Eps 98 - How AI is Revolutionizing Business With Heidi Araya Transcript
Erin Austin: Hello, ladies. Welcome to this week's episode of hourly to exit. I have a very special guest for you today, who I'm super excited about. Heidi Araya. Hello, Heidi.
Heidi Araya: Thank you. Yes. I'm so excited to be here as well. Thanks for inviting me.
Erin Austin: Yeah.
Heidi Araya: Well,
Erin Austin: you know, we can't get enough of AI. And so we have a wonderful AI expert to talk to us today.
We're going to find out more about kind of, you know, the more, uh, the latest and greatest with AI. But before we do that, I really want you to get to know Heidi. Uh, how she got here. So Heidi, before we get started, can you tell us about yourself, your business and your journey?
Heidi Araya: Sure. So my journey, I actually spent the past 25 years helping companies become more effective, efficient, and productive at work.
I just saw opportunities from the very beginning that no one else did. So reinventions along the way. But I saw that companies would. Execute these process improvements without considering people. So then I built a brand around inviting people into the process improvements. And that took me deep into the software companies, agile ways of working, of course, and finally leading transformations inside enterprises.
But honestly, a couple of things happened the system inside most companies. Prevents actual change from happening number 1 and secondly, I saw about 3 to 4 years ago that the job I had of leading large change initiatives inside organizations was going to come to an end soon. So I knew I had to prepare myself for a pivot and I just didn't know.
How much pivoting I would have to do in the course of a year. So, uh, I, long story short, I dived deep into AI, took a data science program at MIT, sort of upskilling and building AI assistants. And now I'm building an AI agency with a people first poke focus.
Erin Austin: Wow. That is amazing. Yeah. I mean, I will say, I mean, we're going to point out like all the things that happened since you, MIT, I didn't know that about you, Heidi.
And so, uh, That, you know, a year ago, I mean, I guess it's a year and a half now. I didn't even know like AI was, is formed to me as like Bitcoin or cryptocurrency or whatever it was like out there, there's no way not touching it. That's somebody else. It's not me. And now here we are. And it's kind of all I talk about these days.
So for you, I mean, so much has changed, like, since you decided to like, kind of dig into the AI, like, how has that process been for you?
Heidi Araya: Yeah, it, first of all, it, it's been very interesting that I'm not where I thought I would be. Uh, I don't think I could have predicted where I am now, but yet I feel like this is kind of what I was born to be doing.
So a year ago, a year and a half ago, I just started experimenting with AI. I was a consultant inside a technical consulting firm, uh, helping enterprises with their transformations and, uh, technical, like, app modernization. And. I saw how it could help me in my work come up with workshop agendas. And I thought, well, okay, this is, this is probably the worst AI we will see, you know, that's what they started saying.
Hey, this is the worst AI you will ever see. And so I began experimenting more and more. I found how to summarize web pages, podcasts and all that, because I love to learn. And I thought this is where I have to dig in more. So I was laid off twice in 2023. But I was already prepared for what I was going to do next.
So I hopped right into the data science program at MIT again. At the time, I thought. I'm just going to learn as much as I can, because I don't know what's out there, but I know I need to know this thing. And I remember having a conversation with the program manager at MIT when you had to have a call so that they would accept you in the program and talk about why you were doing this.
And she says, but Heidi, this is so different than anything you have ever done. Like the business statistics, the Python, the data science and all this data wrangling. And so I remembered my words back to her were, Okay. If not now, then when, like, this is the time. And so I, again, I spent a lot of hours learning that discovered by the end that I didn't want to be a data scientist, but that was okay.
At the time. My goal was to just simply know enough about data science, machine learning, AI to lead an initiative or something inside a company like an AI initiative or team. Then I started to build AI assistants and I got very excited about the possibilities there, especially because my first customers were content creators and authors who had a lot of work out there that they wanted to share their knowledge with the world.
And so I, I thought, well, this is kind of my mission is sharing knowledge and helping connect people with information. So I just kind of all ballooned from there, I guess.
