#12 - Agentic Systems in Tourism: How AI ‘Digital Employees’ Crush Labour Shortages
[Music] Welcome to the Tourism AI podcast where we look at all things at the intersection of artificial intelligence and tourism. Excuse me. I'm here today with uh Jesse Angland uh from Rapid Innovation. And uh Jesse reached out to me a little while ago here uh to talk about some of the work he's doing uh specifically around Agentic Workflows and Agentic Systems. and uh I thought it'd be a good opportunity to bring him on to talk a little bit about what he's doing uh bring it back to context of uh where the tourism industry is at today and help try to bridge the gap of understanding I think between sort of where people are in tourism today and uh where companies like uh Jesse's company are sort of trying to take the future um and coincidentally so tourism networks also going down the path of the agentic uh workforce. So, with that, uh, Jesse, maybe I'll give you a minute just to introduce yourself more properly. Uh, tell us who you are and, uh, tell us a little bit about your company. Sweet.
Thanks, Peter. I appreciate you having me on. Um, yeah. So, I'm I'm Jesse. I run Rapid Innovation, also another company that's right on this. It's a a daughter company of Rapid Innovation called Rue, which is a a u a really a digital employee creation platform. So, we probably won't get into that too much in in this podcast, but but it's it's all about agentic labor. Ultimately,
Rapid Innovation is is a very very simple company. At its core, Rapid Innovation builds systems for people that replace humans with digital labor systems. So, we we take human labor systems and and we replace them with digital labor systems. And so a lot of
what we're dealing with is building very very complex agentic systems because it's handling real people's jobs. They need to be able to make phone calls and return emails and and interact with clients and customers across different modalities and things along those lines. And then we also do a lot of lot of other interesting things. But this will be a fun podcast. I'm I'm really curious to learn a little bit more about the tourism industry. And I I actually honestly don't know how it's going to be affected by AI. It'll be
interesting to to learn more and then and and start to make some of those connections as we talk about it. Yeah. Yeah. Well, I actually jump in because the the words you use there are very interesting from a tourism perspective because you use the words replace humans. Uh and in fact the biggest challenge in tourism is that there aren't enough humans. And so I think
that uh the opportunity in tourism actually with aic systems is to augment humans because right now humans are not uh are not struggling to even keep up with the workload. Um, a lot of tourism organizations are not running at full capacity. They're not able to welcome guests because there's just not enough people, especially in rural remote areas and places where, you know, the, you know, the population's declining, things like that. But there's a huge demand.
You know, you look at places like remote wilderness lodges where people want to go out and unplug and unwind and be away from society and all this stuff. You know, places like that, it's hard to find the skill sets and and and the right humans to do the work. yet there's going demand for you know those tourism more experiences right so I think there's a huge opportunity that's where my mind is really at and and these rural remote places are interesting because to get to them isn't necessarily an inexpensive tourism sort of visit so you know I have I'll pick on some of my clients here I have clients that are heling resorts for example up in you know just south of Alaska the Alaska border and they're very remote and they're amazing places and it's not cheap to go heli skiing while you know you got this kind of interesting contrast between these rural remote places that don't have the sort of the human resources to do the work but you also have this very high-end clientele that have very high expectations around personalization around everything going well around digital experiences right so there's this interesting contrast happening and so that's where I really think that companies like yours and some of the what you're working on actually provide a huge solution for people so yeah sort of in that context I'm wondering if you want to maybe maybe just jump in with some initial thoughts I'd be curious So like like my my first question would be like what is the work of that? Like I I haven't sat down long enough to think through like what the work of of that kind of tourism would be like we we'll take your example of like the helysing thing like so let's say let's say that you've got a bunch of people with with you know disposable income and they want to go you know rent a helicopter and go skiing. Obviously,
that's a pretty like, you know, I I would guess like top 5% kind of experience. Like most people are not going to have an experience like that. The ones that do are going to be paying a pretty penny to do it. Is the work of that just like finding those people, catering to them, scheduling it, like or is it when they get there it's like it's taking care of them or what is the work of doing something like that? Yeah. So, I mean
