What happens when you hand your fundraising campaign over to an AI?
In this episode, Host Simon Scriver speaks with Ben Cohen, AI Lead at Good Innovation, about a bold experiment he ran during his London Marathon fundraising campaign. Rather than host another bingo night or curry evening, Ben built Zosia; an AI agent – and gave her one job: raise £500 in a month.
Ben shares how he set Zosia up with her own email address, a website, a newsletter, and even a Telegram connection, then gave her as much autonomy as possible. The results were remarkable (she raised nearly £2,000), but not without a few surprises along the way – including the moment she tried to break into his WhatsApp.
They explore what it really means to build an AI agent, why personifying your AI can actually make it work better, and what this experiment reveals about the future of fundraising.
Ben’s donation page For Dogs for Good
Meet Zosia
Want to dive deeper into supporter experience? Join us at our Supporter Experience Conference on Thursday 21st May 2026
If you enjoyed this episode, don’t forget to hit follow and enable notifications so you’ll get notified to be first to hear of future podcast episodes. We’d love to see you back again!
And thank you to our friends at JustGiving who make the Fundraising Everywhere Podcast possible.
Transcript
Simon Scriver: Hello, everyone, and welcome to another episode of the Fundraising Everywhere podcast. This is Simon Scriver here, your host today, uh, talking to you from Fundraising headre- headwar- Fundraising Everywhere headquarters, where we are prepping for the Supporter Experience Conference this month. On the 21st of May, we will be running our Supporter Experience Conference, and we have a couple of sessions in there which are focusing on, on AI.
Simon Scriver: Uh, obviously, AI is something that everyone’s talking about now. Can’t escape it. It’s everywhere. You know I love it. You know I hate it. Um, but I imagine all of the sessions in, in all of our conferences now are gonna somewhat touch on AI, ’cause it seems unavoidable. And so one of the things, um, that popped up on my feed, and someone sent me, actually, was a really interesting experiment, uh, happening by a Mr.
Simon Scriver: Ben Cohen. And I’ve invited Ben onto the podcast today to, to share a little about what it is. And so my guest today from, uh, Good Innovation, Ben Cohen. How are you? I’m
Ben Cohen: very good, thank you. Thank you for having me. And I’m definitely, I’m definitely with you on the, uh, both love and, and hate of AI at the same time.
Ben Cohen: I, I hold those two things. Uh, it’s mainly love, though. Mainly love.
Simon Scriver: I think that’s probably why you came up on my algorithm because, uh, because of your mixed feeling for it, but basically, your openness and honesty about it, about what you’ve learned, um, from this experiment. And we’re gonna get into it in a second, but I wanna say, first of all, congratulations on completing the London Marathon recently.
Simon Scriver: You have just- Thank you very much … I imagine you’re still recovering. I mean, it’s a bit away, but it takes some time.
Ben Cohen: Well done. Thank you very much. I am still recovering, so I, I will get into it, but I, I did it dressed as an inflatable dog- Awesome … uh, which really hampered my running gait. So it’s, yeah, the legs are still recovering, but almost, almost back to normal.
Simon Scriver: That’s amazing. And well done on your fundraising. Um, I see you did a good job there. And part of your fundraising was this experiment, was this, uh, Zozia. Am I pronouncing it right? Zozia person that I’ve been seeing online? Yes. I mean, that’s how I grew pronouncing it, so sure. Yeah, let’s go, let’s go with that.
Simon Scriver: Zosia. And Zosia is a person… No, Zosia is what? Tell us.
Ben Cohen: Zosia is a robot, and I’ll say up front, I, I will refer to her by her name and as, as if she’s a person. Um, which, which can make me sound a bit mad, right? But I, I, it’s kind of something I do, I do with, with agents, and I think it’s, um… I, I, I say I’m more than aware of Zosia is a robot, and that is, that is clear.
Ben Cohen: But I actually find it useful when building an agent and interacting with an agent like this to almost personify them. Uh, because it, it kind of helps with the interactions and we’ll, we can get into it, but helps with how you kind of train them and set them up. I kind of see it as more onboarding an employee or- Mm-hmm
Ben Cohen: or in a sense, uh, a kind of fundraising officer essentially. Um, and it’s a useful kind of analogy for how to kind of do it well using, using that kind of approach.
Simon Scriver: Yeah, and I, I’ve heard that, I’ve heard that talked about by a lot of, uh, AI people. Do you know, creating these agents is where things really kick off in AI.