Erin Austin: That's fantastic. And so now you have an AI agency. That's how you, is that a new term or is it, did you term, did you coin that term? Or is that,
Heidi Araya: I didn't coin that term, but it is a term that's kind of new.
That's out there that people are using. So yeah, in the past year I had to move from, you know, ways of agile ways of working to this AI from a non technical role
Erin Austin: to
Heidi Araya: providing technical solutions from helping large companies. To serving solopreneurs, small businesses and from dealing with technical customers who were typically engineers to serving a lot of non technical customers.
And lastly, uh, the last and hardest maybe I think was from employee to business owner solopreneur.
Erin Austin: Yeah, yeah, yeah.
Heidi Araya: Well, I, I am one of her challenge
Erin Austin: clients. Thank you very much. They're wonderful. So we're going to talk about AI assistance. And by the way, when we say assistance, it's like the A N T S like, uh, the noun correct, uh, or plural noun.
And, uh, and so basically the custom chatbots. Is that basically what we're talking about? Right. Yeah. And so, uh, first tell us about, cause you know, I believe it wasn't that long ago that I was asking someone about adding a custom chat box to my website and they're like, oh, it's super complicated. And maybe they just didn't know what they're talking about, but, or has it just.
Progress that much in the last 6 to 9 months. It's like, what, what is the history
Heidi Araya: there? Yes, it has progressed very rapidly in the last 6 to 9 months. When I first started building chatbots, there were very few platforms available for them. They were expensive, they couldn't house a lot of information and a lot of them, like, for the 1 that we built for you and others, you can kind of crawl the website effectively and quickly and those kind of things didn't exist.
You know, 9 months ago. So, yes, great progress has been made not only with the capabilities, but also the storage of information. So it used to be very expensive to to do that. And they were saying it would be, you know, hundreds of dollars per month. So that has the cost has come down a lot from from that as well.
Those are some of the factors. Okay. So what is the difference
Erin Austin: between
Heidi Araya: like, Okay.
Erin Austin: The AI assistance, and like, when I go on chat, if I premium, I can make a custom, uh, chat bot or whatever they call it.
Heidi Araya: Yeah, that is a great. Yeah. Custom. Yeah.
Erin Austin: Yeah.
Heidi Araya: That's a great question. So when you build your custom in chat, you can actually program that with.
Natural language, I want this and I want that. And, you know, how, how it, um, how it responds and you can upload some files there. But the limitations are that it's limited to both. The file upload, so it can't crawl your site automatically and pull in on the data, but also the file uploads are limited. And I think it's 20 files that you can upload.
So, what I, they're the work around is to gather all of your information, transport it, Format to text format 1st, and then you can upload it. Um, so you'd have to do some massaging there, but the. You also can't embed that on your website as a chat in a pop up, like, we have done, or in an eye frame. And so they just meant it to be if you did do that.
Say, you had some of your content there. It would be a lot less content, and it wouldn't deliver the same results as the custom assistance that you could embed on your website.
Erin Austin: Right. Yeah.
Heidi Araya: Yeah.
Erin Austin: I'll say from my personal experience, it is amazing. Like how well it can, you know, gather the information on my site and, uh, and help, you know, people answer their questions just from the from, uh, from a chat.
Very different from the customer service thing that you get, you know, when you're. Trying to get through to your phone company or whatever. I mean, it really is just kind of finding the information that you have and helping, um, helping people access it. So, uh, when people are thinking about using. an AI assistant on their site.
Um, I mean, the beauty of it is it helps them kind of, you know, really make accessible all of their, the information on their site. But I imagine that there's kind of some minimum requirements for it to make sense.
Heidi Araya: Yes. And I've had a couple of customers that I actually said, It doesn't make any sense for them to have a chat bot on their website.
And 1 was a gentleman who had visual solutions. So he would do lighting solutions. And so 90 percent of his website was just videos of the lighting that he had set up in arenas and
Erin Austin: big
Heidi Araya: buildings and all these things. And, and with that lack of text information on his website, I said, well, the chat bot Isn't going to perform for you, you would actually have to have documents where you train the chat bot to answer questions.
