AI for Specialty Business Growth: Automate, Optimize & Scale Smarter

Shawna: Good morning everyone. 

Welcome to today's webinar. We're so glad that you're here with us. It's a beautiful morning here in the Pacific Northwest. We hope it is where you are at as well. And I'm really excited for today's conversation. I think a lot of you are as well. 

Joining me is Jeremy Kennedy, our director of development here at Evosus. And we are in for a treat. Really fun conversation all about AI. I know a lot of people are talking about AI. I think this conversation is really close to home. We're going to talk about just the world as a whole and how AI is affecting the world. Where it started, where it's at, where Jeremy and team foresees it going. And then we'll get into a little bit of what we're doing here at Evosus and how we're thinking about AI, what our policies are around it, what the future looks like for our software and how we integrate AI. So, a lot of things to cover this afternoon and morning depending on where you are. 

If you have any questions through today's conversation, there's a little chat bot at the bottom of the screen. Please feel free to type in your questions. If you want to type them into Q&A, that's always really helpful for us. So, at the bottom, there is there's some little shapes you can click on and then there's a Q&A button that you can click on. Shannon Stein will be moderating today's conversation, so we'll pause about halfway through and then at the end for questions. So, please type in your questions as they come. We'll try our best to answer them, but if there's anything we can't answer for you, we'll be sure to get you the answer afterwards. So with that, let's get this thing kicked off. 

Jeremy, I'd love to start with just a little bit about you and who you are, what you do, what you do here at Evosus. So, can you just give us a little bit of insight on your background, and what led you to your position here at Evosus? 

Jeremy: Totally. So, I'm Jeremy Kennedy. I've been with Evosus for just under two years. Plan on being with them for a long, long time. Plan on retiring here. I have a background in engineering as well as a background in unrelated field legal. But I've been in engineering for the greater part of 15 years. I love everything about programming. I love designing and developing. I also have very many hobbies where I love building things. So I build a lot of things. I'm a tinkerer for sure. I also really am into the understanding of business continuity and improving people's lives through technology. I feel I can make a difference and an impactful difference on not just the way things are done here internally but the customers we serve or the clients we serve through helping build better products because I understand that we all have mouths to feed and everything I do is very impactful on everybody else. So it's very important and I take a lot of pride in my work. Knowing that what I do improves a lot of lives, not just internally but externally.

Shawna: It's so true. I think first of all, I'd love to hear what's one of your favorite things to build?

Jeremy: I love to code. So I am a midnight coder whilst raising three children. I do program when they go to bed almost every day. I also love building cars. 

Shawna: So cool. So fun. So physical like physical touch, hands-on stuff, and then also digital. I think that's really incredible. 

And I think I mean what you said really resonates with me because that's exactly what we're doing here. We're helping business owners. This is their livelihood. This is their legacy in a lot of cases. And I think it's really important that in your role leading a huge department, there's a lot of people, a lot of developers on your team. That everybody's on that mission to just help make people's lives better, build software that helps them run a successful business and pass on their legacy. So that's really incredible. 

I'd love to talk about AI. And I think something selfishly that I would like to know is is AI new or how long's it been around and what's been the catalyst that's just made this thing explode over the last what feels like 18 months. 

Jeremy: Yeah, it's newish in some regards, but it's actually been around for quite some time. The theories have been around. Movies have helped drive reality as we all have seen movies where we make us fear it a little bit. It's true. But it's been around since IBM first explored it and really took it off in 2004. So that was close to over 20 years ago with IBM Watson. And then Nvidia, which is a graphics card company, had their model of Luda and then they had another one that came out in 2012. But it really started taking shape with the foundation of OpenAI.
The foundation which is a nonprofit Open AI is the company that is responsible for chat GPT. And so Chad GPT came out in 2022. And then immediately following was Google Gemini in 2023 as well as another language model that's popular cloud which is claude, excuse me, which is more of a technical model. Chad GPT encompasses a lot more in a brighter broader range of answers and those kinds of things. 