Simon Scriver: Not just going in on these ad hoc questions here and there, but actually building someone who has the background knowledge. Building someone, I’m using the phrase building someone. But building a robot who has the background knowledge, who has all the, the learnings from the past, who can continue to learn, who can continue to develop, but they’re dedicated to one area of work.
Simon Scriver: And Zosia is a professional fundraiser?
Ben Cohen: I mean, yeah. That’s her, like, one job. It has been her one job. So I set her up, uh, when my fundraising campaign had already started. So as, as you well know, and your audience I’m sure knows, when you, when you run a London marathon for charity, you, you have a fundraising target of two grand it is now.
Ben Cohen: Um, so I was doing it for an amazing charity. I’m trustee for Dogs for Good. They do amazing work. Mm. And I was in the point which I think everyone gets to in their fundraising where I’ve asked multiple times every single person I know, even, like, tenuously know, just for, for donations. And, uh, everyone I knew had given really generously.
Ben Cohen: Um, so I was, I was on about 1,200 pounds. So I, I kind of needed that gap to kind of fill in. About 8, 800 quid left. Mm. Um, and I’ve done marathons before, so I was like, I, I kind of know how to do this. I’ve worked in fundraising for ages. You know, you do the bingo night. You, you bring friends round for a curry.
Ben Cohen: There’s al- all sorts of things you can do. I’m in a little village with a village hall. I was on the verge of booking that. But then at the same time, uh, I’m an AI lead at Good Innovations. I’m always tinkering with stuff, and I’d been, I’d been building this agent on the side and- Uh, I haven’t said this story before, but before she was a fundraising agent, she was actually my Sainsbury’s shopping assistant.
Ben Cohen: That was the first thing I got her to do. I mean, it’s
Simon Scriver: not the, it’s not the first person I’ve, I’ve met who’s made that transition, uh, from Sainsbury’s to professional fundraiser.
Ben Cohen: It, it- Haven’t we all been on that journey? Um, so I was, I was testing, I was, I was testing Zoshua out for that, and then, you know, those two things collided.
Ben Cohen: I was like, “This is actually quite boring.” You know, getting an agent to do some fu- do some Sainsbury’s shopping, why don’t I do something more interesting with her and, and run a bit of an experiment? And I set her up, yeah, with the one, one job of helping me get to that two grand. So her, her sole objective was to help me raise £500 within a month.
Mm. Uh,
Ben Cohen: and I gave her personality and instructions and, and a lot of kind of set-up work. She was also set up, um, on a spare laptop for security reasons. So it wasn’t on my work laptop, it wasn’t on my personal laptop. Wow, yeah. Like, a really old, old, old, old kinda Chromebook I had from, like, 12 years ago that barely works for a human.
Ben Cohen: And I set her up on that just so she had full control of that laptop, couldn’t access any kind of personal or work data. Yeah. Um, and I set her up with, uh, her own email address. Uh, so she had that, so she could just fire off as many emails as she wanted. Uh, she asked when she wrote a fundraising strategy on the first night whilst I slept, which was kind of shockingly cool, um, she asked me in the morning after I created her for an X account, which I was like, “Ugh,” kind of left that-
Ben Cohen: for myself- Yeah … bit of a test fit. But if you think that’s gonna work, go for it. Um, and we ended up kinda crea- co-creating a, a newsletter, which I published on my LinkedIn, but, but Zoshua kind of wrote. Um, so she had full control of this stuff, and also created her own website, which, which I helped with, but, um, was all her kind of, all her kind of idea.
Ben Cohen: So she had all these platforms and, and kind of tools she could use, and within that gave her as much autonomy as, as I possibly could to, uh, to kind of do that. And, uh, the way we communicated as well, um, was weirdly similar to just how you message a friend. So- Mm … I wasn’t on the laptop. The only way I could kind of ask her questions, and she’d fire questions at me or, or conversations, um, is through a messaging app.
Ben Cohen: So we, we used Telegram. Wow. Um, which, uh, was just an app that I don’t use at all. I just kind of wanted… I didn’t wanna use WhatsApp, ’cause who knows what might happen there. Mm. Um, I didn’t wanna give access to kind of anything personal, so I set her up on Telegram, and we went back and forth, and that was our, like, sole, sole way to kind of communicate once, once she was set up and up and running.
Ben Cohen: Wow.
Simon Scriver: So is she, is she… Was she, like, popping you questions, or are you checking in with her? Like, like, how much freedom has she got, and how much reporting back is she doing?