So if you're willing to invest in building those documents, then we could train it. And, uh, but if not, then the chat bot can't operate with no information. So that's 1 customer. Um, the other customer I had, which recently I told him it wasn't worth to get a chap on his website. He has specialty cleaning business and.
He had 20 pages on his website, and the value proposition was clear. People weren't going to get lost
Erin Austin: in
Heidi Araya: what he was trying to offer. So they could ask a simple question. And I did a test chat bot for him so he could see,
Erin Austin: uh,
Heidi Araya: and it was, you know, how often should you clean your rugs? Uh, do you do this kind of cleaning?
But at the, at the end, I told him, I actually don't think this would be a value to you because, uh, When customers come to your website, it's pretty clear what you're offering is they're not going to have to sip through reams and have complex questions about that. So, um, so I, yeah, I actually told him to focus more on building his brand and his authority in the specialty cleaning space instead of the chatbot.
Erin Austin: Yeah. Yeah. And I would think, you know, just good old fashioned FAQs. Are adequate for some uses, you know, for, for my site, which was, you know, I guess, finally, all these web, these blog posts and things that I've been doing finally paid off. And so I have, uh, a, You know, a fair amount of content on there, so it's able to the blog posts, right?
My podcasts are all on there. So the transcripts from that, and, uh, and including some that I was a guest on and, uh, was there something else on there? I can't remember that. YouTube,
Heidi Araya: any YouTube videos, YouTube
Erin Austin: videos. Yeah. Yeah. And your other website, the. That's right. And I have multiple websites. Yeah. So it all works together.
And, uh, and so it is, it is truly amazing. I'm loving it. So it is a completely closed data set, correct? So, so that's not going outside of the website when we're doing it, the chat box. Okay. All right.
Heidi Araya: Correct. Yes. And That's one of the reasons why I'm very careful about the solutions and platforms that I offer because some of them, I was using one back in October, November timeframe that actually was hallucinating and it was impossible to turn off the internet access to it.
And so it would do strange things like give recipes for tomato soup. And it, it would also at one time, yeah, it would, so it was a management consultant and it would give recipes for tomato soup that people asked it for, they would, uh, it would, uh, at one point. Someone asked it a question and it gave a, like a link to a different website.
So I knew that it wasn't.
Erin Austin: So,
Heidi Araya: uh, so yes, all those things have evolved so much over the past 6 to 9 months. And so we have better checks controls over to ensure that it's not hallucinating, not going outside. And what I say, the little walled garden that you're supposed to be playing in an answering problem.
Um, the other thing that I really wanted to do when I was working on your solution is to have the citations. It's readily available so that when people get an answer to their question, they're able to click and see where did the answer come from? Because that gives more certitude and kind of confidence that it's not hallucinating.
Erin Austin: Yeah, that is so important. In fact, I was just like, interacting on LinkedIn today. Had a post about AI and, uh, and someone mentioned, yeah, you know, I realized that I never get spurses when I use, you know, chat GVT in this instance. And so, and I think people aren't aware that they can get spurses. Get sources, but it's, you know, people are just becoming way more aware.
Like we need to make sure we know where is this information coming from? We know that, you know, that, uh, generative AI hallucinates, we know that it steals stuff, we know, and so we need to have that comfort of having sources. And I'm glad to know that, um, you know, on, on my site, it's, it's great. Cause it, you know.
Yeah, I mean, you get the answer from the bot, but to be able to go to the source is really important as well. And, uh, and then back to that being a closed dataset. I mean, um, it sounds like some of the risks that everyone is afraid of when using, you know, kind of the chat GBTs of the world. Those are eliminated when we're using our, uh, custom chat bots.
Is that right?