The reason why it's really taken off recently has a couple factors. It's the underlying hardware to run it. So computing power has been a big one. When it first came out, computers were what we may have thought of them as revolutionary at the time. Yeah. But if you look back at a computer used 20 years ago, you're like, "This thing's a dinosaur." And that's what we had to work with. So as computers have advanced they have the ability to process more and faster and and and then and that feeds into the second segment which is large language models which is mass data.
So, so big big data and big big data has been a key to it too because these as they're called models that are trained which give you the answers when you ask some when you ask it something those are from trained models of big data and so over time data itself it's like compound interest and so when you train it and you and you give it an answer you're like okay it knows a little bit more a little bit more a little bit more it starts picking up trajectory just like inertia and speed and then eventually it is able to start training itself. 

So it can take and it can use basically its own ingest its own way of learning and it can learn from the things it does and every time you ask it a question it starts training it on those questions automatically so that way somebody's not behind the scenes having to train that model of that question. So it ingests what you ask it. It goes and scours all the different sources it has. It gives you an answer and then it takes your answer and adds it to the database and so it just keeps compounding and the more that people use it and the more it's used the faster it learns. 

Shawna: So that I mean that makes sense to why I think we all feel like it's taken off and it just is moving. This train is moving so fast but it's been around since 2004 you said in some capacity. It's been a really long time. 

 that that all sounds you could I think you could think of it two ways, right? You could think, "Wow, this is really empowering and this is really cool." Or for a lot of us, we're like, "Holy smokes, that's kind of scary." Is there anything with that that we should be concerned about? or what should people be thinking about knowing that these language models the more we put into them it's going to take that information and basically use it as a you know course for itself. 
Jeremy: Yeah, there are some concerns and considerations but thankfully like companies like OpenAI they have initiatives since they are not-for-profit and their goal is to help drive AI in an ethical way. They do have core standards and they do have guardrails in place that are to help us ingest proper knowledge. But with that some things still are always going to slip through. So AI is you have to think about bias and fairness that it gives you. 

Sometimes an AI model is only as good as the way it's trained. So if it's trained on something that is untrue or biased, it has the potential to give you untrue or biased answers. Interesting. Interestingly, Microsoft faced this issue when they released a product. I believe it was in 2015 or 2016. Their first initial run at an AI powered chatbot and it quickly shut down because it gave biased answers. So that was a big public facing. Okay, we need to put more guard rails around it.And then so very very shortly after open AI was founded with those standards in mind in place because of that. 

Shawna: Smart. Yeah. Learn from history and don't make that mistake again, especially in a public setting being a large company. That makes me feel a lot better because I know I know there's a lot of concerns, you know, rolling around out there about legislation and laws and we need to put guard rails around it. 

So, it sounds like some of those things are already in place via OpenAI and hopefully other companies, one I think they've moved really quickly in comparison to any other company that has an AI initiative. So, that's good for us that they're leading the way in a positive, you know, a positive standpoint. Let the good guy go first and then everybody will hopefully follow that same track if they want to grow as quickly as OpenAI and chat GPT have. In that vein, what would be some guidance you would give to our dealer network, to our clients when it comes to using AI?
Is there anything that they shouldn't use AI for? Is there anything that they should maybe create a process around within their business? Obviously people want to use it responsibly and want to use it to better their business. Like we talked about earlier, these are generally family-run businesses. This is their legacy. They don't want, you know, they don't want to do something thinking it's for the greater good, knowing that it could potentially harm them. So, any guidance you would give to these business owners when it comes to using AI or adopting it into their business? 

Jeremy: Definitely. Generally AI is a tool. It is not a replacement. So, a myth versus reality is people think AI is going to replace our jobs. It is not a tool and we'll consider it to be a tool. It may transform the way we shape our work and it may cause us to upskill our labor force. But it will in turn actually create jobs around it. 

but in tune to your question, definitely one thing to not do, don't do trust falls on it. Don't blindly generate something and say, "All right, we're good to go. Ship it." Or ask it for statistics on something or ask it for advice, life advice, and just follow it blindly.