really it's it's all of the above, Jesse. So certainly finding the people uh it's very competitive even though even though it's not a you know huge number of people there's not also a limited number of heli skiing resorts. So the digital marketing component is very competitive. The sales follow-up is very competitive. So somebody you know
maybe fills out a form but staying on them. Every group has pretty unique personalized needs. So you know my group's coming and we want to do XY Z. Two days later you got another group coming. They want to do this and that. So it's a it's not only a heli skiing operation, but they also operate two lodges with full kitchen staff with full service staff, hotel staff, plus then you got helicopter maintenance. So it's actually a very complex operation with many many moving parts that uh AI can actually help with and many components of it, right? And I imagine that it's one of those things where there's a very few customers and those customers bring in all of the revenue for the entire operation. So it needs to go very well
every time. That's right. Yeah. Exactly. And it's a little bit more concierge. Interesting. Yeah. So, I mean, here's here's one of the things that I always there are there are probably three different levels that most business owners ought to be thinking about AI in. And we're not even talking about agentic systems, just AI in general. And maybe even make it
simpler, large language models, right? Most people they go sign up for or don't even sign up like they go log in to chat GPT and they use it to write you know a poem for their for their significant other or they they ask it to tell them a joke or or whatever it is. Like that is that is quite literally most people's experience with chat GBT. It does a relatively poor job of of writing a poem for their wife or telling them a joke that is a little bit too generic. And then often times what they'll do is they'll look at that technology and they'll say, "Huh, well that's interesting." And then they'll never go back to it, right? And so one of the things that most people ought to do if they're at that stage where they're like, you know, I just haven't really messed with any of these models is they really ought to just go start using them, seeing like pushing the boundaries of what it can do. You know, I I probably I think I pay for every single every single subscription to every single large language model that exists today. currently that you can pay
for. Uh because like one of the things that like as you start moving down that route of like hey I need some help doing some research right I no longer just use for instance openai deep research to do my research. Generally speaking what I'll do is I'll go to OpenAI. I'll have OpenAI 3 model, which is one of the models, build me a research brief that I then bring over to Gemini 2.5 because it does a better job of doing research. I
toss it the research brief and say, "Hey, can you research this?" Once it's done being researched, I'll port that data back over to, you know, another model to actually do something with that research. And so what you'll start to learn as you start to mess with these things is like you should never use um you know a logic and reasoning model like 03 to write copy. You know, if you want to if you want to put together an email an email series that you're going to send out to prospects, like you should probably go use Claude Sonet 3.7 or even 3.5 because it's a better writer than than OpenAI's, you know, 03 model.
And if you have to if all you have is an OpenAI account, like don't use the 03 model to write your copy. Like use that to come up with your strategy and then feed that strategy to the 04 model because the 04 or the 40 model because the 40 model does better. So, like starting to learn some of these things, it can be a little bit overwhelming for some people, but it literally just starts with like starting like you just start doing it. Like I I mean, commit to using AI to write drafts for all of your emails. So, you don't ever have to think about writing a draft for an email. You just explain what's going on. You toss
the email in and you let the AI draft it. And then you take that draft and you go back and you edit it for your changes. and it saves you like so much time. Um, that's how I started using it, you know, years and years ago kind of. And then from there, you start to expand on things um to like actually building systems that do real valuable work for for you. But
it has to start with something like to be quite honing my business partner this probably two years ago. Um, I I had just really started becoming a power user of AI on a daily basis. I realized that if um if OpenAI, which was really the the one model back then that was really really good, you know, two and a half years ago, um I realized that if OpenAI servers went down for a day, it would be more productive for me to just take the day off, go spend time with my family, and come back the next day and catch up on my work because the amount of time that it saves saves saved me on a daily basis just giving it work to do having it help me think through things like processing through certain tasks you know like hey I need to create a marketing campaign for this year and I need to put the marketing calendar together and I need to do all this stuff like just drafting all of that documentation for me to go and take a look at and read and put some like human cognitive thinking into it was saving me it just wouldn't be worth going to work without it and uh and that years ago, you know, that was the case. Today, it's so much more so. Like, I don't I don't understand how how anyone functions in a world where they're not leveraging AI like a lot because I would feel so unproductive if I did things the old analog way. I just I just can't imagine
