September 13, 2024
I had such a great chat with George and Dean on the AI Impacts Podcast!
We ended up talking about everything – from how small business owners are using AI to get more done, to how the big players in venture capital are using it to spot the next big thing.
What I loved most was seeing how this tech isn’t just for the tech experts anymore – it’s actually helping all kinds of people do their jobs better.
George and Dean got really fired up when we started talking about the future of work, especially how more people are choosing to freelance these days and all the new AI tools that are popping up to help us work smarter.
Video
Transcript
Dean Hanson: Foreign this episode we’re discussing the impact of AI on startups and professional services. We’re joined by Mike Smith. Mike is a management consultant, ey, and also a venture capitalist at Scrappy Capital. Hi Mike, and welcome to AI Impacts.
George Little: Welcome to the show.
Mike Smith: Thank you. I’m happy to be here.
George Little: Yeah, Mike, I think we should jump in first. A lot of our guests are in the kind of sales, marketing, go to market space and I think your sort of expertise falls a little bit in parallel to that, but perhaps outside of it and kind of the management consultant, venture capitalist space. So before we jump into kind of the tools and stuff that you’re using and how you feel about all the workflows that have been changed, could you give us a little bit of a sense of like what your day to day is and what some of the things and tasks that are being offset by AI are starting to come into play?
Mike Smith: Sure, absolutely. So usually I describe myself as a management consultant by day and a venture capitalist by finite. And what does that mean tactically? So on the management consulting side, I work mostly on what we call transformations, which are either an acquisition bank buys bank, insurer buys insurer. I tend to be focused on financial services, so banks, insurance companies, wealth and asset managers. And then I have a small venture fund that I invest through nights and weekends. We look at emerging industries, so distribution, logistics, media, and then what I call the future of work, which is human capital management, hr. Really anything that enables where work is going to be in, let’s say, 10 years. So that’s why the whole AI craze is super interesting to me.
George Little: This feels like a good place to dive in. Describing as the future of work and then wrapping in HR is not something that people normally hear. So can you kind of give me a sense of where that line of thinking is and how AI is I’m assuming, impacting that at some point soon?
Mike Smith: Sure. I think like anything else, there’s a bunch of layers to this. So layer one is just the work tools. Let’s talk about one of the big things I’ve been tracking for the last few years, and George, you’re a good example of this is I think we’re moving away from W2 employment. I don’t think it’s going to go away. But if you look at the numbers across the economy, we’re seeing less employment, more contractor work, less. Some of that is driven by benefits, some of that is driven by comp. But it’s challenging for a company to have an employee benefits, healthcare salary, all these Things. So if you see that’s changing a lot in all industries, not just one. What does that mean? Well, I’m not self employed, but if I was, I’d need payroll, I’d need a website, I’d need finance, I need to build clients. So this whole suite of tools are being built to enable people to be one person shops. And that’s before AI started kicking off, that was important. And now if you can be a one person shop plus AI, there’s limitless potential of things you can do.
George Little: Yeah, I mean, I think is there a part, a sort of through line in when you’re working with some of these, some of your clients right in this sort of transformation space where they’re actively considering, you know, replacing a human LED job with an AI workflow or I mean, is it kind of getting to that point?
Mike Smith: I personally don’t see that some of that might be that in acquisitions you’re not generally doing new fancy technology. You’re trying to duct tape the two things together and make sure they work. But I think one of the things you notice, and I have a really interesting lens on this because by day, billion dollar companies, biggest companies in the world, I’ve worked with big logos, super impressive companies, and then startups, no one’s heard of any of these companies. So I get to see both sides of the ecosystem. And one of the things you notice is startups and small companies want to be big companies. They want to act like the Googles, the Facebooks, the EY’s, the Deloitte’s. Big companies want to be innovative. So what does that mean? It means that these large companies really want to adopt new technologies. But it’s a struggle because one of the things I didn’t realize that would be such a struggle with adopting new technologies was the compliance layer of it. So let’s talk about consulting my firm. And I would assume most other firms don’t let you just use ChatGPT, Claude, things like that, because there’s a data privacy concern for our information and for the client information. No client would be happy if I just shoved all their information into ChatGPT and was like, figure this out. For me, that’s probably in 2024, the correct move. Because if I shove some stuff into an LLM, I have no control over it. I don’t know where it’s going. So I think the way that big organizations tend to look at these things first is from a risk profile. These tools are cool, they’re fancy, they’re exciting, but I can’t risk my information going somewhere where I don’t control it. So that’s generally the first lens, which is why I don’t see AIs taking our jobs happening so quickly. I think the first layer is AI.
