Alexis
Hi everyone. Welcome to another episode of Talk Tech with Data Dave. I am Alexis, your host of this podcast along with my dear friend, Data Dave. And today we’re doing an expert episode.
Data Dave
I’m very well, Alexis, how are you? And I’m very excited about this episode. This is a. I am too.
Alexis
This is a great question and I’m super excited. Listeners out there, I know you’ve heard Dave and I talk about Preston before. He is my boss, the cofounder of D3Clarity with Data Dave. He is here to join us today to help answer this question because it’s specifically about AI. So, Preston, thank you for being our expert today. Welcome to the podcast.
Preston Gregg
Hey, glad to be here. Hello Dave. Hello Alexis. Thanks for inviting me.
Data Dave
Good morning.
Alexis
So, our question today is about AI, and it is about small businesses. Let me pitch it to you gentlemen, and you guys can share your insights. One of our listeners messaged Dave on LinkedIn and sent in this question. How can small to midsized businesses leverage AI to gain actionable insights from their data without requiring extensive in-house expertise? So, it’s kind of a deep question, but I think the long and short of it there is- this person does not have a lot of in-house expertise on AI, but they want to use it to help understand their data better. What do you all think a small to medium-sized business can do to try to leverage that data with AI?
Data Dave
First of all, I want to thank the listener, whoever it was who sent that question in. So fabulous. We love it when we get questions directly from our listener base. I think this is a brilliant question.
Alexis
Oh, I do want to put it out there. I know the listener’s name, but I know that I’m going to pronounce it incorrectly and names are super important to me. So, I’m going to tag that listener in our LinkedIn post when we put it out there. So if you want to know who submitted this question, please go check it out there. I just didn’t want to be disrespectful and mispronounce someone’s name.
Data Dave
Perfect. And thank you Alexis. And again thank you to the listener for posting it.
Let me start and give a little bit of a technical perspective on AI and machine learning, and how I believe as a small business and how we use it to a certain extent as a small business. And I’ll let Preston talk a little bit more about it, as a business, how we have approached AI and addressed AI from a technical perspective and from a business perspective.
If you heard what I just said before, a lot of it comes down to the data as much as AI. And there’s a number of ways that AI can be – or machine learning can be – taken advantage of by a small business from a purely technical perspective.
The first one is simply looking at it as a tool and using it as a personal productivity tool. We use it a lot for proofreading documents, for helping coders, for a number of other things. Helping us understand AI. So just using it as a personal productivity tool where you’re using the pretrained massive models that exist in the market, use them. Be careful, they do not forget anything that you tell them. So you have to be a little bit careful and put it in a box and bound it that way because it is proprietary information. So that’s that side of it from a technical perspective.
Alexis
I definitely use it a lot for this podcast. I use it so much so for this podcast that it believes that I am Data Dave. And so, anytime I ask it to write an email for me that’s not about this podcast, like I need it to write an HR email for me, it signs it Data Dave Wilkinson. And I just love that. I’m like, that is amazing that I gave it no prompts to make it believe it was you. But the account that I have with D3Clarity is so used to me talking about Data Dave and talking about the podcast that now believes that I am Data Dave Wilkinson.
Preston Gregg
So I like this. On that, on that front, I noticed something that I thought was pretty funny last week. One of the things that we use AI for is creating the transcripts that we post for these episodes. And I was getting flagged for a bunch of spelling errors that were in the transcript. And so, I went and looked at it. It appears that it’s picking up on Dave’s accent. And when he talks, it spells everything in the British or the English version of the word versus when you talk, Alexis, it’s the English version.
Alexis
Yeah. So, like behavior is a word that I see very often. Like when I say it, it’s spelled one way, but when Dave says it, it’s spelled another way. And I’m just like, “How does it know? Like how does it know?!” It’s picking up his accent and spelling it that way? I’m just fascinated by that.
Preston Gregg
And it’s not once or twice. It’s consistently over every episode.
Data Dave
That’s funny, actually.
Alexis
And I leave it in this transcript, like, I don’t fix it because I think it’s kind of cool. So, I just leave it that way. And I’m just like, “I like this.” If somebody ever asks, I’ll be able to have this conversation or listeners out there. If you’re wondering why our transcripts look like that, that’s 100% why.
