Episode summary

Most leaders don’t get stuck on AI because they lack ideas. They get stuck because they have too many, and no obvious place to start. Keith Stoute, Chief Product Officer at Bursting Silver and the firm’s in-house AI lead, has spent years helping associations, unions, and regulatory bodies move past that paralysis. His answer is simple: don’t try to build the one AI that knows everything. 

Pick a single, focused, measurable use case and run with it.

In this episode, Keith takes Riley through a tour of quick wins that organizations can start on today. He opens with iMind, the member-facing assistant Bursting Silver built for the iMIS Users Group by cleaning up decades of forum Q&A, a project that earned a measurable bump in member retention, and one Keith stress-tested with a “double-blind” trick before any member ever saw it. From there he covers the conference assistant that cuts decision fatigue and drives registration, the subject-matter-expert assistant that answers member questions from bylaws and regulations in seconds (with citations and page numbers), self-hosted AI that keeps member data in-house, and an AI interviewing agent that captures institutional knowledge before it walks out the door.

Two themes hold it all together.

1. keep a human in the loop, AI is a tool the staff use, not a replacement for them.
2. whatever you start with should be simple, safe, and easy to measure.

Keith’s closing advice to any executive director staring down “the AI thing” is to stop waiting for the perfect moment, pick one win you can actually count, and learn by doing.

In this episode

• A prompting trick that turns AI from a chatbot into a thinking partner: structure prompts with #task, #context, and #interview so the AI challenges your assumptions instead of just answering
• iMind: how the iMIS Users Group turned decades of forum and listserv Q&A into a member-facing AI assistant — plus the data cleanup (“garbage in, garbage out”) and the AI-disclosure waiver behind it
  Why a measurable retention bump followed, and why being transparent that members are talking to an AI is non-negotiable
• The case for “targeted and focused” over “the one AI that does everything” — and why narrow scope is the only way you can actually test it
• The AI governance policy as step one: what staff can and can’t use, how new tools get vetted, and disclosing AI to members (Keith notes AI is great at drafting the policy itself)
• Why dumping all your CRM, LMS, and analytics data into AI doesn’t work — and the safer move: extract clean, de-identified datasets and ask for trends you can verify
• Vibe coding without writing code: build a conference website, a join wizard, or a CE calculator yourself, prototype to ~80%, then hand a real spec to your technology provider
• The conference assistant: an AI trained only on one event’s details that answers “is anyone talking about X?” and helps people register before the early-bird passes
• The subject-matter-expert assistant on bylaws, collective agreements, and regulations — turning a 15-minute document hunt into a few seconds, with citations and page numbers
• Keeping member data in-house with self-hosted LLMs — and why that went from ~$10k/month to ~$1k/month
• Succession planning: an AI interviewing agent that captures what lives in staff heads and drafts SOPs without the drudgery
• The throughline: keep a human in the loop, then pick one measurable, safe win to start

Hosts & Guests

Riley Miller – Host

Sales and client success lead at Bursting Silver, helping membership organizations modernize iMIS, data, and AI workflows across North America.

Keith Stoute – Guest

Chief Product Officer at Bursting Silver and the firm’s in-house AI lead. A builder, entrepreneur, and frequent conference speaker on AI for nonprofits — including an AI keynote at the iMIS Users Group — Keith leads BSI’s AI strategy and product work, helping associations, unions, and regulatory bodies adopt AI in practical, governed ways. This is his first podcast appearance.

About Bursting Silver

Bursting Silver is a fully remote consultancy specializing in modern CRM, iMIS, and AI solutions for membership organizations across North America. We’re a 3-time Great Place to Work Certified company that helps associations, unions, and regulatory bodies modernize legacy systems, improve data quality, and deliver better experiences for staff, members, and registrants.

Our unique ability is finding the simplicity in the complex.

Learn more about Bursting Silver

The Modern Membership Org Podcast

Full Transcript

Keith Stoute (00:00)
what are other ways you can measure it? It could be how many inquiries did you field from renewing members last year? You know, was it 200? Well, let’s see if we deploy an AI assistant this year, will the number of inquiries drop significantly? Or how many members waited to the last hour of the last day to to renew to finish renewal? Maybe they started earlier, but maybe it’s hard and confusing for them, so they keep kept putting it off. so that’s another measurement.

roll out the AI and and just compare the numbers year over year. And I would just say pick one that you can measure that’s simple, that’s safe.