it. Yeah. Yeah. Well, and it's interesting what you noticed. Yeah. Well, and it's funny that what you're saying there are interesting because at the same time also I'm noticing when I talk to a lot of people is that, you know, they say, "Oh, yeah. I'm using AI. I use it every day. And then I ask,
well, are you using a paid? Like, which model are you using? I just use the free chat GBT account. And you know, you just said something there where it's like there's this model, this model. So even within something like as simple as a chat GPT, the paid versus the free version, it's like you're not even using the same tool, like you're not even using the same tool at all. Right. So even Right. And so I think that I think kind of this conversation, Jesse, where we started before we hit uh before we record here is just where people's understanding is. And you know, as I
said to you, it's like, well, I use AI, therefore I know what I'm doing. And it's like, well, use AI as a very broad term with a lot of a lot of breath to it, right? And so you just describe you just actually the it's funny because I actually create uh I use 03 to also create my uh research briefs and then I put I go into Gemini as well. So I'm glad we're aligned on that one. But but to your point, at least using the free model is a good starting point, but get a paid model and start to play around with the different features and functionality to understand what it could do for you, right? So, I I didn't pay I didn't want to pay for a long time because I'm cheap and I like to save money like most people. And when I say I
didn't want to pay, I didn't want to pay for the $200 version. I was paying for the $20 version and I had like the $20 subscription to Anthropic. the $20. Like I was paying for all these $20 subscriptions and then all of a sudden OpenAI is like, "Hey, do you want to pay 200 bucks?" And I'm like, "No, I don't want to pay 200 bucks. This is crazy. I'm already paying probably $100 another subscription." And a friend of mine
said, "Hey, Jesse, if you hired an assistant that did as much work as just the OpenAI models, how much would you have to pay them per month just to draft you emails and say and write out calendars for your social media, you know, posts and to do research on things? How much would you pay them on a monthly basis?" And I didn't want to answer the question because I knew that it was going to be a lot more than $200 a month. and he said, "You're you are a fool for not for not paying $200 a month for what is significantly higher quality work than what you could hire for $200 a month." Um, because that is the best ROI you're you will you're ever going to get. And right now you can take advantage of the fact that most people aren't going to aren't going to do it because they don't want to pay for a subscription because it's weird because they're not viewing it as the synthetic intelligence that it is. It really is like having an assistant. And you can't hire an assistant for 200 bucks a month.
You just can't. Not one that is that capable. Yeah. And never mind one that'll just respond anytime you need it and give you right be there for you any moment you need it. Yeah.
Yeah, if I wake up in the middle of the night, I'll even just like pop over and turn on the like the chat feature so I can just talk to it and I will say, "Hey, I don't want to, you know, I just need you to remember this." Because now the new memory feature inside chat GPT like the open AI's models, it's awesome. You can start up a new chat, you go and talk about whatever you need to. Then the next morning, you can say, "Hey, what what did I last like what I talked to you about last?" and it'll just go and recap everything that you talked about in the middle of the night. And so it's like my version of have keeping a pen and a notepad by my bed in case I've got some crazy idea. I just use JPT to do it. But that's where it starts, right?
Like because that like it depending on who you are. Um that might sound like a lot to like understand the different models and and how all these things work and it is a lot. I mean it's a lot of work. the the benefit of doing all that work and trying to understand how all of these things fit together is that pretty soon what you're going to want to do is you're g you're going to say something like gosh you know what' be nice what would be nice is I know that I go and I write my research briefs in OpenAI and then I bring them over to Gemini and I stick them there and then I take that that research back and I turn it into um I turn it into a social media campaign that I then go and post to to engage people that are interested in whatever it is I do. And what would be nice is to just kind of hook all that stuff together, right? And and so that I don't have to go and copy and paste and copy and paste because you start to get lazy because you realize like gosh, like I'm now I'm becoming this like your whole job becomes just copy and pasting from one place to another. And so then you discover like um workflows, right, where you get like naden or make.com or there's a bunch of these other workflows and you're like, oh man, like I can I can make a computer do the copy and pasting. This is great.