Preston Gregg
Who would know it would be able to interpret data as East Texas accent.
Data Dave
East Texas accent, exactly. Yeah.
So getting back to the question, the other thing, which really comes down to bringing the data, what we’ve also noticed and what we do some of and help a lot of small businesses with, because D3Clarity does do this, it helps small businesses leverage their data as an asset. As I said, the next step before you get into the deep understanding and that kind of thing is to start with, the idea is using AI to do essentially reading comprehension, which is because the large language models are getting really quite good at reading documents now, and they’ve got enough background in the reading documents.
What you can do is, without understanding the depth of the model, how to build a model, how to train a model, or anything else, you can actually give it the document as context of the question set that you want to have and then ask it, “What is in this document?” So, you can essentially say, “Here’s the Encyclopedia Britannica. Now I’m going to ask you questions about it.” And it will read the encyclopedia and then it will answer those questions about it. So, it is doing reading comprehension. So, that is the easiest way to feed your data to an AI model and get meaningful results.
This is essentially a good use case for things like as a small business, if you wanted to introduce a chatbot for customer service or answering phones or that sort of thing, which is: take your knowledge base, the knowledge of whatever you’re answering questions on, and then using that as context for every question that comes in, and then letting the out of the box, the libraries of chatbots and other things that exist. There’s a lot of this that exists that is fairly easy to put together, and just using that and saying, “Okay, well, here’s my manual on my product or on my whatever, then use this manual and let this person answer questions against this manual.”
Alexis
I want to dumb that down for a second, Dave, and make sure I’m understanding you. I’m going to use a manual very specifically. Let’s say I get a PDF owner’s manual for how to set up my cell phone, and I have it in a PDF. If I drop that in AI and I say, “Hey, tell me how to turn on my phone,” it can scan that entire document and just give me the specifics to answer that question. Is that what you’re talking about? As far as reading comprehension is concerned? That would be a pretty easy real-life example, but that could really be applied to anything.
Data Dave
Exactly. If I gave you a manual and said, “Alexis, go read this manual and then come back in three days, and I’m going to ask you 20 questions about it.” And I’m going to say, “How do you turn on the phone?” And you say, “Well, from paragraph four…” whatever it is, right? And then you can say, “Well, how do I turn off my phone?” “Well, from paragraph 96, when I’m done with my phone, then this is how I turn it off….”
So, it gives you that ability. The large language model and the chat engines give you the ability to have a dialogue with the AI engine. And then you add on the document and say, “Here’s a context of information that the conversation is going to be about.”
Alexis
Okay.
Data Dave
And it hasn’t been deeply trained. It’s not a deeply trained expert AI system on this. It is simply giving you a dialogue on this set of documentation based on what I inputted.
And Preston, I think you were trying to add something to that as well.
Alexis
Yes, please.
Preston Gregg
Yeah, I would say I agree with everything that you both said there. You know, about the use of AI at a higher level. Whether you’re a small business or a large business, I think that AI is an important factor. I know, like has been said on practically every podcast episode, that AI is not a magic bullet at all, but it is a very powerful tool. And I feel like we’re really just in the very beginning. And I’m not going to be surprised at all if a few years from now, we’re looking back at this period and thinking AI and generative AI on large language models in particular is as significant as the commercialization of the Internet was back in the 90s. We’re not there yet, but it’s important.
I always say internally, “It’s not AI that’s going to take someone’s job,” because that’s an immediate concern. And there certainly are jobs that will be changed because of AI. But I tell people, “It’s not AI that’s going to take your job. It’s going to be someone that knows how to use AI better than you that takes your job.” And I think you can apply that to small businesses and their competitors as well. And so it’s an important thing to look at. The user was asking about how can AI be used, I believe without a lot of expertise.
That’s a bit of a nuanced question because to use AI successfully, you have to be an expert in the things that you do, what makes you unique as a business. Because the more specific you can get, as Dave was pointing out, by providing it a specific document or data set, the better result that you can get out of it. Now, do you need a PhD level engineer in AI? Depends on what you’re doing, but there’s a lot of scenarios where you don’t. And I think that we’ve already seen it get dramatically easier just in the last year. That trend is going to continue.
