Venture Step

Dalton Anderson (00:00)
Welcome to VentureStep podcast where we discuss entrepreneurship, industry trends, and the occasional book review. Tired of hearing AI without knowing how it can help you? In this episode, we're ditching the hype and we're getting practical. We'll uncover simple, everyday ways to use AI to boost your productivity, spark your creativity, and even streamline your chores. Plus, we'll tackle your nagging doubts and fears about getting started with AI. If you're ready to take the first step,

in this AI revolution, this episode is your guide. Before we dive in, I'm your host Dalton. I've got a bit of a mix of background programming, data science and insurance. Offline, you can find me running, building my side business or lost in a good book. You can listen to this podcast in both audio and video format on YouTube. If audio is more your thing, you could find this podcast on Apple Podcasts, Spotify, YouTube.

or wherever else you get your podcasts. Today we'll be discussing how you can get started with AI. And I think it's kind of important to get practical now and have this episode because my thoughts are AI is going to become more mainstream. We have the recent announcement of Microsoft's

Copilot plus PCs. Microsoft is going to be launching the Copilot plus PCs, which will be a AI integrated work machine. They'll have an AI chip that will be used solely for AI workloads and Google.

has had the same thing in their Pixel phones since the Pixel 8. And so it hasn't been that long, but a year or so they're coming out with the Pixel 9 soon. They have AI enabled phones. Then we have Chat GPTO launching with the conversational stuff. We have Microsoft increasing their partnership with OpenAI and launching these.

AI enabled work PCs and they are seeing AI as a way to win back market share from Apple. Then I think there's potentially Apple's getting in a partnership with Google. So they have, they have these big consumer companies or these products that many consumers use on a day to day Google, Microsoft, Apple,

and

these models are becoming more useful and easier to use, right? Like you don't have to type anything out. You can talk to the AI and you can ask the AI like what it needs from you to have the best prompt or get the best results. And so things are becoming more open and it's not only an early adopter space, it's becoming a space that is mainstream.

and everyone is involved.

So today we're gonna discuss how we could build trust in your relationship with AI, practical use cases for different job sets, some of the first steps that I think you should do when you start using AI, a live demo with some examples, and then I'm gonna discuss how this topic came up. I'm gonna start with how it came up. So I'd recently had...

a family event where my stepsister, Gwen, graduated high school. So I went down to visit my family in Jupiter, Florida.

while visiting, there was a topic about me having this podcast and discussing the things I discuss. And Kay, my stepmom, and Sean, my father, were asking me questions on how to, it seems so useful, but I just don't, I just really don't know how I get started. And some of the things that I think maybe it could help me with, I'm just not comfortable. I don't trust it. How do I?

how do I make sure that I'm giving it the right information to where I'm not wasting time and I feel like things are accurate when I review them and how do I build trust with this thing? And so that was kind of the discussion and I kind of helped them out and gave them examples and showed them some cool stuff while I was there. And so hopefully, maybe they use it for,

some of the committees that they run or whatever they're doing. But that's how it came up because when I do these episodes, I'm really excited about this new, the new thing, the new update and how that new update is utilized. And I'm using it as soon as possible. But there's a group of people that haven't even used AI before ever. So I felt that maybe this is a good,

First step for them is, okay, we could discuss how to get started, how to become an AI workflow user, how to build trust.

Okay. I think the first thing that you need to recognize when using these machines or workflows, whatever you want to call them, these models, these multimodal models is.

They're not going to immediately take your job. I think people have anxiety about these models just replacing, like outright replacing them. In many jobs, they need, you need to have human touch. You need to have a human touch. You need to review it. You need to communicate with other people. You need to make decisions that,

don't necessarily compute to an input -output kind of situation. And so I think that the job displacement piece is coming, but it's not there. I don't think that 20 years from now we'll be having the same conversation. But the current state of these models is you're not gonna turn it on, flip a switch, and then you're gone. And...