Riley Miller (00:42)
What’s going on everybody? Welcome back to another episode of the Modern Membership Board Podcast. My name’s Riley Miller. I’m your host here from Bursting Silver, and I am joined by a very special guest. Once again, we’re bringing in another mastermind to give you some secrets and insights. This time with our chief product officer, Keith Stoute. He’s our AI mastermind in-house, and he’s got a well-decorated record as being ⁓ a a builder, an entrepreneur, and he has plenty to share with you today in the field of AI and beyond.

for for nonprofits. Keith, welcome to the show.

Keith Stoute (01:15)
Thanks, Riley. Great to see you. I wow I had no idea you were gonna build me up that much. I feel like I’m chatting with a an AI that’s, you know, giving you all this wonderful feedback about how amazing you are. So I I I feel right at home. Thanks, Riley.

Riley Miller (01:26)
Well, you can see me now ⁓ in the flesh here, boosting your ego. Well, I and I I mean everything I said i is true. ⁓ I’ve gone on conferences with you, I’ve seen you in action, I I’ve seen your talks, ⁓ you know what you’re talking about and I think it’s it’s an awesome opportunity to have you here to share just even some short nuggets of knowledge there.

Keith Stoute (01:48)
Awesome. Well, it’s true. I’ve done a lot of talks about AI. I did a keynote on AI at the conference in Washington a couple years back, but this is my first ever podcast. So I’m really, really honored and excited to be here with you.

Riley Miller (02:00)
Yes, well, locking down the first. Of many, I’m sure. ⁓ there’s there’s one piece that ⁓ that you’ve shared along with the team too, and it goes in today’s theme, just about AI use cases that might be ⁓ unknown or underutilized. you shared it over a lunch and learn with ⁓ our our firm here, about ⁓ using hashtags to be able to structure your prompts. So hashtag task and hashtag context, hashtag interview to really

chunk it up and then provide just better instructions and better prompts for better results.

Keith Stoute (02:34)
Yeah, I

think that’s a really good you know, I I got it from ⁓ some webinar I attended. So yeah, there’s lots of knowledge floating around on these different topics. That’s a very great way to get AI to kind of challenge you, your assumptions and make you look at things differently and think about things differently. I think a lot of people think of, you know, AI is just a chat bot and you ask a question, it gives you an answer. You can turn that around with that technique that you you just described, where you’re telling it to interview and challenge you. so it can be any any kind of strategy you’re thinking of or any

any policy or procedure you’re you’re rolling out. you can have it, you know, again, y it’ll challenge you to defend yourself or what you’re proposing to do to make you look at it in different ways.

Riley Miller (03:13)
Yeah, a and that’s important too. I think ⁓ just like any tool, you you’re not you’re using it and and building on it and iterating. it’s getting smarter by the day, but it’s also becoming better as a idea generator too. You just gotta make sure that you’re challenging it when you can.

Keith Stoute (03:30)
Yeah, great for brainstorming, great for ⁓ being your your collaborative thinking partner. have it flesh out new ideas with you, ⁓ and as we said, challenge your ideas too.

Riley Miller (03:41)
Yeah, and check your work. I I know that ⁓ it one item that ⁓ we we’ve highlighted in an episode previous was your work on iMind, and the iMind solution for the iMIS Users Group, which has become a huge resource for that community. I just was wondering maybe to kick off our episode here, if you could walk us through what iMind is and the problem it solves for the iMIS users group, like how that came to be.

Keith Stoute (04:05)
Yeah, thanks. Sure. I’d love to talk about that. That was ⁓ it’s one of my favorite AI kind of implementations we’ve done. And I think one of the key things that make makes it so successful is we had a very targeted and focused approach of what we’re trying to accomplish. And that’s what we’re always recommending clients when they think about AI, and AI can be so many things and it’s you know changing the world and all these different things. Just try and hone it into a a practical solution or application you can implement. so iMind in particular was actually the brainchild of

Michelle Morgan, the executive director at iMIS Users Group. And it’s an association and its members are, you know, organizations that are using iMIS. And so they can become a member and they can collaborate and network and and really kind of help them further their experience with the iMIS technology. And they’ve been around for quite some time, and part of their offering has been for you know a couple decades, is a user form and/or listserv where

Someone who has a challenge with iMIS or is trying to learn iMIS or sees an opportunity, they might want to try doing something with iMIS. The form allows them to interact with other members and ask a question. So a typical case might be they’re they’re getting an error or they’re having a challenge, they feel stuck. And with this form, they can go and post that question. And people see it in their email or they log into the member portal and they see that. And there’s a back and forth, and inevitably, usually somebody comes up with an answer.