Data Dave
I completely agree with that, Preston. I want to build on that because I think one of the comments in the question was around using our data.
Alexis
Right, to gain actionable insights on our data.
Data Dave
“To gain actionable insights.” I think that is the key here. As a small business, I think your data is key, certainly as a knowledge business. So D3Clarity- we’re a knowledge business. We sell expertise. That’s really what we sell. The mechanics are being automated and made easier with AI.
I read an article a little while ago about “Are programmers going to be put out of job by AI” because they can do it better than we can. And I use AI when I’m programming. But the answer I think is NO, because we’re focusing [on being excellent]- and this is where I think small businesses and certainly knowledge-based small businesses and others need to focus even more than they ever have done in being excellent at what they do. Be excellent at what you do. Know your knowledge, know your base of knowledge and then use AI to exploit that knowledge and to make that knowledge, even from myself or from yourself, more accessible.
Preston Gregg
That’s right.
Data Dave
You write it down, get it out there, get it in structures where you can then, like I said before, the chatbot that can answer about this document, that’s great. Use the AI just going off, not my data. There’s a lot of open-source pieces. AI is coming into a lot of structures now. Use it to do your calendaring, use it to schedule meetings, use it for all these things that allow you to focus on being the best that you can be and being excellent in your area of expertise. And then take that expertise and help AI access it from yourself so that you’re not the bottleneck. You’re not the bottleneck anymore. And you’re using AI to allow you to scale.
Preston Gregg
Yeah, absolutely. That’s potential to level the playing field on a lot of things like you’re talking about there, even your programming example, you know, will a programmer lose their job? Only to another programmer that can write more efficiently because they use the new tools properly?
Data Dave
Right, exactly.
Preston Gregg
You know. Yeah, exactly.
Data Dave
And I use AI for writing programs. But you still see a difference in style between a junior and an experienced programmer, even if they both use the same AI engine.
Preston Gregg
That’s right. You have to get into the mindset that AI isn’t a black box that you just go tell it to do things and expect it to do what you would do.
Data Dave
Right.
Preston Gregg
But if you treat it more like an expert that’s limited and a few specific things sitting next to you that you can ask it for its opinion and get some tips while you’re doing what you’re an expert in, then you’re going to be successful, a lot more success.
Data Dave
Right. I’ve got a great story there because when I started programming as a junior program, I got a job writing low level C code, operating systems, and network communications. This was what, early 80s? So, I’m dating myself. So, I literally had a stack of books on my desk next to me that were Kernigan Rich EC for anybody who knows C Beyond Strous and a couple of other books that were textbooks, literally textbooks on C programming. I spent quite a lot of time just sort of thinking, “How to do it?” and then going and looking up and working through it.
I’ll tell you what, having the AI engines, the Microsoft Pilot and the ChatGPT, and so on on my other monitor these days is way quicker than having a copy of Kernigh and Richie and Beyond Strous Book sitting next to me. It’s just way, way easier.
Preston Gregg
It is, it is. It’s amazing.
Alexis
Let’s take this concept that you’re talking about, Dave, this chatbot idea. I want to think back to, I think it was our conversation with Shannon Kelly. We ended up talking about AI, and she was like, “Listen. Garbage in, garbage out.” And that was her way of saying what the two of you keep saying. Like you have to have a core understanding of your data, and it has to be fit for purpose. And then if you apply the AI to it, you’re going to get a good result. But if you’ve got like, not right fit for purpose data and you try to push it into a bot, you’re not going to get the output that you’re hoping for.
Data Dave
So, let me talk about that in a minute. And this is where I think again, drawing into the question and saying without extensive knowledge and small businesses, this is where to a certain extent, I think large businesses have a little bit of a disadvantage because they have a massive cadre of data, a massive set of data across their entire business and then they try and solve some massive problems with AI as well as the small problems. And so their biggest challenge is with a massive set of data that goes back however many years, they simply can’t validate it all. And they have data that goes back in history, and they can’t validate it all, and they put an AI engine in it, and they wonder why they get bad results.