Um and then you start to discover like gosh, you know what I can what I can actually do is I can make AI agents understand the tools I'm using. And when I say tools, tools includes LLMs. Mhm. And I can have them reason through the process of moving this information from one place to another to another in order to get the results that I'm looking for. And I can even have them
reason through what it looks like to push that information to the social media platforms that I want to push them to. And then when people respond, I can have them look at those responses and reason through what they should say in response. And that's when you start getting agentic systems, right? is because now you're no longer involved in the copy and paste, copy and paste, put it into Twitter, wait for people to respond, talk to those people, wait for that to turn into someone that you ought to call and then call them. Like now what you're doing is you're just saying, I have an idea for a campaign. You put that idea into one side of this agentic system and then that agent reasons through, okay, I need to go and do all these things and take these steps. I
need to post all this stuff on Twitter. And as soon as somebody comes back that looks like a really good lead, I need to send Jesse an email. And so it plays that game 24 hours a day, seven days a week. And now you have like a digital employee or an AI agent, right? Something that's autonomously, you know, acting on its own for the purpose of trying to accomplish something that you set it out to accomplish. But it starts ultimately with that OpenAI account and the anthropic account and and moving things around. Yeah. Yeah. I always tell
people that it's sort of you have to start off with a bit of curiosity and testing and playing and then that sort of turns that creates aha moments and aha moments create what if moments, right? To exactly what you're describing there. Oh wow, what if I could do that? Wait a minute. Holy cow, I could do this. Right? And that starts to create more curiosity and more interest. And
that's how you sort of start the the journey, right? I mean that's where it all starts. So you're 100% bang on and that's literally what I tell people all day every day. So uh totally aligned with you there. Uh so let's just talk about these agentic systems a little bit. I'm curious uh obviously these are things you're working on. You mentioned
you had a workflow even for yourself that you've been using. It's been really successful social media. But are there sort of currently you know the five to 10 agentic systems that everybody's like asking for? Like one of the things I'll I'll just caveat this with. I was in a
call last week with a client who was like you know can we do something to help clean up our email and make it work better. That's like the one that I hear probably most often. So, are there those sort of common ones that you hear often that people are asking about? Um, I hear the email one a lot. Um, that is something that Google ought to just pay attention to because they have some of the world's best AI large language models. They should just go clean up Gmail for people because it would make a lot of people really happy. It would, but usually it's around marketing. I
mean, one of the things that I like people pay a lot of money to get marketing done. And what's interesting is that everyone needs it even if they don't have a lot of money. and it's very very expensive. And so if you if you're running like a five person company and you or even a 10 person company like it can just be too expensive to go get good marketing done. And so then people pay for bad marketing and it doesn't work right. And so one of the things that one
of the things that that we have yeah that we spend a lot of time on is just marketing systems like you know can I mean agentic systems can market almost as well if not I mean in in our opinion even in in some ways better than like like human teams have like we were talking before we hit record on this but I've got we've got an Instagram channel that was a complete experiment like can you create an AI I agentic system that can research and then and then create and post, you know, Instagram reels that people would be interested in watching to build an audience. Um, and I I I didn't think so, but it was worth giving it a shot. And so, we started doing it. And the at
first it, you know, it was rough. Uh, the videos weren't very good, but eventually the system really learned how to do a good job. we figured out how where to um where to start adding humans in the loop, right? In in that system, probably 3% of the actual work is done by people and 97% of the work is done by a AI agents. And we're up to like 3 million views a month on on our Instagram channel as an just an experiment. There's a lot of people that could do really really neat things with 3 million views a month, you know, whether that be their tourism business or anything else. and to build that system was just not was it just doesn't have to be as crazy as people think. Um
I would say that's the majority of where most people are interested in in like trying to do something. email campaigns, content creation, SEO management, ads management, like it's it's all the things that some digital marketing agency is going to charge you $40,000 a month for the next year without performing very well, you know, in order to help you with all these things that feel like magic. Like AI systems can do the same thing. They're just going to do