Dean Hanson: Augments jobs in terms of your own personal tool use when you’re, when you’re doing like the brainstorming, etc. Would that be, would that just be like a chatgpt and then just on your own personal device without providing it any specific contextual information or not very like not nothing particularly sensitive.
Mike Smith: We don’t put anything client sensitive into the tool. And the tool is not chatgpt. It’s something that EY has made. I don’t know the technology, technology behind it. It’s very it. I’ve used ChatGPT in my personal life. It functions the same, same Q A dialogue thing. And then the other piece, which is an interesting layer of this is client computer, EY computer, personal computer, client data. I would never put client data into any of these tools, but I can’t even get it into our tool because they’re on separate systems. So part of it is the way that I kind of approach things is my job right now is to build a plan for integrating a portfolio of, of bank loans. I cannot say LM figure this out, but I can put in to our version of Chatbots. How would I think about building a plan for integration of bank loans?
George Little: What are the abstracted version?
Mike Smith: What are the abstracted thing and that like whether the client is bank A, B, C, D or B. Like it’s a good starting point to help me think about. Oh, I didn’t think about that.
George Little: Right. Yeah, I mean, I think we’ve, we’ve had a through line in a lot of the conversations we’d have about sort of a creative partner or sort of a business partner that these LLMs or these, you know, chatbots more than LLMs can actually, you know, assist you in the abstract or guide you in the right direction without giving you the answer. So fascinating to hear that even at like the scale of Fortune 500, that’s something too.
Dean Hanson: Quite often as well, we find that it’s not even necessarily that they give you the answers that you really want. It’s that they stimulate thought patterns and sort of, you know, lateral thinking, which thinks, oh, okay, that’s not quite what I was looking. But actually, how about I might use this idea instead or it leads you to another place. So it’s just that, like you said, that brainstorming or thought showers, I Think we’re supposed to call them now? I think yeah, it’s that ability to generate ideas quickly.
George Little: Yeah. So I’m curious about some of the negative impacts that you’ve experienced with these. Whether it’s at the sort of management consultant scale or at the work you’re doing with startups. Where do you see the downsides in working with these tools?
Mike Smith: I think the starting on the startup side, I think, I don’t know if it’s a downside, but I think one of the first considerations is the AI ification of everything. Every single technology ad I see now is we use AI to do xyz, which is fine, but AI means a bunch of different things to different companies. A website builder uses AI. I mean you can make the argument that a WYSIWYG editor is AI before AI became a big thing. But I think there’s, I think it’s hard to understand what is truly worldbreaking AI future model on the frontier versus oh, you have like an automation. So I think teasing that out is really hard because I get a couple, I get a fair amount of companies across my desk that are, we use some and I’m like, I can’t tell if you built some really proprietary AI model that’s going to change the world or you have a pretty design wrapper on ChatGPT and there’s value in the design layer. Like there’s a lot of tools that just make a better interface for ChatGPT or fog. And one of the things we’ve noticed about computing is the design layer matters. You know this as 20, 30 years ago it was, you’re writing something in a command line, it just needs to work. But now, before I can even determine if it’s going to work, I have to see if it’s design friendly. So on the startup side, figuring out what’s actually AI versus air quotes, AI is a big thing. And then on the larger side I’m thinking about drawbacks. I think hallucinations are not the biggest concern, but something to be mindful of. I would never send something out the door without giving it a review. I think it’s really good for a rough draft. But in this world where we’re being paid as a consultant quite handsomely for our expertise in the area, it becomes risky if I pull something from ChatGPT and it says this statement is fact and then you send it out the door. So I think proofreading becomes more important, stress testing things becomes more important.