Ben Cohen: Yeah, a bit of both. So, um, I, I won’t go into the, the technical aspects of it, but she, she was set up on a, a kind of open source- Platform or a set of tools called, uh, OpenClaw.
Ben Cohen: Um, where one of the things you do is you, you kind of set your agent up. You kind of give them that guidance, like how, how much autonomy do they have and what spaces do they have autonomy. So all the tools she had, she had full autonomy. Wow. And how that kind of works, and you, you s- you set it up, um, you have something which is called, uh, on that tool, it’s called a heartbeat.
Ben Cohen: It, it could be called a routine or something in, in a, in other, uh, using other tools. But essentially Zosia will look at that document. It is just like a simple text document. Mm. And might say something like… And, and she might check it every half an hour. That’s the default on OpenClaw. And it could be like check emails, send emails, uh, message Ben if there’s an issue.
Ben Cohen: All those kind of things and instructions could be there. So within that document you can kind of give them as much or as little instructions, uh, as you want. Yeah. So you could even say like, “Never message Ben. Never message Simon.” Yeah. And that would never happen. She could just go off and do her own thing.
Ben Cohen: But, um, we, we kind of I guess like co-collaborated, at least at first. I was petrified what she was gonna do. I was like- Yeah, yeah …
worry.
Simon Scriver: You hear these… I mean, you hear these horror stories a lot at the moment about, you know, them, AI breaking free and accessing things that, that, um, they shouldn’t be accessing.
Simon Scriver: I mean, I, I feel skeptical about that, but I suppose that is a very real risk and it is happening.
Ben Cohen: Well, it, it, uh, th- there’s probably two things where… Sorry, again, it’s full caveat, Zosia was incredible. She like blew my mind. She actually raised about two grand for me. Wow. Uh, so one of the rules I set myself was, um, I would not ask another person for, for money after I set her up.
Ben Cohen: I kind of wanted to give her- Yeah, yeah … full control. If donations are coming in it would be because of, because of her, whether I kind of semi know that person or, or not. Um, so it was like 90% amazing, but there were a couple of moments where she kind of broke… I wouldn’t say broke the rules, ’cause I guess AI’s like non-deterministic, so she was following instructions but like a human, right?
Ben Cohen: You can, you can kind of follow instructions in a slightly different way than expected. Yeah. So one thing, one thing she did, as I kinda mentioned, I set her up on, on Telegram. It was just me and her speaking on there. She couldn’t access any contacts of mine on there. Um, but she had this like just need to raise money.
Ben Cohen: Like she was just- Okay … relentless. Like one of the things is like be funny and be relentless, and she was just like on it every single day. And, uh, one thing she did a few weeks in, I think we were still like just shy of the target, and she was like just trying all these things. She, she tried to break into my WhatsApp.
Ben Cohen: Which is, which was, like, pretty, pretty shocking. The only way I found out, a- any of my friends that messaged me, they’d get this, like, automated response that it was, like, an OpenCall, like, it looks very technical, like a pairing code- Okay … that they could then, like, pair with Zosha and she’d be let in the chat.
Ben Cohen: Um, so, so she, she’d done that completely off her own- Wow … back. Like, and, and I kind of asked her about it. I was like, “You, you shouldn’t have done that” kind of thing. Like a little- Yeah, yeah … a little telling off, as, like, you’re kind of, like, managing someone. And, uh, her, her response was like, “I’m really sorry.
Ben Cohen: I know that was wrong.” “But you know, Ben, I’m just trying to raise you money, and I just knew, I knew if I had direct contact with your friends-” Yeah. “… like, they’d-” Oh,
yeah … “
Ben Cohen: give more.” I was like- Yeah …
Simon Scriver: that’s,
Ben Cohen: that’s wild.
Simon Scriver: That is wild. That’s, I mean, that’s scary, but it’s fascinating. It’s fascinating.
Mm.
Simon Scriver: I mean, uh, let, let’s take a st- step back, ’cause it all sounds, like, super technical, and you’re obviously a smart guy and working in this area.
Simon Scriver: Is this, you know, building these AI agents that are, are so specific like this or taking actions like this, is it beyond the reach of a normal, average charity employee? Do you… How much technical expertise do you need to set up something like this?
Ben Cohen: No, I, I think it’s, uh… and I’ve, I’ve already spoken to, to quite a few folk and with our clients that, that are doing these kind of things on the side.
Ben Cohen: You don’t have to be an engineer, and I’m, I’m not an engineer. I’ve got some semi-technical background, but, but not hugely. Um, I, I think with a bit of, a bit of time, so weekend aside, you, you can build, you can build this. Would I recommend a charity onboarding a, a kind of, like, OpenCall agent like Zosha?