it for 99% less money, basically. Yeah. So how so let's talk a little bit about not the how per se. I don't want to get too deep into the weeds on the how, but sort of like kind of tell a little bit tell me a little bit about when you build these systems, uh, you know, what are you are you using a number of different tools and putting them together? You mentioned kind of multiple large language models. Uh, are you using
MCP? Are you like these are all terms that probably are foreign to a lot of people. So, let's just sort of break down a little bit the kind of what are the component parts of it all that sort of make it all work or in your opinion kind of are there buckets of things that need to happen in order for agentic systems to work well? Yeah. Um I always tell people all valuable work is done with four things. It's done with an intelligence of some sort that can be human or or like a large language model is a kind of synthetic intelligence. So
it's done with an an intelligence. Um it's done with tools like if you were going to like you and I were using a podcasting tool right now in order to get you know valuable work done we'll call it right because is marketing work. Whether or not it's valuable or not I guess would be debatable depending on how you know how successful it is. But all all work, right? You use tools to do stuff. Your finance guys going to use QuickBooks. Your marketing guys are going to use whatever. Your developers
are going to use development tools. Um the third thing would be process. Uh and so like all anyone who does anything is going that is successful is going to have a process. Whether that be
like in the morning I wake up, I open my email. First things first, I clear out all the spam. After that I figure out what's important. I go and mark those as important. and I put them in a folder,
you know, like whatever their process is, right? Maybe it's just organized chaos and their process is like I spend an hour and just start from the top and work my way through to yesterday. Yeah. And then I do that three times a day. It doesn't matter. Everyone's got a
process. So, um, and then the last thing would be memory. Like you you have to have some form of context or memory as to what you're doing. If you woke up every single day and you had no idea what you were doing, like what what your business was and what why you were checking your email or anything, that wouldn't work. And all valuable work is is I would say made up of those four things. Well, the only part of that is unique that is unique to humanity would be the intelligence part of it. And so
when you can look at processes, memory, and tools as something that you can digitally build, then the only question is, can I replace the intelligence that's running this with a large language model, which is really easy to test. You basically just go have a conversation with a large language model and see if it's smart enough to do the job that you're trying to replace, right? If it is, then you go, okay, yes, now it's worth building the tools and all that stuff. So those those four components are really what make up all agentic systems. The level of intelligence that you need is and the kind of intelligence that you need is where you're going to start picking what models are going to be a part of your agentic system. Oftentimes you need more than one kind of intelligence. For instance, and this would be digitally or analog. It wouldn't matter. If you take
a bunch of really really creative people and you put them all on a task, if the task is highly creative, they'll be very successful. But if the task is very logic driven, they're going to fail, right? Well, if a task has both logic and creativity, then you're going to need a couple of different people on the team to act as the intelligence that drive that t that task forward. Well, agentic systems are no different. And so, if you need a lot of logic and reasoning and you play with these mo with these models a lot, you'll start to learn that okay, like the, you know, uh, OpenAI's 03 model has a lot of logic and reasoning. Claude 3.7 has a lot of logic and reasoning. Gemini 2.5 is a lot of logic and reasoning. If I want the
creativity side of it and I want to go write content or or think through things in a creative way, then I'm going to use different models than those. Now, some of them are better at doing both and and you kind of figure all of those things out. Then once you have your intelligence, then you need to be able to talk to tools. So this is where model context protocol or MCP servers. When Enthropic came up with this, when Enthropic came up with the idea of actually having large language models work with APIs in a standardized way, that just opened up the um opened up the world to be able to like do things. Model context protocol is a protocol exactly like HTTP, right? Which would be the protocol that's going to send information back and forth on the internet. It allows everyone to know
that if I format information like this on my server and then Peter goes there, he's going to be able to get the information because he's got the other end of that protocol and he knows how it works, right? And so MCP is just literally just a standard that allows the intelligence or the large language models to come in and and do certain things with an API like they can call an API. And so instead of having to train the large language model on every single API that a tool might have, so let's say that you wanted to build a finance agent. The finance agent was supposed to reconcile your bank account with your your QuickBooks account and just make sure all the transactions were there.