Heidi Araya: Yes, correct. Yes. And none of the it's not trained on your information and material. So it's all secure and private on a sock to GDPR compliant. I so 27, 001 and all those compliant platforms. So, so, yes, that's another reason for doing a custom AI assistant instead of just going through a chat one.
Erin Austin: Yeah, very nice. Very nice. And so are there any best practices? Practices that we need to think about, like, what, what are, like, when does someone
Heidi Araya: know that they're ready for you? Let's put it that way. Interesting. So perhaps there are a solopreneur expert that has a lot of information out there. And 1 example would be an author with a lot of.
Books like my 1 client who has 28 books and or the, the, uh, 1 of them has, I think about 500 and 500 blog posts over 17 years. Wow. You have a lot of content to share even courses, right? I have another client who's interested in. Updating their courseware with a real time streaming AI avatar. So it would be, there's enough information there where the AI assistant could operate in a way that it gets good information.
So if there's not a lot of information out there, then I would recommend that person build their brand and create information first so that we could train the chatbot. But if you already have a lot of information out there and you want to share your knowledge or, or monetize that knowledge in some other way than you're doing today, then an AI assistant could be that one.
That thing for you
Erin Austin: go back to what you look, you mentioned with the course creator, some kind of live streaming. What was that?
Heidi Araya: Explain this to us. It is. So it is now possible to have an, a live streaming avatar where someone would. So say you have a course that you wanted to teach
Erin Austin: and
Heidi Araya: you have an AI, avatar that could kind of be waiting for a question.
So you'd go to a portal and then the person would ask a question either by voice or type it in and then have your avatar reply to that question based on your knowledge base.
Erin Austin: Whoa,
Heidi Araya: that is so cool.
Erin Austin: So it's basically teaching. It's like having a teaching assistant, basically.
Heidi Araya: Yes. Yes. Um, so it's, It's new and not a lot of companies or people are going for that solution right now, but you could imagine how much more engaging it would be to learn something if you had sort of this person there in front of you, instead of just engaging in the chat.
So that's one of the things that I, I'm super excited. So I'm exploring for my own self, how much it would take to have that happen for people. And, um, the other thing is I have a client who's interested in updating their old courseware. And they have the transcripts for that, and the content is still good, but yet the videos are probably 10 years old.
So this person did ask me and I told him, oh, you could certainly create an avatar of yourself and that would be your own voice and your own persona, you know, physically. And then we could actually. Train it on your courseware. So you already have the transcripts
Erin Austin: and
Heidi Araya: all we would have to do is update it to however you looked currently and make any tweaks to the content.
And then your avatar would just have your course. So it's a much I'll say lower course creator. Who's got some stuff they want to update if you have your old content. So that's another opportunity. That's not a real time streaming avatar, but it's still kind of the same where you could have your avatar there.
Erin Austin: That is amazing. So as you mentioned that it reminded me of the way old video games used to be made where I think they had to kind of produce multiple options and then somehow it would choose the right option based on what you did. Was that a I or is that just,
Heidi Araya: well, It's a, it's something that I was, I actually was wondering if we would get to talk about today and that it's like a decision tree.
Yes. Right. So it's more like, okay, it's kind of falls in machine learning where you would have to go a decision, you know, a decision tree or so. So, not really the kind of conversant. More intelligent that we have today, but those were annoying because remember, if you call up the phone number and they say, do you want you to talk to this or talk to that?
You press a number. And if you don't answer exactly what it wants, it gets confused and it won't respond. Right? Yeah. So we've advanced a lot beyond that. But, yes, some, uh, those were. Probably typically decision trees, I would think.
Erin Austin: Oh, interesting. And now we have something dynamic and generative. So we can take that is so interesting.
Well, that I mean, I really look forward to seeing 1 of those when 1 of those live streaming lens with avatar is public. Please let us know. I would love to see how that works out. That is so exciting. I think people are going to be. Extremely excited about that. Um, so what else? I mean, obviously, that is more of the trends that you're seeing.
I mean, things are happening so fast. I know it's probably hard to keep up, but what can we also can we look forward to in the last half of 2024?