that question has been quite effective. ⁓ and so Michelle’s vision was that they had all this and along with other knowledge they have at their organization, what about if they could take that information, all these questions and answers that have been posted for a long time now? Could they use that to train an AI that would allow their members to actually now have a chatbot with that information so they could kind of get real-time answers to the questions that they have about iMIS

So that was an exciting challenge for us. And ⁓ we we took that on and and built that for them. It was a multi-step process because one of the key things when you’re we’re trying to think about an AI solution is what are you what information are you gonna give to the AI so they can do what you’re asking to do very well. One of the key challenges we have with this sort of ⁓ project is that a lot of the back and forth between the members was done via email listserv. And so you can imagine ⁓

any email you open has a lot of information in there that’s not relevant to the in this case iMIS questions and answers. ⁓ but the email signatures and some emails you know could be paragraphs. so if you just grab all the email information you dump it into an AI assistant to have it trained on that and and set it loose to the members, it’s gonna it’ll you know garbage in garbage out, right? I think we’re all familiar with that term now. So we had to clean up that data.

⁓ and do several different steps to get it to a a level where it’s actually very good at what it does. But now members can use it and they are seeing quite a lot of the people that are aware of it are using it quite frequently now. Michelle has indicated they did see an uptake in member retention because it’s just another value added benefit. You have to be a member to access this. So it has been quite a success and and I just participated in a webinar one of the one of the ⁓ members put on a webinar sharing her experience with it and

Sharing her successes with it. but yeah, some some really good lessons there too. I think ⁓ you know, what what they did as well is made sure that people understand that they’re interacting with an AI. I think that’s something you know all our clients are member-based organizations. So some things to think about is ensuring that if you are having your members or the public interact with an AI, make sure it’s clear to them that it is. In this case, because it’s actually giving people advice or helping them maybe solve problems or

realize opportunities, we put it behind a a waiver, like a form that someone had to sign indicating they understand that it’s AI and it’s not going to be a perfect answer. ⁓ just to protect them even legally, but also to make sure that the members understand that you know, if if the AI gives you some bad advice, don’t just follow it blindly. So anyhow, that was that was a great project.

Riley Miller (07:58)
Yeah, and ⁓ I think you you highlighted a couple of points there about even from the iMIS user com community there it improving their retention by offering ⁓ a new resource. You’re giving some transparency efforts in terms of using this technology and not pretending it’s something it’s not. It’s a tool and it’s a it’s an outlet, but ⁓ it ensuring that there’s also that ⁓ level of ⁓

⁓ honest ⁓ approach to it as well. It’s it’s not always perfect.

Keith Stoute (08:28)
Yeah. Yeah. And I think, you know, like I said earlier, the objective was quite targeted and focused. And I think that’s also why it was so successful. If you if you’re thinking about rolling out an AI solution that’s gonna kind of be the be all end all, know everything that your organization does and be an expert in every policy and every procedure, I don’t think you would have as as a successful an outcome. ⁓ so I think, yeah, that’s what we’re always trying to strategically guide our clients to understand that

The the best way to succeed with AI is to have very clear, defined objectives of what you’re trying to accomplish.

Riley Miller (08:59)
Yeah, I I mean this is just ⁓ coming to mind too because of the the talk that I did about chatbots at ⁓ the iMIS users group. It was actually a proposal you put together to to train about ⁓ how to implement AI at an organization starting with a chat bot. I think some of the foundational pieces there was starting with your why or what you’re answering before you even get building. ⁓ I think that was

key instrumental in in terms of how to adopt this technology. Are there any other points that you would raise for teams that are looking at adopting certain tools or technologies?

Keith Stoute (09:36)
Sure, yeah. Well, ⁓ I mean, I think having an AI governance policy ⁓ is the first step to take. And it doesn’t have to be a complex, long ⁓ policy paper. It really is what you’re allowed to use AI for, what you’re not allowed to use it for, what tools you’re allowed to use, and then talking about what the staff are allowed to use or not use. what to and what the process is for if staff want to use a tool, what is the process to getting that vetted and approved? some of the things we kind of

Touched upon as well. ⁓ it’s important that if the members of the public are interacting with an AI, it’s important that they are aware that they are and don’t try to fool them into thinking that they’re they’re talking with a a person. So, you know, an AI is actually really good at helping craft these AI policies. So that would be the first thing to have in place. but I think having a very targeted and focused objective, like I’ve already said a couple times now, is really key if if ⁓

it was important for us with the IMind to test it thoroughly before it was used. And if you’re trying to have one AI that knows everything and does everything, how are you even going to test that effectively? What was a nice opportunity for us to so we know obviously we know iMIS very well. So we could do some of our own questions and answers into to iMind before we let it talk to the members or interact with the members. But ⁓ the next level of testing that I was able to do was to actually kind of