Shawna: So, no legal advice. 

Jeremy: no. I mean, it'll help you cite facts. Okay. But you want to it's it's it's it's essentially not trust. It can be a trust, but it's always verified. 

Shawna: Okay. So, don't fire your lawyer and hire Chad GPT. 

Jeremy: No, you can help it guide you to have a more intelligent conversation and better understand what people are talking about. You can help it to generate you some predictive analytics, but you do want to go look at those and you do want to say, okay, are these numbers correct? How true is this advice? 

it can help you generate content like say you're a spa owner. You're a pool builder, you work in hearth and you have and you sell products or you sell services, you need help or or just want to take what you already have and improve it as far as like a product description. Yeah. Or a service description. You can help it say, "Help me write a description for such and such." And it can help you do it. Or you can take what you have and you can put it in. You can help me improve this and it and it can help you do it. And then the part where human intervention is needed is to go through and read that. You want to make sure it's accurate. Make sure it doesn't give you inaccurate information because that information is going to be critical to your product or your service.
 
Shawna: Totally. 

Jeremy: So you do want to verify everything it says. Say you have an email marketing campaign and you say, "Help me generate a fun email that's playful but yet does this and articulates this." You want to go through and you want to read that for the factual basis of it and also make sure that that is the tone you want, 

Shawna: right? Yeah. Your brand voice is really important. And I've noticed that personally. I mean, I run sales and marketing and that's my background. And I've noticed that it just is so apparent to me when companies plug something into chat GPT, get an output, and then post it right away. It's like it's not your brand voice. That is the voice of Chat GPT or AI. And it's very obvious. 

So, I think I mean that's really important that you make sure when you plug something in you're using those adjectives, right? Make it fun, make it real, make it personal, make it serious, make it professional, whatever brand voice you have identified, which starts with having a brand voice and having brand guidelines and having all those basics in place so that you can create your own guard rails around what content you create.
I think the other thing that I've heard a lot is I think you also need to be really cautious when you are grabbing information when you're grabbing outputs out of chat GPT you need to edit them in some way you need to revise the words you need to make it it's it's really kind of a rough draft in my mind it gives you the framework it does what I think is the hardest work which is just getting the framework work down on paper, out of your brain, down on paper. But you want it to remain your content, right? It's not someone else's content. It's your content. So, make sure that you're making appropriate edits. It's your brand voice and you really just are at this point, I think, using Chad GPT or AI for that initial framework versus like the final final draft like you said. 

Jeremy: Yeah, it's good for the broad strokes, but if everybody did a trust fall on it, every email or product description would come out with greetings or salutations.
or every product would look at this amazing product and everything would say that. So there are little euphemisms that we speak in and like you said the adjectives and different nouns. So you do want to double check it. It can be good for the framework, the broad strokes, but you do want to refine it for sure and fact check it. 

Shawna: I really like that. I'm going to use that. Don't trust fall on it. I think that just says everything. You know, you got to trust fall onto a person, and that person is one of your team members, somebody on the team who's double-checking things. They have the final say, and it's a person who's sending it out to market versus a piece of technology. And I think Evosus really aligns with that when we build software like Lou, right? A lot of features within our software will automate the process for you, but we're still going to ask that you double check it in the end, such as digital door hangers.

When you complete a job and you invoice it, we'll populate a digital door hanger for you, but we still want you to look at it before you send it out to the end customer because you might want to put your flare on it. Something might be incorrect that wasn't caught when you were out in the field or the technician wrote something that maybe is in your speech and not the consumer's speech. And so, you want to change the verbiage there. You know, we joke about that all the time, but it's real. 

 so there's a lot of things within LOU that are automated, but we still want that final checkpoint before it goes out to the end consumer. So I think we're really aligned with the advice that you would give. 

 What about any advice on how consumers or our clients could use AI internally? Any advice on that before we move on? 