So you should think about these models as like your little helper and your little helper helps you.

do XYZ and it helps you save time with this assistant that you have.

these AI models are willing and happy to help. And so you can give it your little tasks and it could help you out and save you time. And maybe it's not something where you're saving an hour plus every day, but if you have a task that you routinely do every day and it's like 20, 30 minutes, and then you start with something simple, like maybe it's something that.

you know, it takes you 10 minutes a day and you can use this AI, you know, open AI, Claude, Gemini, these models to enable yourself instead of it being 10, maybe it's a minute 30. You copy paste what you want. You get the output, you review it, and then you carry on with your day.

You should start small with non -critical tasks, things that aren't important, maybe it's just yourself. So you're the only one who sees this information, it's strictly for yourself, so you can slowly build trust with the AI.

I encourage that you experiment with different prompts. The different prompts will get different outcomes.

If you say something a certain way, it will do something pretty much exactly like that. It's an input -output thing. So if you tell it like, I want an image of red apples or something, and it will make an image of red apples. But if you really want it green, then you have to say so. I mean, it's a simple example, but.

It matters, the words that you use in your prompt matter to the output. It will have a direct influence. And so how you say something is very important. Where it's clear, it's concise. It doesn't necessarily have to be concise, but for me it's good. I like to be clear and concise. And I always say please and thank you. I'm very polite to my AI friends.

Just in case, you know, just in case. But, so I encourage, you know, experimenting with the prompts and understanding that like, hey, like, you're not gonna get everything you want on the first go. And that's all right. Like, you can train the AI, you know, I'll have examples later on in the episode on what you could do to do so.

Okay, so I think that there's some examples for different job sets. We'll start with technical jobs, because I think that's a little bit easier. Technical jobs, and the job isn't necessarily easier, but using it, I think that user group uses AI more often than the others.

So technical jobs, you can use it to help think through difficult problems and.

I do this at work where I have a situation where I need to make a decision on the implementation of something. And maybe it's an internal thing with just myself and I have this coding project to where I have to make a decision on which library to use, how should I structure my for loops, how should I go about structuring my query, should I do a

CTEs or should I make temp tables? What would scale better with this situation? What's more readable? Is it more of a readability thing or is it a runtime thing? And I can ask questions like, okay, for this library using this formula, what are my other options? I think that I could do it this way, but I also, I have this other option potentially.

I don't necessarily have all my information. Do you have any thoughts on which library would, when I say library, I'm just being very general, but which approach would you recommend? And not necessarily recommend, but what are the pros and cons of each approach? Because pretty much in any of these technical questions, it depends on the actual.

outcome that you're looking for. And so with this person that I can bounce off ideas from, it enables me to make a thoughtful decision.

quicker on the approach. So I spend less time on mind mapping, mental mapping the approach and spend more time.

implementing, if that makes sense. Because when you first start one of these projects, you kind of want to make a mental map of how everything should be approached, how you should go about it, what structure are you going to do. And then from there, you start. But if you implement the project incorrectly the first time, it

might require rework later on either because it's too slow, too difficult to maintain.

or just it didn't get the right outcome. So you can use AI to help you think through difficult situations for your projects and help you bounce off ideas to get something to bounce back. And so you kind of have this conversation that you would normally have with yourself or a peer. You just have it with the AI machine.

Okay, so you can also troubleshoot your code. So it's very common to where you might inherit someone's code and they didn't put comments. It's just a mess and it's difficult to understand. And I'm sure everyone who's been programming has had that issue before.

You can use AI to, I mean of course don't put anything proprietary in there.

But if you inherit someone's code or you want to take someone's code off the internet and you don't necessarily understand what the functions are, can you have AI break down the code and explain it? Of course. So AI will be able to logically explain each step and why that step was taken from the script. And so it kind of walks you through line by line, like, okay, like.

This is why we're doing this. This is the potential desired outcome and here are some potential issues with this line. There's a bug in this line. Actually, this is calling the wrong function. It should be calling this function. And so it allows you to understand someone else's code. I almost said my code. I understand my code. Understand someone else's code. Learn new approaches.

by being walked through like complex code off the internet or off GitHub.

help with troubleshooting with bugs. All of that stuff saves an incredible amount of time. Like mind blowing stuff.