It’s often the quality of the data, and the nature of that data, and the fact that there’s history inherent in that data. And you’ve heard me say this before, which is data is just evidence of history. So, if you’re looking at the history and then you put an AI engine in it or machine learning engine or whatever, and you’re extrapolating a bad history forwards, you’re going to make a bad result, you’re going to have a bad decision. They’re trying to get too big with the problem set.
As a small business, if you stay focused on your expertise and being excellent, and then say, “I’m going to use out of the box AI engine to answer these questions on my expertise, and here’s my expertise,” you can stay very focused and get some very good results. And that happens in large organizations. That might happen at the departmental level, but that’s not where the big massive benefit is.
Now, you don’t have to go all the way into understanding how a neural network works, and how a generative AI network works, and how these things do it, that can be useful if that’s an area for your expertise and where you need particular expertise. But even that is getting more and more straightforward, more and more packaged, easier to put together. My comment there is really, as a small business looking to do this without massive expertise. Focus on your expertise, understand your expertise, and then use the AI engines to help you publish your expertise and work it that way. Because then you’re using the AI engine largely off the shelf, you’re using it as a pool, and it’s only if you need particular areas of it, and this is areas where we can help you, other people can help you, and it shouldn’t be a huge amount of work. If you’ve got your expertise and you understand the expertise that you’re trying to put out there.
Preston Gregg
Couldn’t agree more. And I would add, be selective with your use cases. You can go buy the world’s most expensive paintbrush, but that’s not going to automatically make the art that you produce with a museum quality or even something somebody would want to pay for. Right? It’s the same with these technologies. And what’s right for one use case isn’t necessarily right for the other.
Data Dave
Right. Absolutely.
Preston Gregg
The use cases that we see with our clients that apply to most small businesses are certainly data-driven decision making in analytics and automation. Routine processes that you do over and over again. Personalized marketing. You know, we talked about the chatbot scenario as well in each of those.
It’s easiest for me to think of it as like a layer cake, because AI is like such a broad term and there’s so much noise in the market and permission. It’s like, okay, where do I start? But the way I think about it is in layers. And this is really to address that, “How do you do this without a lot of AI expertise?” And so the top of the layer is most of the tools you’re already using are starting to incorporate AI features, and they may not be very powerful right now, but even in the last nine months, the number of AI components in Power BI from Microsoft or Amazon’s version of BI, QuickSight, Google has one. You’re seeing AI widgets in your email application in Microsoft Word and other packages. So, that’s a way to get started. And Alexis, I know you have a bunch of examples on that. And so that’s just incorporating AI in the tools that you already use.
And then the next level down where Dave was talking about is really the pre-built AI platforms that are being developed by the AWS’s of the world, the Googles, the Microsoft, and then the more boutique players like OpenAI and Hugging Face and Anthropic. There’s a list. But the thing that gets my attention is these organizations are literally investing billions of dollars per quarter in building out these ready made platforms to solve specific tasks that cover a lot because you can start to customize things. And that’s the example where you’re not having to train models. But one thing we did is, we trained a preexisting model on our employee handbook that has all the regulations, it’s 30 pages, and it’s all that. Then we could have a chatbot that could answer questions about our employee manual. And it was a test to see how well it worked. And it actually worked.
Alexis
Really well, yeah, yeah. I asked the hardest question I could think of, and it literally, it did what Dave said earlier. It was like, “Oh, in section 7.2, it says…” and it gave me exactly what the handbook said. And like, I know that handbook back in front and it was really, really cool that it did that.
Data Dave
So this is good. I just want to bring up another phrase from the question, which is the question asked in particular about actionable insights within the data. And I think this is key to that. Uncovering and being able to ask questions about your data, your things, and then actionable insights is one of those strange terms because in the big data world, but everybody said, “Okay, throw all your data in a big bucket in a big library, lake, whatever you want to call it, big data lake, and magically you will generate actionable insights.” It’s only actionable when you decide to take action on it. And I’m just being a little pedantic, but certainly putting your knowledge, putting your base, your structure into that and asking questions of it is a great way to get action. The action? I look at the action as being, “How do I accelerate my response to a decision point?” That’s really action.