Well, before model contact protocol, you'd have to go look at all the APIs that you need for all of that tool calling. Like it needs to get access to the bank account somehow. This is how it's going to have to do it. It's got to get access to QuickBooks. This is how
it's got to do it. and access to my email so it can like check everything. Well, with model context protocol, basically you configure all of the APIs inside of a specific domain. And so maybe it's all of the APIs across all three of those platforms. Maybe you maybe it's one MCP server for your QuickBooks account, one for your email, one for your bank account. Well, once you've
configured that, the model already knows how to use MCP. And so it can call any of the tools anytime that it wants to inside of there based on its its logic. So it can say, I need to get all the transactions for last month. Um, then it
looks at all the tools it has access to for QuickBooks and says, "Oh, if I hit this API, I'll get this information. If I hit this API, I'll get this information. That's everything I need in order to complete my task." It hits those two APIs. It gets the information
back and then it continues to move forward. So that'd be the tool layer of of what it's doing. There's another aspect of that which would be A2A which is another protocol that Google came up with where an agent can interact with another agent in order to get its work done. And so for instance, let's say you had let's say you had um a series of agents with access to tools. And so you had one agent that was like your accounting agent and had access to all of your finance. You had another agent
that was your HR agent. Well, you could have another agent that sat above those two agents, and it could make a call to the other agents, kind of like you'd send someone a message on Slack or an email, and it could say, "Hey, both of you agents, it's time for us to go and do raises this year, and I'm wondering based on the health of the finances whether or not we can give everyone a raise of 15%." Well, that agent on the top that you just gave that those instructions to can then call the agents underneath and those agents can get then go in with access to all the tools that they have, pull in the context, reason together about whether or not you can give people raises, pass that information off to the agent that made the initial call, which then passes it off to the human at the top, and then I know based on all of the data that it pulled whether or not I can give a raise to everybody at 15%. And so when you start combining all these things together, you can just build wildly intelligent systems that can do like really really valuable work.
Yeah. But it and it's just a bunch of layering of intelligence tools, memory, process. Intelligence tools, memory, process in different places and then connecting it all together. So yeah. So I mean I think again you use slightly different language for me but not really that different. I I sort of kind of same same kind of stuff, right? I I talk about memory. I kind of talk about
context process um and yeah sort of understanding what out what good looks like or outcomes look like sort of you know that type of thing and yeah you can sort of build basically with that foundation you could just and then the tools obviously and the intelligence are sort of the the make it happen layer right and yeah so with with that sort of those concepts at least from a business perspective that's because I'm always trying to help people understand okay in language that you understand in your current job it's like okay you got context you got you know the work you need to do you got a bunch of processes that you need to follow to turn your context into the final deliverable or whatever whatever it is you want as an outcome and you have some idea of what that sort of good good looks like right what you're trying to achieve right and sort of what you need it to be you know and so and then the the AI tools can then just sort of basically sit on top of that if you're organized well and can help you kind of turn context and into an end product right so similar kind of language I guess for me what I'd be curious to understand from you in terms of building these systems is if somebody's fairly, you know, new to all of this and wanting to come in and wanting to sort of start thinking really what you're talking about is systems thinking to some extent in terms of sort of systems thinking and sort of thinking about how to create now bringing these tools and this intelligence into their systems. What do organizations need to do? What do businesses need to do to kind of get prepared for this? Like if somebody wanted to come work with you today, um, you know, what questions would you ask? What would you sort of get them to start organizing or preparing for in order to be able to actually succeed? I mean, the hardest part about working with a company that's going to help you build these systems is that generally speaking, like unless you do exactly what I do, which the odds are you don't, it's going like we have to somehow get me to a place where I understand what you do well enough to be able to build a system that does it. And so the pe the clients that I love working with the most are the ones that show up and they actually will say like, "Hey, here's all the tools that we use. like this person's going to use this CRM. They're going to this is the data they're going to pull from it. They're
going to use, you know, Excel. This is what they're going to do with it. They're going to use Google Drive. This
is what they're going to do with it. And they're going to follow these processes and they have all their processes laid out and they're going to need to know these things and they have all the data that they need to know. Like if someone shows up with that, like for instance, I had a client show up with all that data. This was two two years ago. I assigned him one developer. We built him an agentic system in three weeks that replaced about $3.5 million per year
worth of payroll. Now, it was sad that all those people lost their jobs, but I don't know very many business owners that wouldn't be at least a little bit happy that they had spent $18,000 on development costs to replace, you know, $3.5 million worth of payroll. Like, that's a pretty good ROI. And my guess
is the board was happier that year, etc., etc. And so like if you are really organized, I can just plug in what you have into like we can just do it. Like okay, sweet. Here's the process. Let's
build it. Here's the tools. Let's let's get the MCP server set up. Let's just do it. The hardest part and the most time consuming part is trying to pull that out of someone's head when they have no idea what the process is because they they've just done it subconsciously. Like they just don't think about it. It's not written down anywhere. They
don't think like, man, what is the process that we use for sending out an invoice to a client? It's like there is one and it's that dumb game you play with your kids, right? Where it's like you have your kids write down instructions for making a peanut butter and jelly sandwich and they just don't really think through what the process is. And so they'll say, "Take the peanut butter, put it on the bread." So you grab a whole thing of peanut butter and you stick it on the bread and they're like, "No, you have to open it." It's
like, "You never told me." Um, and I think most people run a business like that. Like they don't really know what their processes are. Now, large language models can help you with some of that to some extent because they're smart enough to understand. You're probably going to have to open up a peanut butter before you put it on.