Lurk on the chat group form. And then when people posted a question, I would behind the scenes paste that question into iMind before it was released and get that answer and then paste that into my response as if it was coming from Keith. And then a few days later I’d follow up with that person and said, Hey, just curious, that answer I gave you, did that help? Was that was that a good answer? And then and people would say, Yeah, that was that was good. So it was kind of like ⁓ blind testing, right? Like a double blind study testing. people

Didn’t know that they were getting the answer from an AI and they still scored it ⁓ that it did a good job. So that gave us the confidence to know that it was doing a good job. So being able to test and validate that the AI is giving good answers before you obviously expose it to anyone else is is one of the key things too. So again, if it’s targeted and focused, you can do that. If you’re trying to make the one AI does everything, it it’s just gonna be ⁓ you can’t have confidence that it’s gonna give good answers all the time.

Riley Miller (11:50)
Yeah, I like how you’re running it through the Keith filter too before you send it. Great strategy.

Keith Stoute (11:54)
Yeah. It

was a bit sneaky, but it was fun.

Riley Miller (11:57)
Well, I so we think about ⁓ all of the use cases that have been unlocked now just with the advent of AI. One idea that comes to mind is being able to interpret large sets of data. just when it’s too big to to research or or you have all of this information, how to find patterns or trends within your membership.

In your experience, when you hear about distilling a lot of information into an AI system, what what are some of the thoughts that come to your mind?

Keith Stoute (12:27)
Sure, yeah, like I think to me it’s a bit aspirational, at least for for our clients to to think of you know, it’s a bit of a dream, right? You take well, let’s just grab data out of our CRM and then the data out of our LMS and and throw in the website analytics and email analytics and and even a few spreadsheets and throw that into AI and have it come back with like some grand vision or trend that it’s identified throughout all these patterns. I I haven’t seen that work, I think.

You know, if you hook all those things up into Power BI or something like that, I think it has some AI capabilities. but just me personally, I I that’s not a strategy I would recommend to to you know our clients. there’s a lot of questions we haven’t touched upon this yet, but ⁓ there’s a lot of I would have a lot of concerns about the the privacy of the of the data. because it you’re you’re talking about a lot of private member data. and if you have all that data thrown into the AI and it comes back with

you know, some sort of strategic plan or some great analysis it says it’s done, how do you even validate that it is giving you good answers, right? most likely if you’re just throwing a lot of data into it, it’s going to get confused. It’s gonna prioritize some tables in your structure and some data that is actually irrelevant and give you strong recommendations based on something that’s irrelevant. So again it comes back to garbage in, garbage out type ⁓

That’s that’s the philosophy. so I I I’d instead of doing that type of thing, I would suggest that you you could extract key data sets from your systems and remove any any sensitive data, like personally identifiable information of your members, and you could send up a few spreadsheets from those and ask it for it to look for trends or to you know do some analysis with you on that and then it it it’s high quality data.

And you can also ask it to validate, you know, why are you recommending this or what are you seeing? So yeah, I think, you know, I as the AI technology is better and better and it becomes more embedded in the tools that we’re already using, I think those opportunities will come more and more. But I just don’t think I I personally haven’t seen that succeed.

Riley Miller (14:28)
Yeah, and wise words, I think that’s something that might just be overlooked. We come into these ⁓ big new shiny tools ⁓ with the hopes that we’ll be able to unlock and ⁓ uncover new trends. But it’s important to still maintain that trust with your membership that you’re not gonna be feeding all that information out into the the machine So ⁓ yeah, coming back to that point that you said as well, starting with why and having your your objectives pretty clear.

Keith Stoute (14:50)
Yeah. Yeah.

Riley Miller (14:57)
in instructing any adoption or or use of AI help really helps instead of throwing caution to the wind.

Keith Stoute (15:04)
Yeah, yeah, exactly.

Riley Miller (15:06)
So for for other use cases, ⁓ another term that came up, and I I saw this at a another recent Innovations Road Show stop, the use of vibe coding. Now, vibe coding is ⁓ sounds like a trendy word, ⁓ but a lot of leaders I I think nowadays still assume that when it comes to d designing anything on their website or or spinning anything up, it it means hiring a developer i with this new

vibe coding approach. What do you what in your opinion has changed with what an individual and organization can realistically spin up now?