Heidi Araya: What a great question. So, the next thing I think people really start to dial in on is this idea of AI automation. So, more than just an assistant. So, a simple assistant can, you know, you can chat back and forth with, but it's not really taking any action on its own.
Erin Austin: Okay.
Heidi Araya: So, agents. There's something called AI agents that will actually take autonomous actions on their own. So they might go through complex workflows. There's 1 person on LinkedIn. That's kind of advertising his advanced AI agents that he's created a series of of them that operate intelligently on their own.
They go out to the Internet. They solve coding problems and they come back and, you know, when he wakes up in the morning, it's created 100, 000 lines of code overnight with his series of agents and contributed to open source projects. And so. That's kind of coming up and also scary to think about.
Erin Austin: Yeah.
Heidi Araya: Um, but the other thing that might be relevant for small businesses and people thinking what's next with AI would be just, uh.
AI automation, so for content creators, they might have some way that they automate their blog post creation in a way that it, it generates an image as well. So it could go through this series of gathering information, generating a draft blog post, creating an image, and then have you, of course, add your own voice to it before you go out there.
People who automate the whole thing. And I don't recommend that. Right? Um, and the next thing is, is voice in AI. And that's something I'm also experimenting with. So this weekend I set up a, Okay. Uh, phone number where people could call in and schedule time on my calendar and so it's not ready for prime time yet.
But, uh, voice AI assistants are going to be big. The AI agents are going to be big. And I think that the avatars are going to be big. It's still very new, but those are the kind of things I see coming.
Erin Austin: Yeah, I think I saw it was in some, uh, mastermind, I don't know, I think it was one that we share, but someone was mentioning that their bank had offered to have voice, uh, authentication, like, so that instead of, you know, pins or things like that, but is that a good idea?
If we have, uh, these voice avatars, if someone can take a recording, they can take this podcast. Recording and create a voice avatar and then use it to get into my bank account.
Heidi Araya: So actually that's a risk that no one had thought about when they put the voice prints because that was, I think that came out maybe five or six years ago when I was asked to add that to my financial services account.
And so no, probably that alone isn't going to do it anymore.
Erin Austin: Definitely.
Heidi Araya: I was risky, but now they have other more advanced applications to figure out whether it's actually you, you know, uh, so I know they're advancing rapidly. I don't. I don't know of any banking or any serious application that's only using a voice right now because of that, that risk.
Erin Austin: Yeah, very good. Yeah. Well, we know that we're having issues with actors and actresses because of the ability to reproduce their voices, uh, very accurately. So, all sorts of fun times for, for litigators. I'm not that kind of lawyer. So, but someone's going to make a lot of money from
Heidi Araya: this stuff. You know what?
I was thinking, Aaron, I, well, I was doing my MBA years and years ago. So it was about 2011. I looked it up to February of 2011. I'm in my MBA program. We had to take a class on innovation. And there's this thing called the theory of inventive problem solving by this guy, Alt Shuler, who was a Russian guy analyzing patents, and we actually had to write a paper on our, uh, what we thought would happen based on his analyzation of, uh, sorry, analysis of patents.
Patents, um, and how people could create new patents and basically the future of things. And so 1 of the things was removing the human from the loop or combining things. So, just as an example, it used to be that heaters and air conditioners were separate, but now we have heat pumps, which effectively can operate.
Both 1 of them was the other 1 was, of course, removing human in the loop. So my paper that I wrote back in 2011 was soon we won't need actors at all. And there's going to be an interim state where. Where the actor licenses their likeness to a movie because we all want to see Tom Cruise or whoever, uh, but soon we won't really care because it's going to be good enough and we'll get used to not knowing that actor.
But then imagine people who write books that they have this vision and the series can go on for 20 years, like the James Bond series, and they won't get disappointed because all of a sudden Sean Connery is too old to play.
Erin Austin: That's exactly right. We have a whole Dorian Gray issue here, I think, going on. So we'll just.