Jeremy: Yeah. Always keep in mind the ethical concerns, the privacy concerns, sharing proprietary information that is very that you hold near and dear to your company.
You know, be careful of what you share because that language model can train on that and you do not want it to know proprietary information. You don't want it to accidentally give somebody else an answer that is very very near and dear to your company that is a trade secret. 

Shawna: Totally. No trade secrets in there. Are there any tools or options for people to have a kind of a closed loop AI option? Like are there chatgbt programs or plans that you can sign up for that keep things close to you and they don't train the model or are we training the model no matter what? 

Jeremy: No, there are options like when you do say use chat GPT you can go through the settings and you can say don't train based on my answers.

so we've started using some AI enhanced IDEs which are code editors and we very much have that option. It is a policy that we have internally as well as we control it through our IT department but we do not allow the models to be trained on our codebase. So that way it is not shared with the world whatsoever. 

you can do the same with when you use chat GPT. And often times whatever tooling you use that are out there available for the internet, such as there are some other builders, things that can give you templates, there most often is always an option that says do not train the model based on your answers. And you do I I would I would argue say that you do want that always checked.

Shawna: I would definitely agree with that, especially when you're putting things like proprietary information, pricing structures, compensation agreements, things with people's names on it. I mean, that's super confidential stuff and you wouldn't want that shared with anybody. 

Jeremy: Blueprints, plans, yeah, different language campaigns, product information, those kinds of things. You don't that is something that you hold and you want to essentially protect those secrets from being trained into a model to where say a competitor can go in and also use them too, right?

Shawna: Totally. Whereas things like maybe a job description or something generic isn't that big of a deal or you're just asking it a question maybe not a huge deal. 

Jeremy: Yeah. You're essentially googling it. Yeah. And so yeah, you do want to think about what you ask for and make sure like if it does not have that guard rail in place, be very very cautious of what you ask for. 

Shawna: Yeah. Yeah. I think that's really good advice. 

Jeremy: Something that you touched on just to clarify too, AI to a myth is AI is conscious. It is not conscious. While it is in place, it does help us make decisions based on data. It is not sentient. Not even close. So, that is a myth. A lot of people think it can think, it can do things.
 
It cannot. It is not sentient. So in quick like just a quick rundown in non-technical terms of what AI is. AI is the ability for machines to mimic human intelligence. Tasks like learning, reasoning, and decision- making. 

But it will not go forth and it will not give you therapy 

Shawna: It doesn't have a conscience. 

Jeremy: Right. 

Shawna: Okay. Got it. Yeah. I think that's a really important distinction to make. And I think that that leans even further on what you were saying about don't trust fall on it. This could be inaccurate and it doesn't know the difference between ultimate truth and false, right? because it's just it's kind of regurgitating the way it's been trained. 

Jeremy: It does not have a moral compass. Yeah. It is as good as it's trained, right?

Shawna : Yeah. So, a tool, not an end all be all? On that note, I'd love to just talk about how I think people are wondering how we're thinking about AI. We're kind of the legends in the software industry for Pool, Hot Tub, and Hearth. And we've been consistently really the thought leader. 

So, can you just share with everybody on the line kind of how we are thinking about AI? Are we just slamming it into everything? Are we not doing anything? What's our mindset when it comes to developing this product? 

Jeremy: So, we're very strategic about it. We're not just slamming it into anything like that. We think about it in two parts. One is internal tooling and one is external tooling. External tooling will roll out a little bit slower than internal tooling because we really want to create a really great user experience and we also have competing priorities as well. From just the industry's need. Yeah. And so we pivot based on what we feel like people need and what will benefit them most.

Internally, our strategy is to use it to improve quality. It is very much that we are using it in the way we code to help find errors to help find bugs to help keep best practices and standards. And we use it in our QA process and we're mixing it more into our QA process to help us create predictions of possible errors and think about use cases that a human may not because we try and think of everything. We really do. But the fact is is we're human and we can only ingest and remember so much even as we record it. And a machine can break things instantaneously at a much faster rate, right? And so the goal of QA is to break it. And so we use it to help us find use cases to break it.
 
