evil, but it's getting harder and harder to spot that it actually is AI generated. Craig: Yeah, thanks Anna. Yeah, it's a really interesting story and saddening in lots of ways and shows that EU AI Act and the regulation that will probably come to make this easier to spot and it shows the importance, I guess, of regulation and doing it. what we need to be careful of is we don't throw the baby out with the bath water. There is some really positive cases of using AI with four charities and we'll look at that side of it as well. But yeah, there's no doubt this is concerning and worrying. Anna: Yeah, for sure. Any other oops, AI did it again from you, Craig? Craig: Well, I've got something I'm going to save for next time but I thought finish off and end we've got our big debate next week where I'm going head-to-head against you to talk about some of these pros and cons of AI and you know really excited that I'm going to be doing some research over the weekend and you know I've got a few hopefully a few interesting talking points and some interesting you know points of view hopefully that will be know again go beyond the the generic arguments we hear time and time again around AI and the pros and cons of it and have you been sharpening your debating skills for this? Anna: I absolutely have you better you better be careful, Craig, Craig: I'm going to have to work my game. But Anna: Ha ha. Craig: yeah it would really interesting and we'll be recording that and sharing that as an exclusive podcast episode as well. So we better get back to the debate prep then Anna and I hope you have a fabulous weekend and thanks everyone for listening and we'll see you next time on Beyond the Slop. Anna: See you next time.