Stick the actor in some room who never see him again so he can age while he stays beautiful on screen. I like that. That was very prescient of you. It's sometimes it's really hard to wrap your head around the possibilities, honestly, I mean it's. They're just endless on, you know, so, oh my, okay. So as you know, this is a pretty meta podcast, uh, mostly for female founders of expertise based businesses to help them, you know, make sure they're building a business that is scalable and maybe someday saleable.
And so you fit into that. Avatar, so I'm wondering what your plans are for your business and are you thinking about maybe scaling your sale selling or what the next chapter would be?
Heidi Araya: Yeah, that's a very interesting question. Because 1st of all, I never thought that I would be a solopreneur and end up in this, but I absolutely am thinking about it because I see that there is.
There's desire and need and I have to turn down projects every day because I can't handle them or they're just not in the area of scope of my business. And so I, I have 3 interns and already someone else helping me on the side and I'm just wondering at what point it will I. Will I choose to scale or prepare my business for sale?
So I'm already planning on reaching out and, uh, and and thinking early about those things in case it does happen. And of course, you'll be 1 of the people that I reach out to to make sure that I'm doing things correctly along the way.
Erin Austin: It's
Heidi Araya: never too
Erin Austin: early to plan. So I like to say that the things that we do to prefer.
Prepare for sale are the same things that we need to do to scale. So it's not like, you know, you need to be thinking 10 years down the line, those things that you need to kind of help decouple your income from your time. Those are the things you need to do now too. So very good. All right. So a couple of last questions.
So, you know, I believe in creating a more equitable society and I'm wondering if there's an organization or a person who's doing great work in helping that happen that you'd like to share with the audience.
Heidi Araya: Yeah, so I recommend the Malala fund because it really focuses on girls. And of course, education girls getting their proper education is so, so important.
So I love their focus on on girls getting their education because without an education, we really can't begin to compete and know what our potential really is. So I love the Malala fund.
Erin Austin: Yeah,
Heidi Araya: absolutely.
Erin Austin: I agree 100 percent and there's very little more important than making sure we have that.
Independence through education, so we will make sure that is in the show notes and to that end. Also, um, is there an offer? I know you're probably overwhelmed. I think you are going to be even more overwhelmed, but is there an offer that you'd like to share with the audience or what's going on in your business?
Heidi Araya: Yeah. So actually I'd love to hear from solopreneurs and small businesses who are really struggling to scale, maybe struggling with overloaded. Uh, sorry, struggling with too many phone calls, unable to answer them, or perhaps like you, they got to sit in a meeting that could have been just answered by the assistant on their website.
Um, so definitely, we're looking to help those folks scale and grow. And also there are some folks reaching out now who are ready for that next step into automation. So, automating some of their processes and I think, um, so that that's my offer to just reach out and have a chat. If you think that might be you.
But also I have heard a lot of stories that people implement automation or AI assistance without talking to the people who are impacted. And so 1 of the things that I take very, very seriously is including the people who are impacted in the conversation when we're thinking about automation. So I lead these AI exploration workshops that are also very people focused and we talk about opportunities inside our company for automation.
Um, so another conversation is. If people are just uncertain where to start with AI in their company, we can do have an AI exploration workshop with your workforce.
Erin Austin: Yeah, we don't want to leave people behind, right? That is that that is the danger, isn't it? To leave people behind in all of this conversation about about artificial intelligence.
So where can people
Heidi Araya: find you? So you can go to my website at BrightLogicGroup. com. You can find me on LinkedIn, Heidi Araya, and yeah, those are the places that you can find me hanging out.
Erin Austin: Awesome.
Heidi Araya: We'll
Erin Austin: have all of that in the show notes as well. Heidi, this has been a pleasure. Thank you so much for sharing your expertise with us
Heidi Araya: today.
Thank you so much for inviting me. I look forward to meeting you again and watching the podcast when it goes live.
Erin Austin: Awesome. Thank you.
Heidi Araya: Thanks, Erin. Have a lovely day. Bye bye.