So, if you use some of these AI features that are being built into all sorts of things, even if it’s as simple as I take a photograph of a car part or whatever, I’m a mechanic, take a photograph of a car part that’s failed and use that picture to go order the one that image recognition, that image matching. That image structure that says, “This part is available for this much and be delivered tomorrow.” That is AI. That is an actionable insight on a failure state that you saw in that vehicle and took action to get a part quicker than you could have done without it.
So also look at new data that hasn’t been available to you all. The image processing is a form of AI to a large degree. So, now you’ve got images that you can use as easily as a part number in some instances. And this kind of thing, you’ve got more things that you can explore.
We’re doing a project at the moment where they’re photographing everybody as they come through turnstiles so they can match them up with action shots of them in a park and then sell them to them later. It’s something they couldn’t do before, and they were doing it, but they couldn’t match the person with the picture, so they were manually searching thousands of pictures. And this kind of thing now you do much better.
Preston Gregg
You know, another example back to the data question, is predictive analytics. That’s traditionally one of the harder things to try to do in an automated way with data, and there’s a variety of ways to do that. But most people that are interested in using data, have some form of report that they use, whether it’s their profit and loss statement or their sales pipeline or inventories for retail.
Now, I think pretty much every BI platform, whether it’s Amazon Quicksight or Microsoft Power BI or Google BigQuery, they now come with out-of-the-box predictive model that are tuned to standard business data sets. And, you might need to modify your data a little bit and put it in a specific format or label it appropriately for the model. But out of the box, they’re very powerful and can give you true insight that wouldn’t be obvious necessarily without having some help. But the most important part is the ease of being able to actually visualize it in the form of a chart or a line that means something that’s specific to your business. And that’s just the out-of-the-box tools, right?
They all offer ways to go deeper and kind of get into it and learn how to tune it further. But I think that’s a clear place to use the data that you already are using to run your business and run it through some of these predictive models.
Alexis
So, I’ve heard you guys give a ton of really specific things that people can try. I’m not going to try to list them all. Please go back and listen to the podcast. But the one thing that I’m hearing from the two of you is, “You just have to try it. You might not be an expert in it, but if you go in and give it a try, you’re going to find that you’ll be able to use it in one way or another.”
And Preston, that’s how you taught me how to use generative AI. You sent me a link and said, “Here, ask it a question, see what it says.” I just started playing with it, and eventually I learned how to use GenAI pretty well. So, I think for our listeners out there, and specifically for the listener who asked the question, take some of these steps that the gentlemen have shared with you and give it a shot. Bet you’ll find that you can use it pretty well. Would you guys agree that’s the long and short of it, right?
Data Dave
The other comment I would make is just using it to validate as well because it will find in your data evidence of what you know. While you, “I think my experience, I’ve been doing this for a long time. I’m an expert in this field. I think I need to do this next month because that’s my experience, but I’m not quite sure why.” You can certainly feed that data, get that data, ask it the question, and validate that answer. “Does the evidence support this? Does the predictive analytics support this?” Now you’re getting something much better and that will give you confidence to ask it some more robust things. And you’ll be able to act on that insight probably quicker than you would because it’s not a feeling anymore, it’s a fact.
Preston Gregg
Use it for a second opinion.
Data Dave
Yeah, yeah.
Alexis
I think that this is a great place to end the podcast. Preston, I really appreciate you joining us today to answer this question for our listener out there who submitted the question. Again, I apologize, I don’t want to be disrespectful about your name, but we really appreciate the question, and thank you. And then for all of our other listeners, if you have a question for Data Dave, you can always send us an email at talktech@d3clarity.com, you can submit a question on the D3Clarity website, or you can connect with Data Dave or I on LinkedIn and send us a question there. We would love to hear from you!
Dave, it’s been a great day. Preston, again, thank you so much for being with us today. It’s been awesome.
Preston Gregg
Yes, thank you.
Data Dave
Yeah, excellent. Thank you.
Preston Gregg
Thank you. Enjoyed it.