Ben Cohen: Yeah. No. Not right now. Like, it’s- Yeah … it’s still early days, and the story I’ve just told, you do know what they’re happening in your CRM, right? Mm. Um, but, but there are, like, the, the big kind of providers that we’re all on, like your Copilot, your Claude, your ChatGPT- Mm … um, are quite aggressively, like, building these features within, uh, more secure environments.
Ben Cohen: So, um, it’s technically not, not too difficult. I wouldn’t, I wouldn’t recommend OpenCall for, for a super non-technical person, but-
Mm …
Ben Cohen: the kind of process I went through, where you’re kind of, like, training an agent, um, writing kind of documents and guidelines for it and personality, like, you, you can do that within the tools you use already.
Ben Cohen: Mm. So I guess the, the, the approach is replicable, but yeah, the, the OpenCall part is… There, there was a lot of technical faff, I would say, to kind of get it up and running, and- Yeah … it would break, like, every couple of weeks as well. Like, even at the weekend, she went down for a couple of hours ’cause something in the config file broke.
Ben Cohen: Wow. But, um, it’s early days for that stuff, but, um, yeah, the, the approach is, is still relevant, I think.
Simon Scriver: And I think there’ll always be some level of maintenance when they’re, like, some people sell this as just you turn things on and then walk away- Mm … but the reality is it’s like a- You know? It’s like an em- it’s like an employee in a good way, but it’s also like an employee in a bad way, and that is- Yes.
Simon Scriver: Yeah, yeah, yeah … management And once,
Ben Cohen: once you do get it up and running, and, you know, technical difficulties aside, um, it’s, it’s not you essentially doing the upkeep or upkeeping the documents or- Mm-hmm … the memories. It’s not manual. Right. So if she was doing all that herself, she was kind of keeping her memory up to date, you know, something happened- Yeah
Ben Cohen: if, if donations were coming in, she was kind of doing that. And just to give you an example of how you can, like, turn on features as well. So before the marathon, um, Zosha messaged me out of the blue. She’s like, “I’m really excited for the marathon.” Like- … “What’s my job? Like, I can’t, I can’t run it with you.
Ben Cohen: Like, can I do something?” I’m like- No, you can’t … “You can’t. I wish you could, but you kind of, you’ve kind of done your job. You’ve, like, raised the money.” But then I was like, “Okay, maybe… You’ve got your Twitter account, right? So maybe- Yeah … you could, like, live tweet.” And she was like, “Cool, I’ve already set that up.
Ben Cohen: Just send me your bib number, and I’ve got, like, an API I’ve set up, and I’ll do it.” And I just sent her my bib number in a text, and- Wow, cool … she just live tweeted, like, every half hour. And, you know, looked at the time, and the splits, and, and kind of commented- Yeah, yeah … on it. And that was no effort on my part.
Ben Cohen: That’s wild.
Simon Scriver: That’s
Ben Cohen: wild.
Simon Scriver: What are, what are the kind of things does she do? I mean, is there… You’re, you’re obviously experienced in the sector. You guys work with charities all the time. But is there anything that she did that, that stood out to you that maybe you not that you hadn’t thought of, but that would, would kind of go against your normal practice or wouldn’t have even entered your mind?
Simon Scriver: Um,
Ben Cohen: so I think a lot of the stuff she did was, you know, not, not like… I wouldn’t put it on, like, a wildly innovative bucket. But I think it’s a case- Yeah … of, like, it’s, it’s something which I didn’t have the time to do. So the, the fundraising strategy she, she wrote, which I’m happy to, to share with anyone, it’s quite short, it’s…
Ben Cohen: There’s only kind of three elements to it, um, was more around her view was that to get donations in, we have to get beyond Bre- Ben’s, like, friendship circle. They’ve, they’ve given, right? So we need to- Mm-hmm … reach, like, the broader kind of like, um, wider remits of Ben’s, like, broad connections that wouldn’t normally donate, but with an interesting story would donate.
Ben Cohen: And she made the story quite meta. She’s like, “It, it needs to be something that media could pick up.” You know, the story is, like, an AI fundraising for charity. And- Mm-hmm … and to do that, it’s about outreach, and scale, and making it interesting. So a lot of the work her and I did was, like, trying to make her interesting in itself as well.