Keith Stoute (15:38)
Yeah, ⁓ vibe coding really emerged, I don’t know, I think within the last year. And ⁓ it was the ability for people that were coders to kind of now not do the most of the writing themselves and they could build an application really quickly, maybe you know, in a few hours versus weeks. ⁓ but even since then the technology’s come along so far, I don’t I I think the term vibe coding needs a a a replacement because you don’t need to know any coding whatsoever now. and I think

people thinking, well I can’t I can’t vibe code because I don’t know any code. that’s that’s no longer the case. You can just interact with these systems. You know, cloud code is a good one or cloud cowork. you can literally just have a conversation with it and and describe what it is you are looking to do. And it will create it, you know, right in front of your eyes. So you can build an entire website without writing a single line of code. so where are those opportunities for our clients? Like you could they you could build your entire annual conference website yourself.

you know, and it can be beautiful. you can link it in with the data from your event management system. So it’s there. You can link it into your registration system. and and you can test it for all the important things like you know, mobile devices and AODA or WCAG type ⁓ guidelines to make sure it’s accessible by everyone. so there’s just so many ways you can now start to surface if if you have an idea to actually make it a ⁓ reality.

digital reality, but a reality. you know, other other things you could I could see our our clients using it for, or customers using it for. could be like a a member, like right now, a lot of let’s say an association that might have a join process that speaks to everyone the same. Whereas you could have like a wizarded approach where you get taken through an entire journey based on, you know, filling in just a couple questions about who you are or what industry you’re in or what stage of your career you’re in.

You could present a whole different series of you know reasons why to join and benefits, and that that’s really appealing just to that targeted segment. what else can you do? I a lot of our clients that are more regulators, you know, they’re they’re they’re their registrants or their members have to do a really complex, can be quite a complex ⁓ continuing education curriculum or program to stay qualified to practice. And this can have a multi-year cycle, and there’s lots of different learnings and activities they do, and there’s a point system.

It can be pretty complex for the for the member to understand. You could just build your own calculator yourself, right? Again, these are things that you can do without hiring an external technology consultant that you know they’re not cheap. And and you could do pretty much the whole thing yourself. Now I would I would I would caution that when it comes to actually reading or writing data from your system, you wouldn’t want to, you know, vibe code that solution yourself. But you could

To build it yourself, to get it through the, you know, to circulate it with the stakeholders at your organization or have actual members that you kind of trust to play around with it and give you feedback on it. You can do all those things yourself. You could extract data out of your system, you know, again, remove any personally identifiable member information, but have that as a data source for your experimenting, for your prototyping. And then when you’re ready to hand it off, then you you’ve kind of handed off a full, fully qualified and spec.

solution to your technology provider who can then make it a reality, make sure the security and and and you know the real crucial things like security and performance are there. ⁓ but you’ve just done about eighty percent of the work yourself. So it’s it’s kind of like a a superpower, right? When you can think that now people that don’t have any technology background can start building these applications themselves almost end to end. yeah, it’s it’s pretty amazing what it can do.

Riley Miller (19:11)
If you can dream it, you can build it, right?

Keith Stoute (19:13)
Yeah, yeah,

definitely. Well said. Well said.

Riley Miller (19:16)
Well,

so it’s one of the major resistance for a seamless membership experience we’re seeing at ⁓ a lot of these associations and and regulators is being able to find the right information easily and efficiently. Just from a conference example being able to find out what hotel you’re staying at, if there’s a group rate or what the the early bird rate is. If the the list goes on for for that

How does adopting an AI assistant, for example, it support just that that flow of information and and what separates ⁓ having a a a generic assistant or chatbot versus an actual embedded assistant?

Keith Stoute (19:57)
Yeah, I think that’s one of the most popular use cases with the clients that we’ve talked to is, you know, we call it a knowledge finder or sort of a subject matter expert. Again, not an AI that’s gonna know everything about everything and answer that to every different audience, but ⁓ something prescribed, an assistant that’s available in the context of that information that it’s supposed to know about. ⁓ so you gave a good example about a conference. a lot of organizations or clients have annual conferences, and these can be pretty complex.

⁓ you know events that last several days and might have different tracks and have a hundred different sessions and there’s member pricing, non-member pricing, there’s parking, there’s accommodations. ⁓ to have an assistant that’s really trained on all that information can help, you know, get people to register. There’s a lot of you know decision fatigue and trying to try ascertain ⁓ when you’re approaching a thinking about going to a conference, at least I know when I’m thinking about going to a conference, to find out if the topics are going to be of interest to me.

can take a lot a fair bit of time because there is so much information and so many different sessions and tracks. Whereas if you have an assistant that’s trained on that specific conference, might not know anything about you know membership or certification, but it’s there on that annual conference website. So people aren’t going to be asking it those questions there. It can help, you know, really drive registration up to the conference because people can just say, well, is anyone talking about you know artificial intelligence and member-based organizations or whatever it is I’m interested in.