They also can do predictive analytics, but I don't necessarily.

use it for that? I don't believe so. Like it can be done, but for simple stuff. But if you're doing something very complicated, that's when AI can't do it. If you're asking very, very complicated code requests and expecting them to understand the full situation, it's difficult. I would use AI to help you make a

predictive analytics framework, like write the framework code, like the data wrangling section, the...

the partitioning and all of that stuff and maybe make some models that they might be good, like a kitchen sink approach and then run your models that you have from the AI and kind of cut down to the best model for the data set. Because you could tell what the data set and potentially in what you want in the structure of the information and it could potentially tell you like a like,

I think that these three models might be good. You should try them out and see how they go. That kind of stuff, of course, but to actually do it, close your eyes and just say, all right, send, run, and then send out the report, no way. It does do predictive analytics, simple stuff. For example, I forecasted my podcast growth and I did that, I think at 50 downloads or something, I asked, they are like, hey,

Here's my download growth. I plan on posting an episode once a week. Can you give me a forecast of where I'll be? I want to discuss certain milestones. I want to do 200 downloads. I want to do 1 ,000, 10 ,000, and I think 5 ,000. And so it gave me the projections and the dates that I'm expected to hit those. And we're pretty accurate so far where I am.

right on the 200 download mark for the podcasts and the forecast is pretty accurate. I think it was saying June 7th. So from that logic, I will hit the forecast like two days early, something like that. So that's good. Management, and so management you could do,

Various of things. It depends on like what level of management you're at. Any management job, you have to send a lot of emails and notes, meeting notes, or not necessarily send out the meeting notes, but you want to have meeting notes just in case. You need them to recall separate events or different teams interacting, or maybe troubleshoot a root cause, like why things aren't going right.

with this project. So you can create meeting notes, you can write emails. If you have a certain email that you write to the team every week, like an update to everyone in the department, and you have a certain structure that you do every email, and it takes you a long time to not only create your thoughts, like write.

write everything down, but then transform that into an email, AI can do that. So you could dictate, like use voice dictation to give an update in Word, right? So you would talk about the progress of the department, major updates, some strategic decisions that were made in the last couple of weeks.

and you kind of ramble on unorganized thoughts, you train the AI to know your template that you send out every week. And then from there, it would take that unstructured thought process that you have in that Word document and structure it into your template email. And it would do everything for you without you having to manually

put your thoughts together into the template. And so that would be something that you could do at a management position.

For processing positions, you could use it for document processing, like maybe that you have to read through.

various newsletters and then summarize them with key points. You could ask AI two things. I think you could ask it, hey, can you summarize this document? Two, can you tell me the key locations that I should review myself? And so you could get this information where you put in three documents from three different newsletter groups that you need to provide to the general audience at the company.

to give them an update on what's going on in the industry. You could upload those documents, ask for a summary, and then...

ask for like, okay, what are some key points that I want to read? Like what pages are this from? Like is this like page three, paragraph four, what's going on? It can give you that information. I do think that you should read, you should definitely read the newspaper, the newsletters, not the newspaper, the newsletters yourself, but gathering the information and writing down your concise thoughts to the group.

takes time, how do you determine what's important? How do you give a solid summary? That can be done with AI.

Or if you have a document where you don't necessarily need to read the whole thing, but you just kind of have to have a general idea of what's going on, like maybe a law update, and you want to make sure like, all right, are there any action items that I need?

it can give you a high level, like what were the interaction between the legislative groups. If there's anything that potentially might need human review, we could call that out. Of course, when you're doing any of these things, you need to verify the information and make sure things are correct. Like don't just take AI's word for it, because AI does have hallucinations.

and it'll also have rejections. Hallucinations are things that are answered, don't necessarily make sense. Like they're not related to what you're talking about or like they're just completely wrong. Rejections like, I can't do that. I'm a large language model. When they have the capability, they just, I don't know, they just get confused. Sometimes it happens. Pretty common on Google's Gemini for some reason. It's kind of annoying.