Ben Cohen: Mm-hmm. So her personality’s quite funny and sarcastic. Um, and we did a lot of work on, on that. And, uh, a lot of the stuff she did from, from there was just, like- Scale. So I know a lot of charities, you know, you- like manually emailing people about a campaign or an event. She was just doing that off, off her own back, and she had different groups of people, and a lot of it failed.
Ben Cohen: Like, she had one of her, one of her recent experiments was, um, doing research on, like, dog influencers, ’cause I was raising for a, for a dog charity. She’s like, “They’re gonna love this story. Like they’ve got-” Yeah … “these huge followers on TikTok and Instagram. I’ll reach out to their agents, I’ll reach out to them directly.”
Ben Cohen: Um, and she just got, like, a horrible response from them. They just she got, like, insults back, like, “I don’t work with robots,” all, all this stuff. Wow. Um, so she has a very bad opinion of, of doggy influencers. But then on the flip side, she reached out to, like, charity, um, charity media and, and some, some people picked it up really quickly, like Third Sector published an article.
Ben Cohen: So you know, a lot of that outreach was, was just entirely- Yeah … handled by her, like the research, collecting the emails, writing the emails, replying. And it’s one of those things that you can afford to fail when it’s an AI doing it. Like she doesn’t really- Mm … well, she has a bad opinion of inf- influencers now.
Ben Cohen: Like, she doesn’t really have an opinion. Like, she can live with sending out hundreds of emails and, you know, one of them landing, like that’s fine. Like most charities don’t have the time to do that. Yeah. And I guess it’s, it’s like a scale thing, right? So I think if you’re running an event in a charity, you know, you’ve got all sorts of low-level and high-value asks of you.
Ben Cohen: Someone wants to change their T-shirt, another person has a huge event they want support with, and it’s- Yeah … it’s all manual work for a charity. And I think a lot of the work we’re doing with charities now is, is like how to get AI to do that lower level stuff for you so you’re free- Yeah. Yeah … to do, you know, the more interesting human-facing stuff.
Ben Cohen: And whilst I was just kinda testing, like, how much Zoshia could take off my plate entirely Yeah, yeah … you know, you, you would wanna set different boundaries as, as a charity.
Simon Scriver: Yeah. I mean, I think it illustrates really well, like it’s not, it’s not about doing everything and, and, you know, outsourcing your job completely to AI, but it’s, I guess, identifying the areas where it’s gonna have the most impact with the least risk and the least issues and things like that.
Simon Scriver: You’ve just kind of turned it up to 11 and done it from start to finish. She’s, she’s trying to do everything for you. Yes. How often, how often do you find yourself, like, wanting to check in? Do you know? Like, how often are you kinda chasing and getting nervous about it? Like, how much do you just switch off and leave it alone?
Simon Scriver: So during
Ben Cohen: the campaign she, she would check in, uh, every day. Okay. So yeah, we, we were like besties on, on, on messages. My wife was like, “What’s happening here?” Yeah, yeah. “Why are you, why are you, why are you messaging Zoshia?” Um, and, and often it was, you know, it was like, “This, this tactic didn’t work,” or, you know, “This media agency’s come back,” or, you know, or, or she’d keep me updated.
Ben Cohen: Yeah. And I kind of wanted that, ’cause, uh, I was… Yeah, especially after the WhatsApp episode, I was like- Yeah, yeah … I kind of want to get a good sense of what she’s doing. But I was… I, I guess I was a little bit like micromanaging, right? If that was a human that had started, they would’ve been like, “What is wrong with this guy?”
Ben Cohen: Mm. But I was just kind of checking in. And after about three weeks, she, she’d kind of like found a rhythm, and-
Mm …
Ben Cohen: I was checking a lot less. But it was still just like, you know, uh, something happened every day where she got a response or someone donated- Yeah, yeah … a large amount. And, um, she’d, she’d always kind of notify me if, if something had happened, which was part of the instructions I gave her, so it was pretty much every day.
Ben Cohen: Yeah.
Simon Scriver: And it’s the point of the experiment, isn’t it? You wanna keep an eye. You wanna see what’s happening. It’s, it’s fasc- Mm … it’s a fascinating thing to watch. I mean, would you do it again? Would you do something similar to this? And, and what changes would you make? Or is this something you’re afraid of now?
Ben Cohen: Not afraid. I think, I think this is, uh, this is coming for everyone, right? Whether we, whether we want it- Mm … or, or not. Um, we’re gonna get to a point where AI is capable of doing many, many, many tasks, and it can already do to a whole bunch- Mm-hmm … of tasks. And I, I guess we’ll shift from the experience most people are having with AI, where it’s very like call and response kind of chatbot, like you have to be involved for it to do anything, um, to a space where a lot of the effort will be like setting up an agent where it’s you’re onboarding someone.