And I can I can find out and validate it is interesting to me or where it is or how much it is really quickly, that should increase the amount of time that people choose to register then and there versus putting it off, and maybe they put it off too late and they don’t get the early bird or they just ⁓ they sign up for a different conference where they they can find out that the of of you know it’s gonna be a value to them. so that’s one really good example. what we also find customers find are they’re benefitting by AI with in terms of

Finding knowledge is just like the sheer volume of information that our clients have as part of their well, part of their mandate, right? So an organization might have a constitution, ⁓ they have bylaws. If they’re a union, they’re gonna have ⁓ collective agreement. and this this you know, this is quite significant, you know, the number of documents they have and the government regulations they might be operating under as well. And a lot of the information that their member needs from them is is is contained inside of their.

These can be hundreds of pages. ⁓ so what does a AI solution look like to help in that case? Well, it’s trained on those and and really just those, right? The specific core documents that ⁓ run the organization or or or the organization is mandated to operate under. and a member will inquire, like, well, you know, if I retire a union, they might have an inquiry about a member who wants to know if they retired a certain certain stage or a certain pro.

stage of the career, how does it impact their benefits for their retirement? Or if it’s ⁓ a regulator, maybe the someone retired and they want to come out of retirement, how do what do they have to do to qualify to practice? Well, these things are all documented and written down, but it might be on page 149 of a a 300 page document. The inquiry goes into the staff. The staff have to research that, find that information, they have to craft a response to that member and and that can take a long time. It can take

a lot of time there. But you can have an AI assistant that is fully knowledgeable about those documents and you can you know copy and paste in that inquiry from that member and that AI can get back to the staff with a crafted response in seconds and also cite the document and even the policy and the page number where it was found. So you’ve taken something it could take 10 or 15 minutes and and turned that in just a few seconds. And so you can see the multiplying effect it can have on frame

The staff up to be more productive and do something that’s more value added for the organization. So that is I think a really powerful way to use AI to help serve members better and help empower staff. Now I I think it’s important to note that you have to be careful with that type of application because what I just described there is taking the inquiry from the member and putting it into an AI. and you need to make sure that the the data.

the AI tools that you’re using are are protecting that member data. There are options for that. we’ve built some and ⁓ you know you have tools that can actually redact some of that or all of that or almost all the the personally identifiable information to make sure that the member data ⁓ doesn’t leave. What we’re really starting to look into now and I’d be happy to talk about it more is you can actually host some of these AI solutions yourselves these days. And I I don’t want to get too technical but

⁓ typically when we think of AI, what we’re thinking of is large language models. Like that’s kinda like two and a half years ago ChatGPT came out and it kind of changed the world for everyone and understanding of what was possible with AI. Well, that’s a large language model technology. And there’s sellers several of them we’re all familiar with now. There’s ⁓ OpenAI, there’s Anthropic, and Google has Gemini. and we think of these as, you know, big, huge cloud providers and the big huge data, AI data warehouses that they’re building, you know.

Contra you know, some controversy around around the world. ⁓ and so when your staff are sitting interacting with an AI, the the information they type in there gets sent to one of these big huge you know, big tech firms, VC backed tech firms. That doesn’t have to be the the case now. We’re actually starting to roll out AI solutions that are hosted on a client’s server. So ⁓ an actual large language model can be hosted ⁓ by

you know, your your own servers. You can do it yours do it yourself, LLM. are these gonna be like as good at answering any question in the world that are they do the do these large language models that you host yourself do they have a PhD in every topic under the sun like you know the latest big tech LLMs do? No, they don’t. It’s but all it has to do is find that regulation that is corresponding to the members inquiry about you know coming out of retirement or their benefit retirements. It just has to know a PDF really well, which you provide to it. So

I’m I think you can tell I’m super excited about this because it allows our clients to have AI and expose member data to it and that member data doesn’t leave their their infrastructure. So ⁓ yeah, that’s kind of what that’s kind of like the next thing. And the reason that’s possible now, when we first looked into it about a year ago, it was gonna cost us about ten thousand dollars a month to host one of those, it’s a GPU server. that cost is dropped dramatically. so now it’s about a thousand dollars a month to host something that can do that.