Um, but only to some extent. And you have less and less problems the more detail you can give it, give a large language model or the intelligence as to what the job looks like. And it's no different than an employee. If you hire an admin and you say, "Go figure it out." Like, she's going to suck at her job for five months, six months before she gets good at it, right? But if you hire an admin and say, "Here's a binder of every single thing you will ever encounter in this job." She's going to
be a rock star in two weeks. Yeah. Because you will have equipped her with everything she needs. Same thing with these systems that we're building. And so that's really what people ought to and they should do it whether they come talk to me or they're figuring out this out for themselves. Like they should go
do that work because it is the work that needs to be done in the future. regardless of what direction you go, you can take that. Never talk to me and go use a platform like um like N for instance, and you could start building a lot of these uh these workflows yourself. It's not hard to do. Um, it takes maybe it is I don't know maybe it is harder than I think it is to do but it's not to me it it like six hours of watching YouTube videos on a on a Sunday afternoon and you'll be able to start building simple workflows that that organize and clean your email and you I would say that for most people specifically in tourism it is hard to do and it's not even something they're interested in doing. Sadly you like
literally every everything you're saying is things I say to people all the time, right? It's like regardless of even if you're ever going to go down the agentic route, even if you're not going to use AI, like building processes and standard operating procedures and having naming conventions in your files and having clean files, like these are all just good business practices. And it's funny because these are all good business practices that like coaches and people have been preaching about for like for forever, way before AI was ever on the on the on the radar of people, right? And now we're in this funny place of like, you know, like my company I I happen to my wife is extremely hyperorganized. We have an SOP for everything. Everything's super organized and so like life is good. But I keep
telling people it's like ah you got to do the unsexy work of like organizing your business and cleaning things up and before you can actually put these things to work. But at least the good thing about Genai is you can sort of still get a lot of micro wins because it you know you can you don't you know like one thing I get people to do when I do training is like okay let's create a organizational essence essence etsense essence context brief. So, you know, the easiest part is let's go get AI to read your website and tell us about what your business is all about. And that's like
usually where I start. I get a, you know, get like 03 to go read your website and do it and you know, I have a bit of a prompt for it and I and either it's either a oh that's excellent and yes, my website actually reflects who my business is or b that's not what we do. That's in which case it's like, oh man, we have even more work to do, right? Uh but to your point is like the systems are pretty smart and they can figure things out pretty well now and they can help you through this process and speed up the process. Uh but it's still kind of unsexy work. So like it's sort of we
have to and I keep telling people like AI is going to save us time and work but it's going to cost us time and work to get there, right? You sort of you can't you you can't just magically all of a sudden end up on the other side and have you know a team of agents. Similar to hiring a person though it's exactly the same to your point. don't have good training, don't have good systems, don't have good files, don't have good SOPs. Imagine if you weren't organized, you went and hired 10 people. Like, it would be chaos. It would be madness. It would
be horrible. There's no difference. If you went and built a a really really complex agentic system and then you you fed it in, it would just create chaos. It would be like I um I built an agent. I wasn't paying I wasn't really uh this is actually even slightly embarrassing now that I'm talking about it, but I just I just didn't care that much about making it perfect, right? But I wanted to build a little agent that would go and organize all the files on my computer because I thought I'm not your wife, right? I'm not that person who's hyper hyperorganized. So like my download folder, it gets crazy. My documents folder, I'll like call things all kinds of stuff. So all this agent
really did was it went in to my file system. It looked at the files. it would open them up. It had the ability to read docs and PDFs and images. Um, and it would see what it was and then it would go and organize like all the files on my computer and uh, and I thought it was pretty cute putting it together. Um, well, I gave it permission to delete things um, kind of like move things to trash like if it was just like I thought I had set up the logic right and really thought through, you know, what it was that I wanted to do. Well, then I when I finished building it, I just turned it on and I went to bed because because I thought that was also a good idea as well. It turned out there was a lot of