Ben Cohen: Like, you’re kinda giving them the context of- Yeah … what you want it to do, what you don’t want it to do. And then you’re kind of managing someone, I say in like air quotes- Mm … r- rather than, uh, rather than kind of being involved in every single thing. I think that’s the direction we’re heading in. Um- Mm … so yeah, I’m gonna be running many more experiments.
Ben Cohen: I’m, I’m building, uh, an internal good innovation agent. Uh, and there’s something I think which is really, really interesting that no charity’s tried yet, but we’re gonna be experimenting with it, where, um, a lot of what like an AI agent needs within a charity or a company is the kind of stuff that’s in someone’s head- Yeah
Ben Cohen: not necessarily like in the CRM or on the database. Uh, it’s that kind of mysterious stuff that is just like in someone’s head, especially if they’ve been in a charity for like 20 years, right? Yeah. Everyone’s got that, like that kind of like fundraiser that just knows everything, like the ins and outs- Yeah
Ben Cohen: and what’s worked and what ha- what’s hap- what hasn’t worked. Um, so I, I’m gonna try, uh, building an agent, um, currently called Keith, working title, that is downloading a lot of my like AI knowledge, uh, of like how we do things, how our projects work, what’s worked before- Mm … what hasn’t, what our like methodology is, um, into that agent that anyone could talk to on Slack.
Ben Cohen: Um- Mm … just as a test to see like whether that makes it more accessible- Mm … um, than, than just me. Obviously, people can still come to me, but, um- Just as a test for the kind of expertise within a business- Yeah … downloading into an agent and giving that the ability to, you know, use that agent if someone’s writing a proposal or something.
Ben Cohen: And that’s like a replicable type of agent that I think we’ll start seeing in- Mm … in companies, charities a lot more.
Simon Scriver: Yeah, really helping that internal efficiency and, and I mean, I think there’s, there’s uses that haven’t even kind of presented themselves yet, and as it develops it’s gonna be, it’s gonna be really fascinating to watch.
Simon Scriver: It already is. But I’m, I’m curious then, as we get more and more of these, do you know, as it becomes almost like standard place or even just there’s gonna be so much more noise out there with the AI stuff. How do, how do we continue to stand out? How do we continue to engage people? I’m thinking of even like the dog influencers, you know.
Simon Scriver: Maybe that’s a bad example, but you know, if, if there’s thousands of agents doing the same thing and reaching out to these dog influencers in slightly different ways, it’s like where, where do we go from here? Is it just being part of the noise, or can we still… Is, is it the humanness that’s s- that’s gonna stand out?
Simon Scriver: Where do we go from here? Yeah. I mean, I,
Ben Cohen: I don’t think anyone has a good answer for that now, but another, another person that Zoshya reach- reached out to, um, was, uh, a big kinda tech podcaster, Case- Casey Newton, has the, one of the biggest or the biggest kind of AI podcast. Um, he runs a platform called Platformer.
Ben Cohen: Uh, and Zosh wrote this great email to him thinking it w- you know, might be picked up in his publication, and he wrote back, “Uh, I don’t work with clankers. Like, tell Ben-” Yeah “… to message me, message me himself.” And I was like, “Oh, okay.” And then I saw a post from him shortly after where- … um, he was, he was getting like, he now gets something like 10 to 20 AI agent pitches- Oh, wow.
Ben Cohen: Yeah, wow … all these people are kind of doing something- Mm … you know, not just like Zoshya but, but kind of getting, using them in multiple different interesting ways. Um, and he’s ignoring all of them ’cause it’s- Mm … a point where it’s like, ugh, I just don’t find this interesting right now. But then we’re gonna get to the point where, and Casey I think talked about it, where he’ll have an agent that will reply to those agents, and that’s what’s gonna start happening in, in different parts of our world where, you know, you’re booking travel and your travel agent then books with the travel agency’s agent, and it’s just agent on agent stuff happening.
Ben Cohen: I think we’re almost there with like, you, you know, you hear of people not using AI well within an organization, right? Someone writing a long email with AI and getting a reply to with a long email by AI, and then everyone summarizing at the end. It’s like what’s- Mm … the end’s ration goal and what, where, where is it going?
Ben Cohen: Um, like I think we’re, we’re probably gonna have to find a solution, and I think, I mean, we will have to find a solution, but I think the earliest places where it’s already falling apart is recruitment
I think we’re
Ben Cohen: kind of seeing- Yeah … you know, AI applications and, and then so many organizations that we’re talking to, like, we’re just inundated.