⁓ and then if it’s an internal facing for your staff to use, you know, you probably only have to run it eight hours a day, right? so then you cut that cost ⁓ down by two thirds. So these are some of the things we’re we’re excited to roll out for our customers. ⁓ I think yeah. I I I get excited about this type this type of thing.

Riley Miller (26:39)
Naturally.

The AI mastermind, I can see it.

Keith Stoute (26:43)
Yeah.

Riley Miller (26:44)
Well, in the interest of time, you you did mention a good note here too about having core documents and staff with ⁓ that servicing members, but when we look at staff too, so much of an organization lives inside the staff members’ heads, the the seasoned war vets that have been in the office for years. however, nowadays ⁓ how do you see AI ⁓ as an an opportunity

to be able to support effective succession planning for maintaining institutional memory and onboarding for new staff members.

Keith Stoute (27:14)
Yeah, I think that’s that’s really important that you know we’re all subject to so much change these days from technology but just just in society in general. So organizations that may have had a an operating manual w ten years ago probably is completely irrelevant to to their operations now and and people are kind of doing a lot more, you know, just triaging and fly by the seat of their pants as as they they’re asked to do more with less resources. but that is a challenge when someone new comes on board or someone leaves you know.

Trying to understand what it is they’re actually doing per department can be hard. And I just think like there’s so many, so many applications AI can help with. In this example, I would I would say like you can spin up an AI kind of like an agent that would act as ⁓ an interviewing agent. And you could just have it just have each staff and each staff member in each department.

be interviewed by this AI, just a chat format. And the AI could be structured to solicit information about what it is they’re doing on a daily basis. What are their key standard standard operating procedures that they’re doing, SOPs? When a new member applies, what are the steps that you take to do that? So it’s going to impact, you know, the CRM, maybe emails get sent, maybe there’s some word templates or Excel files that get filled in, set up in an LMS.

the the AI would interview each staff member individually, and this would just be casual, right? Just the way you and I are chatting. Maybe typing, but it would be it would be chatting. and you do that for each person, and some common elements are going to keep coming up again and again, like the CRM and and ⁓ the email templates or word templates. You do that across the entire department, for example, and you could then take each response and roll that into an AI that would consolidate that and

Write a a a draft of an operating procedure of what is happening at that department. It wouldn’t be perfect, but it would it would be kind of get you 80% of the way. So again, look for ways for AI to really kind of leverage your your team’s knowledge without giving them a lot of manual work. Like nowadays, if you sit someone down and say, well, write the the the manual procedures for your department, like I can’t think of anything more boring and painful.

AI can make that you know much easier and can do a lot of the the grunt drudgery work. And that that’s always like what we’re always trying to help our clients do. Get rid of the drudgery data entry, kind of boring work that AI can take on and kind of grab that structure of, in this case, the different people, the different roles that they’re doing at the department, and write a pretty good draft of the operating manual for that department.

Riley Miller (29:47)
Yeah, and I think too, in all the examples that we’ve mentioned today, it is important to also highlight that as AI i is being positioned as a tool, an enablement and opening up staff for more valuable work, it’s still maintaining the human in the loop. It’s keeping that one team member as an extension or a tool of their position to be able to service those things.

reviewing policy docs or or spinning those things up, there’s still a person that has to execute it and proof it and v validate it and do the the Keith filtering, s so to speak.

Keith Stoute (30:20)
Yeah, yeah.

It de it definitely is just a tool that the staff get to use. It’s not replacing staff. And you still need the human being who’s responsible for the output. So yeah, each example I gave it isn’t the AI just firing off responses to to the member of the public on important topics like retirement or that type of stuff, but it is helping the staff get that member ⁓ get that information quickly and help them craft the response just to s just to save them time.

I mean not that you can’t then also have like the iMind is a is a is a bot that is facing towards the their members. but like we said, you have as long as the members understand that this is this is an AI giving its best kind of guess at at a good answer. and then if it is some of a more important topic, it’s more about helping them find that information quickly.

By citing the p the file name or the page number, and you can link directly to the PDF and the page number where the AI is finding that information. So if you do want to have member doing members of the public even doing some self-serve with AI, it’s important that the AI is in position as the authority. It’s is as an it is an assistant to help the member, if it’s in this case self-serve, find that information, which is from the official source, which would be the actual webpage or the PDF where that information is contained.

Riley Miller (31:34)
Exactly. I just looked at the time. ⁓ what a quick way to burn half an hour. Keith, it’s been excellent talking to you. I I often like to to kind of end on ⁓ more of a speaking to an executive director level or if there was a person that came to you and they’re they’re looking to adopt AI or they’re looking for some sage advice from from an expert in the field, what would you say to that that leader and a non profit?