So that way we can do that. From the coding side, we help it to use coding suggestions, standards, and help accelerate that. And we are rolling it out very responsibly. So we're not just full force at a million miles an hour, but we are rolling with it at the speed of the industry and working on being pioneers in it as well.

From the end user point of view, we are thinking about ways we can use that into things like generative context, generative text, generative emails, generative images, and those kinds of things. And we're a little bit cautious but we're responsibly doing it. We do have some things that are going to be used in different facets but we are very much intuitively doing this as responsible as possible. Because we want it to be a good experience and there's been companies out there that have done major major trust falls on it and it's come back to bite them.

They're these multi-billion dollar tech companies. They're not in our industry. But we take and we learn those lessons because they are throwing billions at it, right? And we're seeing how they, you know, misstep and we're learning from them. And we're building on the mistakes that they have made. 

Shawna: I think that's really, I think that's really smart. I mean, that's always been my approach in my career. I want to learn from other people's mistakes so that I can take the learnings and move faster, right? Because if we're the first ones and we're tripping over ourselves trying to figure this out, it's kind of silly when people, like you said, are throwing billions of dollars at it. They're making the mistakes. Let's learn from their mistakes, take the learnings, and then just do it right the first time versus wasting a lot of time making mistakes and trying to figure it out.

 so I think that's a really smart approach. I'm really excited that we're using AI to build a better product. Hopefully it will cut down on the time that QA takes to test things. Hopefully we'll be able to scale things really quickly and test at scale like huge amounts of data that a human would never be able to produce. So I think that's really exciting. 

and like Jeremy had mentioned, there is a new product coming out in the next few months, lip sealed. I'm not going to tell you guys what it is, but over the next few months, if you keep up, we've also dropped some hints in the past, so if you've been paying attention, you might know what it is. But that will have AI really in the interlacing of the product and it will be at the forefront. So that's really exciting. And then as time goes on, like Jeremy was saying, we will start to weave AI into LOU into the framework.
you're going to see it in the service app in you know the next you know in the future so it's it will be in the product but we are being really thoughtful about it we're not throwing it into everything we're not just saying new feature slam AI into it because we want to make sure that it's done correctly and LOU is used it's a full business software right you are managing your financials you are managing your taxes that you pay you're managing managing your customers, their invoices, all of your inventory, your vendor purchasing.

 I mean, we could go on and on about what you're managing, but we can't afford to make a mistake. You guys can't afford for us to make a mistake and put it in the wrong place and use it the wrong way. So, I feel really good about the approach that we're taking to it. And like we talked about, there's a lot of automation already in LOU. Reorder alerts, digital door hangers, mass invoice and pay. There's tons of automation sprinkled throughout Lou. There's just checkpoints at the end because we want to make sure that you're really happy with how it looks and how your business is coming across before it ends up in the end consumer's hands. 

So that's really important. I'd love to pause for a minute and open it up for questions. I think we have a couple of minutes left. We could probably talk about this for hours. We do it at an executive level. But Shannon, can we pass it over to you to see if there's any questions out there?

Shannon Stine: Can AI help me with images or graphics for social media?

Jeremy: Absolutely. It can absolutely help you with that. That is one of the top use cases currently for it. And you can take it and generate you an initial image. It can enhance images as well as those images that it generates. You can continue to iterate on them as well. 

Shawna: Yeah, totally. And you can even go as far as put your logo in, put your brand voice, right? Put a short description of what you want your brand voice to be. Do you want to be more fun, professional, real, cutting edge, like all of those words, right?

Think about really creative adjectives that describe your business and plug those in with it and it will pop out a pretty solid image most of the time. You might have to iterate a couple times, but the quality that I've seen is good.

Shannon Stine: Another question is which AI bot do you recommend for email drafting?

Jeremy: Yeah, that's a good one. I would recommend ChatGPT 4.1 or 4.0. Those ones have the most trained models for content generations. That that one would probably be your best bet there. Because that one is the largest language model with very common things and uncommon as well. There are other ones that are very industry or tech specific, but that one's going to be the best one for content generation. 