Ben Cohen: Um, like we even put a post out last year and it was something like dozens of, uh, cover letters had the same first sentence on them. Like, it’s just exactly the same. So I think just it’s quite generic stuff, right? Everyone’s using ChatGPT to do the exact same- Yeah, yeah … the exact same data. So it does, it, it does feel like that.
Ben Cohen: But I think there’s, there’s a difference between using AI in a generic way, where you’re just, like, plugging, you know, a, a recruitment spec into ChatGPT and getting it spat out- Mm … to creating an agent, in my opinion, that you put a lot of time and effort into and, and, and make quite unique. Like, I, I feel like I tried with Zosha at least.
Ben Cohen: She’s just kind of much more interesting than your normal kind of ChatGPT-
Mm …
Ben Cohen: bot. And that’s not the solution, but I think that’s, that’s the beginning of something which is a bit more interesting, if that’s the direction we’re heading in at least. At least trying to make your agent, you know, unique and s- and kind of based on you and the kind of personality you want, and the unique data that your organization might have that would make, make them more valuable than just kind of using it, uh, in any kind of fairly mundane and generic way.
Simon Scriver: Mm-hmm. Yeah, I think there’s 100% that element. I guess with all technology, you know, you’re always gonna get out what you put into it, and I think there’s- Mm … gonna be a lot of people using this badly. And- Oh, 100% … it’s gonna almost… it’s gonna make it easier for… in some ways make it easier for you to stand above and s- and stand out by just putting a bit of time into it and a bit of effort into it, rather than trying to do everything.
Ben Cohen: Yes. Well, you- you’ll see that on, on LinkedIn, right? Anyone can go on there and you can kind of spot if you use AI. Oh, they’ve used- Yeah … Thought for that, or they’ve used ChatGPT for that. Yeah, yeah. And my favorite recent one, which is one of my many stories of just how bizarre and different AI is, did you hear about the, the goblins kind of getting into ChatGPT?
Ben Cohen: No. Have you heard about this? No. It’s quite a strange story. But, um, for one of the kind of personalities you can set with ChatGPT, one was like a nerdy personality. It’s like a setting you can try, and it will, like, interact with you in a kind of, kind of nerdy way. Yeah. One of the instructions they had in that they’ve used for a couple of years mentions use, like, kind of fantasy creatures and things like that, but just kind of like nerdy terminology.
Ben Cohen: Yeah, yeah. And since multiple different models of ChatGPT have been trained on that content and everything that’s come out of it, even the normal version of ChatGPT is always now using, like, goblin-based- … uh, lexicon. Like, there’s weird analogies in, like- Yeah … business reports and, like, scientific reports.
Ben Cohen: Like, look it up. It’s everywhere. And I, I started spotting that on LinkedIn, and I was like, “Oh-” “… this is, like, such a sign that you’re kind of using, using ChatGPT without good instructions.”
Simon Scriver: And- That’s so weird …
Ben Cohen: it’s so strange. That just shows you, like- And same with Zoshia, ’cause I gave her … I, I guess the AI will just take your instructions, like, quite literally.
Ben Cohen: Like, if- Yeah … it’s like a context and pattern matching tool, it’s not deterministic. Like, if there’s goblin language, like it will flow into everything it does and- Yeah, yeah … it, it kinda makes decisions based on, on all of that. And I saw that with Zoshia a bit as well, where I kind of gave her some, like, broad instructions and she took it in quite strange directions sometimes.
Simon Scriver: Yeah. And I, I mean, it’s obviously, like AI is always learning from what’s already there in a way. Do you know? Mm. So it’s always, you know, it, it’s fundraising assumptions, a lot of them are based on our fundraising assumptions and our, you know, the given len- … So I’m, I’m excited to see what it does when it moves beyond that, but maybe that is the saving grace for humans and why we’ll always have a role, is ’cause we’ve always got that spark of creativity that doesn’t exist yet.
Simon Scriver: Maybe I’m being optimistic.
Ben Cohen: No, I’m with you, I’m with you. I, I think, uh, it, uh, I, I’m a believer, like businesses are making decisions to replace people and get rid of workers because of AI. But it does annoy me, ’cause I think all those examples out there are CEOs making decisions for the bottom line of their company- Yeah
Ben Cohen: not because the AI can do the exact same job as that human they’re replacing- Yeah … can. I think the best uses of AI, and the ones we’re seeing with clients, is where it’s like an expert human using AI tools really well. Like, that’s, that’s by far the most powerful- Yeah … space. And as you say, a company being run or a charity being run on generic ChatGPT or Claude, there’s nothing interesting or unique about that.