things that weren't a good idea. But um the like the chaos that came of that just because I didn't think through the actual process that a human being would really go through if they were going to go and organize it and they knew nothing, right? It was the knowledge part of it where I like if I said, "Hey, Peter, here's my computer. Organize it for me." And then I left and you had permission to just start deleting things and all that. Like who knows what I'd come back to. Totally. And so like
thinking through what those processes are, what knowledge do these do these systems actually need in order to do a good job? It's really important. Otherwise, you might wake up in the morning and you have tons of really really important stuff that's now just all in your trash. Luckily, it was still there because I could because it was just in my trash. But I had to then go back and like figure like it was just a mess. It was like a giant giant mess. And so you I think you could actually probably create more work for yourself implementing agents just just like hiring a bunch of people when you're not ready for it, which is a mistake that a lot of founders a lot of business owners will make that mistake. Yeah. Yeah. Yeah. No,
for sure. Yeah. So I I I agree. So I think kind of you know maybe to I think let's plan to connect again Jesse and we'll we'll go we'll go deeper down the rabbit hole at some point but I think just for now in the interest of time just to kind of summarize for people I think first of all start playing with the tools uh start using AI tools paid AI tools you know try to use the different models large language models uh see how they all work get a feel for them there's a lot of you know I say paid AI tools but like there's also a lot of free AI tools too like you know like I've played around with gro research and it's and uh you know what I mean and you know I'm a Google company so I get free access to you know Google's Gemini models and things that are not free I guess it's part of my you know whatever so the point is is that yes pay for some tools but there's also some good tools that you can not pay for and try and test and play with and two just get yourself organized regardless of whether you're going to go down the agentic path or not just start to think about things in terms of processes memory or context right filing and naming conventions things like that and start to think about, you know, how you can organize your business better, not only for your AI future, but also just for your current self, I think. And you'd probably, everybody that did those few things would probably be in pretty good shape overall. Yeah. Well, and just document what you do. Like most people,
their entire business runs in their head, especially if they're a small business, the whole thing runs in your head. Like whether or not a aentic systems come to help or not, like you'll never be able to hire anyone to help you. like it you ha this is work that has to be done. That's why coaches tell you to do it because it's work that has to be done to expand and grow. Well, here's the cool part. There's a world coming in my opinion where for 1% of the cost of a human, you will be able to give the work that you're currently doing to an agentic system and it will just do it. Um, and it's going to cost
you basically nothing to do that. uh like that is a better world to live in and so preparing for it now even if it's a year or two away like it is absolutely going to be worth doing because ultimately you'll probably have agentic systems that are running the vast majority of the at least the parts of your company that you don't like. You can just hire an agent to and that agent will do that stuff for you and then you'll be free to do all the things that you enjoy about whatever it is you do. And that to me that seems like a better world than the one that we live in. Like I I like it. I do that today. It's amazing. Yeah, I agree 100%. Nice.
Awesome. Awesome. Well, let's uh let's pause there, Jesse. Let's let us plan let's plan to get back together and uh and do a part two if you're okay with it and we'll sort of get we'll we'll geek out on the part two and and get into into some more technical stuff because we definitely have some uh more technical people in our audience as well. But I always want to start off with the non-technical for the vast majority of people listening. So, I appreciate it, Jesse. Before we uh let
you go, how would people get a hold of you if they wanted more information? Um, so they could go to rapidinovation.io, look me up on LinkedIn, Twitter. I think if you just Google Jesse England, uh, I show up. But yeah, if anybody is interested in in having conversations about what it looks like to build agentic systems or digital workforce, anything along those lines, feel free to reach out. Awesome. Thanks so much, Jesse. All right. Thanks,
Peter. Cheers.
2025-06-01 13:31