Keith Stoute (31:36)
Yeah. I know.

Yeah, like to get started, a lot of you know, that’s it’s kind of analysis paralysis to a certain degree because there’s so many different ways you can use it. And it’s kind of a scary topic. I mean, all of our clients are not for profits and they and they’re they you know the in the for profit world there’s this pressure that you have to adopt technology faster than your competitors or you’re gonna be put out of business, right? But you know, that’s not the world that the member based organizations, unions, and regulatory bodies, associations maybe to a larger degree, ⁓ live in.

But

I do think it’s important to do something because it’s learning to use it, right? It is a learning process. It’s not you can’t just wait and expect that someday you’re gonna turn it on and your organization’s gonna, you know, nail it out of the park, or some technology is gonna be just so great fit for you that you just wait till that technology comes, you turn it on and and you’re done with this whole AI thing. You’ve solved the AI challenge. ⁓ I do think, you know, as I said, like find something that’s measurable and focused where you can

see some success and measure that success. Like ⁓ an example I like to give is is a lot of our

clients have very complex member renewal processes, right? If you’re ⁓ a health professional, the renewal process for you can be, for the member can be quite arduous because there’s just so many things, you know, I mean it’s good, these are regulators. We gotta make sure that our healthcare professionals are are are meeting the standards that they have to, very high standards, and that that renewal process covers a lot of that type of embedding.

But it’s complex, right? And these are a series of forms, often could be a dozen forms with a lot of different information and conditional logic within them. If you’re this and you’re that, then you have to answer this one and this, but not this one under this context. and so ⁓ we’ve had clients that then like create a large PDF that helps their members use it. Well, how about instead of a large PDF that could be a hundred pages, have a little AI assistant on the renewal pages that

Help answer questions just about renewal. Like I said, it’s focused, right? It’s not knowing everything about your organization or your bylaws or everything like that. It’s just, I really know this renewal page really well. Ask me any questions about it, and I’ll give you my my best answer. You only have to train it on that large PDF

knowledge. And then measure it, right? So, like what what do we mean by measure? Well, you can you can of course you can test this well in advance because you can throw all sorts of questions at that to make sure it’s going to give good answers before you turn it loose on the members during renewal. But

What are other what are other ways you can measure it? It could be how many inquiries did you field from renewing members last year? You know, was it 200? Well, let’s see if we deploy an AI assistant this year, will the number of inquiries drop significantly? Or how many members kind of like waited to the last hour of the last day to to renew to finish renewal? Maybe they started earlier, but maybe it’s hard and confusing for them, so they keep kept putting it off. so that’s another measurement. How how

And within the period of renewal, how many kind of renewed before the last hour of the last day, or how many had to, you know, were late and had late fees applied, roll out the AI and and just compare the numbers year over year. And ⁓ just there’s just so many opportunities. I would just say pick one that you can measure that’s simple, that’s safe. doesn’t, you know, don’t start with you know the example earlier you gave where it’s just a huge data set doing a huge analysis for you.

do the one that doesn’t like the example I gave, you don’t have to expose any member data to the AI. It would never, you know, you you structure it in a way that it doesn’t know anything about who the members are, their names, or anything like that. It just knows that giant renewal PDF that you trained it on and answers questions about renewal really well. ⁓

Hope that helped, Riley.

Riley Miller (35:28)
Yeah, absolutely. Well, I Keith, I really appreciate the time. Thank you for coming with ⁓ so many nuggets of knowledge. And ⁓ I’m sure the way AI is moving so fast, there’s gonna be plenty to talk about even in twenty four hours. So we’ll we’ll definitely have to have you on here again.

Keith Stoute (35:42)
Yeah, yeah.

Well, looking forward to it. This is this will be my first podcast. It’s been a lot of fun. Thanks, Riley. Probably no one else would invite me, I bet, but we’ll see. Okay, great. Cheers. Thanks.

Riley Miller (35:49)
It could be your second as well, who knows.

No, there’ll be more, I’m sure. Right on. Thank you so much, Keith.

All right, that is a wrap. Thank you so much for tuning in to the Modern Membership Work Podcast. My name’s Riley, and that was Keith Stoute. Thanks again to Keith for spending the time to give us some knowledge bombs on AI. If you enjoyed this podcast and you want to hear more, please consider subscribing on anywhere that you adopt or consume your podcasts. And ⁓ we release episodes weekly. If there is a topic that you would like us to cover, feel free to reach out. ⁓ info at burstingsilver.com. We would love to hear new ideas or even new speakers. till next time.

We’ll see you then. Thanks again.