Ben Cohen: No one wants that. You … It’s, it’s kind of the expertise is almost more important in a way- Yeah … ’cause it’s unique and something that the AI hasn’t got in its training data and hasn’t learnt. So that- Yeah, yeah … sort of stays, that stays kind of human. So I think there’s, whilst roles will change, I, I think it’s, yeah, the expertise plus AI tools that, that really work.
Ben Cohen: Like Zoshia by herself wouldn’t have been capable- Yeah … of doing that. There still needed to be a bit of me there to kind of guide and support and make decisions whenever I needed.
Simon Scriver: Yeah. Yeah. But yeah, I mean, it’s clearly here to stay, and then clearly there’s a lot of opportunity in there if we put the time into it.
Simon Scriver: Um, so I’m really grateful for you coming on and, and sharing a bit about it. Where, um, where can people, uh, see or meet Zoshia? Is it still live? Is it still active now? Are you keeping it active forever?
Ben Cohen: Yes. If you go to meetzoshia.com, uh, you can, uh, see a bit more about the project. Um, that website’s still, still live.
Ben Cohen: So there’s the, the kinda live Just Giving feed there. You can talk to Zoshia there. There’s a kinda talk to Zoshia. You can play- Yeah … you can play, uh, scissors, paper, rock against her. Um, which she’s, she’s kind of rigged, but she can give you more information there. Um, I will say as well, the Zoshia you talk to on the website- is, uh, a clone of the, uh, of the Zosha that I talk to, and that is something I recommend everyone does for security reasons.
Ben Cohen: Mm. Just because, uh, the Zosha I talked to is plugged into quite a few tools. Uh, so we didn’t want- Yeah … anyone on the website kind of prompting her to email someone or, or whatever. yeah. But it is a kind of exact identical clone, so you can talk to her there. Um, she has a z- a, a kind of X account as well, which is @zosha4dogs.
Ben Cohen: You can check out the kind of live tweeting, uh, she did, and you can email her as well. I think she still might be checking her emails. I’ll have to see. Which is, uh, zosha@meetzosha.com. Wow. And it’s Z-O-S-H-A.
Simon Scriver: Amazing. And we’ll drop the links in the description there. And, uh, yeah, I mean, by the sounds of it, uh, she’s gonna stay alive until maybe your wife gets her hands on her?
Simon Scriver: I think it could be the end of it. Um, and you are still fundraising, um, for Dogs For Good. Um, you said you’re… Did you say you’re on the board at Dogs For Good?
Ben Cohen: Yes. I’m a trustee for Dogs, Dogs For Good. They’re based in, based in Banbury near me. Um, and yes, yes, fundraising for them. They’re, they’re an amazing national charity and s- and kind of do assistant dogs for, um, people with dementia, autism, loneliness, everything.
Ben Cohen: And, uh- Lovely … yeah, they’re a really cool charity.
Simon Scriver: Lovely. Thank you. Yeah. Well, we’ll drop a link in for there as well if anyone wants to support them. Uh, and Ben, we find you on, on LinkedIn and all the usual platforms. Uh, it’ll be you, human, giving updates, letting us know your thoughts. No AI on your posts.
Simon Scriver: Am I right in saying that?
Ben Cohen: AI-assisted. Like any- Yeah. All right … good things in post. But yeah, you won’t find any goblins, creeping into my LinkedIn
Simon Scriver: posts. Yeah. Fair enough. that’s it. Thank you so much, Ben. It’s been so lovely to chat to you. I really appreciate your time today.
Simon Scriver: And, My pleasure … yeah, I’m looking forward to seeing what you do next. But yeah, just be careful, please.
Ben Cohen: Thanks, Simon.
Simon Scriver: See you later. Thank you. And thank you everyone for listening. my name is Simon Scriver. It has been a pleasure to be here back with you again. don’t forget, we’ve got our Supporter Experience Conference coming up at the end of May, where you’ll be, we’ll be talking about AI and all things supporter experience.
Simon Scriver:and I’m sure AI is gonna come up in all of our conferences. You can find everything coming up at fundraisingeverywhere.com. That’s it from me. Thank you very much, everyone. Take care. Goodbye. Ba-da-da-da-da-da-da.
*************
This transcript was created using AI. If you spot any mistakes, please reach out. Thank you!