Very annoying, honestly. Like you'll have an AI chat set up with Jim and I, and for the longest it would do exactly what you want perfectly. And then one day, like three months later, it's like, I'm not capable of that. I'm a large language model. Really? No, you are. You are capable of this. Come on, get it together.

Okay.

So how do you move forward with AI?

I think you should identify some pain points in your day to day life where you think that...

there's a potential opportunity to slowly edge your way in to using AI in your workplace, at home, wherever.

Where that pain point is, I don't know, but you should think about, okay, what are some reoccurring tasks that I do consistently? Do I always send out weekly emails to the team? Do I have issues being disciplined and getting meeting notes put together for the group after meetings to make sure that we have the right action items? Am I having a difficult time getting ahold of my peers to...

you know, think through some of these difficult problems. Like, I need someone to bounce ideas off of and make sure that I have the right approach. Who is peer review my code? Like, if you have a peer review before your peer review, then the likelihood of things coming up are less, less likely.

You could use research tools for your industry. There's AI tools built for each industry. I know Kay has something where they're in discussions of using AI tools for law. And so it would know the case, what happened, how much people got paid, crazy stuff.

And over time, I think that you will gradually scale as you become more comfortable with AI and as society becomes more comfortable.

All right, so let's hop into the live demo of.

Hold on. I'm using Riverside on.

my goodness, no. All right. So I guess I'm not gonna be able to give a live demo, unfortunately.

So I'm using the app this time and I'm not doing it on the web. And that has different issues with me being able to share my screen. So I've enabled sharing my screen but.

because Apple's locked down versus not locked down versus like, Apple's more locked down than Windows is. And so I have to quit this recording and then reopen Riverside.

But I can't do that because I'm already 30 minutes into this episode. So we have a hit and a miss, but I'll explain the things that I did. So I had a meeting notes example created for us. And so I asked AI, hey, I want to create meeting notes template and I'll provide you the exact example when you're ready. And AI responded, OK, whenever you're ready, please send the details.

And so I send them meeting notes, like okay, like meeting notes for meeting name, the meeting date, the attendees, the subject. So each subject would have, you know, during the meeting, you have these meeting topics, I call them subjects. The bullet points discussed in that subject, what are the action items? Like who is responsible for what? And it keeps going through, subject by subject, all the way through.

and I had a transcription that I asked another AI to make. And so this transcription has a whole bunch of different people in there. I think there's seven people in this transcription with different names, different voices, or just general statements from different groups. So it's a mixture of things. There's people, individual people, there's just groups, and it's pretty confusing. So like,

I guess I could read a little bit of it, but voice one, no way we could spin up that environment by Friday. Not with someone interrupts says, look, the client's already signed timelines, non -negotiable here. We promise. But the mockups aren't even approved. How can we get the code win? Sandra, where are we at with the database schema? That's a blocker for us too. That was from it. I sent that over last night. Did no one check there?

Sondra, that's not helpful right now. We need solutions, not blame. Product, can those wirefimes get finalized today?

I'm on mute, can anyone hear me? This is why Agile is a nightmare. We can't sprint when the GoPros keep moving. And so it just, there's just like constant information coming from different people. And it's confusing and difficult to comb through to make your notes and like who's responsible for what. Maybe if you've read through the whole transcript, but it's like.

seven plus pages. And so in the meeting ends and then resumes, so someone jumped on late. And so this person's like, I'm sorry I'm late. I was on another call. What did I miss? It's like multiple people groan. It was just a complete derailment of the project timeline. No biggie. well, I have some good news. Marketing is ready to push a campaign as soon as we have a launch date.

A campaign that barely a product, a campaign for a product that barely exists. That's bold.

So it kind of just goes back and forth. There's some sidebar conversations where people are seemingly frustrated about the situation, but it doesn't necessarily have clear action items. So I copied and paste that into the chat and I said, Hey, I said,

It's super long, so I have to scroll up.