Shawna: That makes a lot of sense. And from a non-technical person, if you have a CRM, large CRM like HubSpot or Salesforce, they have solid AI options within those as well.

I don't know what the language model is behind the scenes that's feeding that information, but I do know that Salesforce and HubSpot have great AI options just built right into them. So it's not an additional investment. 

Jeremy: They most likely are Chat GPT4. 

Shawna: Probably probably

Shannon Stine: The final question is how would I train my employees on AI? Are there some courses that I should take first?

Jeremy: There definitely are courses on it. There's companies like Udemy and Pluralsight that have published courses on it. As well as how to do it. And then I would create a set of rules around it to use it. Like always verify, always read it for anything that is you know inappropriate or biased or inaccurate and then use it responsibly. But there are definitely courses out there with those two aforementioned services that will be able to help you on how to use it.
But it's generally very much the good thing is it's common sense and it's creating a set of rules and standards around it. You can also search for the open AI initiative and you can look at the set of standards research initiatives that they have around it too and you can help guide your guard rails around that and their mission statement because open AI being a nonprofit their goal is to really and they're the ones who are responsible for chat GPT their goal is to push forward the boundaries and the initiative of AI responsibly and so they have a very very solid set of rules and guidelines that they go by. So I would very much model them around them. And then you can use the other two services too. But between those three, you should be able to gather a very solid set of idealisms and standards and rules around using it responsibly. 

Shawna: Yeah, I think that's great. That's great advice.

I just wanted to add that I went to a conference recently a couple months ago in San Francisco. It was actually a SaaS and AI conference. And OpenAI, you can imagine the executive team spoke on nearly every stage every day about different things. 

OpenAI also has some really incredible training options. If you just Google them, they'll show up.

But I want to leave you all with kind of one final thought. What we believe here at Evosus is that the future, the next couple of years, really is all about AI orchestration. And what that means is that your team needs to get really good at telling AI what to do, letting it do the work, and then they come in at the back end and confirm, validate, and push forward to market. So it's really like your employees doing the first 20%, AI doing 70% and then they're taking on the last 10% of the work. So, if you were going to spend any time training your staff or creating an initiative, I would have it be AI orchestration, write down a bunch of prompts.

What are some prompts that you would recommend your team to use? Create a flyer for our sale this weekend about blah blah blah blah. Here's the logo. Here's our brand voice. And then teach them how to iterate on that, right? set expectations as far as what you want to see, what maybe an image of what you guys have produced in house prior and make sure that the person double-checking AI's work is living up to those standards at least, right? It needs to be at least this good, if not better, before I would feel comfortable with it being posted on social media or sent off to print. So I think it's about setting really clear expectations, having business processes that hold people accountable. This is not the wild wild west. And then and then helping them come up with prompts, right? Have them come to the table in a meeting once a week with an issue they ran into and specifically what they asked AI.
When I was at that conference, everybody had prompts that they used within their business. AI has I think a list of them that you can Google or just suggest prompts that you can use to get started. So there are a lot of resources out there. I will try to find them. If I can find them, we'll include them in the recording email tomorrow when it gets sent out so you guys can have a starting place. But there is a lot of information out there as far as training your training your people. 

Jeremy: It's very much just like everything practice practice practice practice. The more you use it, the more comfortable you get, the more you learn about those responses. With that giving knowing that it's not ever going to be perfect. Just like nothing is perfect because perfect is the enemy of great. So it'll give you great but never perfect. 

Shawna: Yeah, that's okay. Shannon, any other questions out there?

Shannon Stine: No, it looks like you got them all.

Shawna: Beautiful. Okay. Well, I just want to say thank you all for joining us. I know if you're in the pool industry, you are really busy right now and it's not light on us that you would spend the time to hang out with us this morning, chatting with us. We really appreciate it.

The recording of this webinar will be out to you all in your inbox tomorrow. So stay tuned for that and we will see you on the next one.

Thanks again Jeremy for joining. That was really valuable.