I said, hey, I have a meeting, a meeting transcription, can you please turn this into meeting notes? And it says, fine, no problem. And so it says, it names it the Project Phoenix, the MVP Sprint, the date of the meeting, the time, the attendees, so the attendees were voice one IT, voice two management, voice three product, voice four IT, and then Sandra IT.

Frank, voice five, voice six, management, voice seven, marketing. So it goes through subjects. So it goes project timeline and MVP strategy. Point one was discussing type timeline and inability to make the initial deadline. Point two, decision to focus on minimum viable product by Friday. Action items, and there's other bullet points, so I'm gonna read them all. But it says action items, product team.

finalize wireframes and revise scope by end of day. Sandra from IT, manage the database setup and scalability issues ongoing. IT team, prepare backend infrastructure for NVP by Friday. Next subject, NVP features and adjustment. So it talks about removing non -essential features, discussion with management or managing client expectations.

action items, product teams strips down, feature set to essentials by end of day, Frank coordinate with the development team on core functionality implementation by Friday, marketing discuss the marketing campaign for MVP materials, and then it just goes on and on and on with action items. And it really breaks it down in an amazing way where it takes all this unstructured information and structures it for you.

This is something that would save you, I think, hours of your time. Where instead of combing through this AI script, or not this AI script, but the script, this transcription of the meeting, you don't have to do that anymore. You can have a template of what you want your meeting notes to be, where it's consistently your identity, and you can have AI structure that information for you.

for your review. At that point, you can add things that you don't necessarily think that is there or missing or maybe you wanna emphasize a little bit more. But on a general sense, it's gonna hit the mark with stuff like that. For the most part, it's not gonna make things up. It may not get everything you're looking for.

but it might get 95%. And so you have to get that extra five, but you're not necessarily having to create everything from scratch. So you're saving hours of your time. And for these big, big meetings like this, it's really important to have meeting notes and you have to make sure that everyone has these action items. And if these groups, like when you have multi -departmental projects,

You need to keep each group accountable to make sure that the project moves along. And so if people aren't meeting their action items, then when it's clearly stated in the meeting, and then you send a meeting's notes and say like, hey, blah, blah, blah, you know, if anyone has anything else to add, but this is what I've got from the meeting, here are the action items for everyone and what was discussed.

And it's very clear on what the subject is, what was discussed, what are the action items for that subject, who's responsible, the actual person.

it's very impactful and it will save so much time. It's in a crazy amount of time. I have another example with email. So I ask, hey, I need help writing and I would like you to use my writing style. I would like to train you with a few examples. I'll enact training when I say start training and I put it in all caps and I will end the training when I say end training. Does that sound good?

AI is like, yeah, it's fine. Let me know when you want to start. So I said, start training. And I fed AI five emails of how I would like AI to communicate with other people for me. And so basically the emails have a little bit of humor and they're clear and direct. Okay. So it says, I got it. I have your writing style.

I end the training and then I say, send an email to Sarah requesting for additional information as you're concerned about the timelines and wanna regroup. And so AI says, hey Sarah, I hope you're doing well. I'm reaching out because I'm a bit concerned about our timelines and I'd love to get some additional information. Could we possibly regroup to further discuss? Your insights are always super helpful and I want a quick chat to set up the right time or set us up on the right track. Looking forward to hearing from you.

And I asked another request was sending email to Josh requesting him to send the code for my review. Hey Josh, hope you're doing well. Could you please send over the code for my review? I'd like to take a look and make sure everything's on track. Thanks. And those are very simple examples because I didn't want to make like a huge email thread and I didn't want to make all these massive artificial

emails to send, to train the AI to get that started. But in a general sense, I use AI to help me with meeting notes. I don't use AI to help me write emails. I will use AI to structure meeting notes, because that's very, very helpful, or it saves me lots of time. I will use AI though to write tickets. So I have to submit some...

support tickets and support tickets typically have a structured format where you have the problem, you have how do you, what is the problem, what are the business expectations, how would you.

say, like, enact the bug. Like, what are the steps that you would take as a user when this bug occurs? So, like, I sign in, I go to this button, and I go to this place, and then the bug happens. You know, what's the error message? Where are the examples? All those things is the same exact way for the bug. Each bug, similar structure or service test ticket for bugs or enhancements.

similar structure. So you can have a template for both. Like, hey, I want an enhancement template. Like use this template, train it, use the bug template, train it. And every time you have a bug, you could send in your unstructured thoughts, you know, and then you have AI make the email for you or however you're submitting it. If you're submitting it on some ticket platform or ADO, I don't know.

Okay, you have AI do its thing with the template. From there, you add in your supporting documentation, your attachments, your screenshots, whatever you need. And my other example I made was a coding demo example where I sent in code that had an error. This error was related to...

the calling the wrong variable, it had some logic errors, it was filtering the data wrong, it had maybe six errors. And so I put in, hey, can you help me fix this code block? Someone left the company and I need to fix this code for my live demo. And so I submit the code to AI and AI comes back with a corrected version of the code. And then it also said like, okay,

Here are the changes I made. I fixed the typo in process underscore data. The function was missing a closing quote in total underscore value plus equals item bracket value bracket. And then I added the closing quote. I adjusted the filtered data list comprehension to make it more readable. So it says, I changed filter data equals item, blah, blah, blah. And so it fixed the code and made some improvements to it.

And it saved me. It didn't save me hours, but that kind of stuff does save a lot of time where, you know, as I talked about earlier, is like if you're having trouble understanding why code code works and what it's actually doing. You can use AI to explain like line by line like, OK, like this is what's going on, like this. This is this is the approach. You could do it this way to be a little bit faster, but it might be less readable. So it depends if you're running this all the time or once a month.

And so if you want something more maintainable, then have this approach. If you're purely going for runtime, I think this approach might be better.

That was the fake live demo. I didn't get to do a live demo, but I prepped, I did all the prep work. Really sad. I should have, should have kept it, kept it casual and kept the same, same process and, and did my recording on the web like I normally do. Okay. So we went over some takeaways, right? Well, the main takeaway was, okay, we want to build trust with AI. We want to start small.

and we have practical use cases for each job position as discussed earlier. Management processes technical positions. You wanna start small, you wanna think about non -critical tasks, you want to discover your pain points of where you could potentially slot in some AI help, your AI helper, your assistant.

to reduce the amount of reoccurring processes that you have to do yourself. You wanna be the reviewer, not necessarily the process person. You can automate these things with AI as much as possible.

And if you do that, you'll save a lot of time and you'll have more time to spend on more important items. The items that you want to do or the items that are most impactful for the business.

So I did a demo, a live demo with the meeting notes, code review, and email writing. And each of those I had different ways of approaching, but the majority of the time I'm training the AI for what I want in that chat, and then I'm giving it the information. And the chat or AI in that chat is using the information I provided earlier.

to influence decision and how the output is given to me.

I would like all of us to explore AI. If you're listening to this, I want you to leverage AI in your work and just give it a try. There's many opportunities for you to do so. And if you become...

an expert, I wouldn't say an expert, but an expert at your company, you can obviously open up additional opportunities for you as an individual professionally to improve your career outlook. Because AI is coming and it's gonna become mainstream and if you adopt now, you have the potentially to be the expert at your company internally. Of course, if you have any thoughts, please write them down and put them in the comments.

and I'll see you next week. Next week we're gonna have an entrepreneur on the show. It's gonna be great. It's gonna be a great show and it's gonna be our first. We'll start discussing entrepreneurship and we'll consistently do that. Remember the first 20 episodes were about me finding my voice and finding myself as a podcast host, because it's not the simplest thing. And so I'm still getting used to it and improving, trying to improve every episode. But of course, have a great day.

Have a good night. Good afternoon, wherever you are in this world. See you next week. Bye.