George Little: I mean, I think both of those. Yeah, yeah. And I Think both of those things go hand in hand, right? Like the, like, even the tool that we use now for this podcast is Riverside, and they often offer, like, we’ll make this AI episode for you or whatever. And, you know, as it turns out, like, it does, but it’s like, by definition, but it’s not nearly the episode that we want or the decisions that it’s making or the limitations that have been put on it. Make it such that it’s like, I feel like you’re trying to sell me this AI tool, but in reality, it’s a feature that I can run a handful of times in these, like, very, you know, succinct small areas. I’m just curious, like, with some of the startups that you’ve seen, seen, what’s the kind of level of that? Like, a lot of them, do they feel like more like they’re features that are trying to build on top of something or, you know, like, how do you even get into understanding whether or not there’s any reality to the growth around some of these tools?
Mike Smith: Oh, man, that’s hard. Because I am not an AI expert, I’m not an LL developer. But I think an important point that you just mentioned, and I think a lot about this for crypto, I think the smart companies use AI as AI or crypto as a thing that enables a bigger business process. The top line sentence is not. If I see the top line sentence as we do AI, we do crypto. I’m like, cool, that’s a buzzword. I have no idea what that means. But, for example, one of the companies we’d invest in is a company called Credenza. They do blockchain technology, ticketing, but their top line sentence is not blockchain. Blockchain, blockchain. It’s not. Look at us, we’re doing blockchain because I think it’s hard to sell an outcome around AI. Sorry, AI and blockchain are not outcomes. They are functions, they are processes. So you can’t go into a sales situation and scream, AI, AI, crypto, crypto, crypto. Because what are those things leading to? So his sales pitch is, we manage your data better across the different silos that exist. That’s what gets the customer to buy, not rah, rah, AI. Because any rational buyer wants to solve a problem. So whether it is 100 people in the Philippines doing the work behind the scenes or it’s AI, the customer doesn’t really care. They want the solution. So I think the thing that I look for first is what is the solution you’re selling? Because I don’t really care how the sausage gets made. I don’t think most people do. And if you’re selling first with AI, AI, AI blockchain. Blockchain, you’ve obfuscated what the actual solution is.
George Little: Yeah, that’s. I mean, yeah, I was gonna say we have a background in blockchain. It’s actually how we met and I think that was a big common through line of what we would talk about in our marketing conversations is we’re talking about. It’s as if we’re selling tractors, but we’re only talking about like piston technology on the shock absorbers or something. Like it’s this function that people sort of have elevated as like a means of community in a weird way.
Mike Smith: And are you even selling a tractor or you’re selling the fact that fields are being plowed?
George Little: Because I think. Yeah, it’s true. Right, exactly, exactly. But yeah, I suspect that that’s happening quite a bit with a lot of this sort of AI wash. The same thing happened with the sustainability movement 10 years ago. Like everything was a greenwash of, you.
Dean Hanson: Know, and I would say as well with a AI, another thing that’s, it’s the, it’s, it’s. As we touched on this earlier, it’s the definition of AI. What does AI even mean? Right. Because I’ll speak to like, you know, old school guys or whatever, call them our customers and they’re like, well, that’s not AI. You know, that’s, that’s, that’s clearly not AI. It’s just some sort of machine learning, some sort of whatever else that’s not AI should be. They have this expect expectation of what AI is, which is, you know, like how or something like that. Whereas you’ve got other people on the other end of the scale which like you say, confuse or conflate automation in AI, which is really, really, you know, not AI at all. But then there’s this sort of melting pot in the middle of what do we currently mean by AI? Is that definition fixed? Is that definition going to evolve over time?
George Little: Right. I mean, it definitely seems like the term itself is used as like the forefront of whatever the current artificial intelligence technology is, right. Like six years ago it was machine learning and a companies were only talking about how they have machine learning and that was the forefront. Now it seems like it’s moving towards this LLM status. Right? Like if, if your system is not backed by some sort of LLM that can spit out sentences and communicate with you Then maybe it’s not actually AI at the moment, but who knows where that’s going to go, Right. Quantum could end up in there. I mean, there’s loads of trajectories, I’m sure.
Dean Hanson: I saw an AI powered toothbrush. Oh, I saw something that was talking about an AI powered toothbrush. It’s like, you know, when like cars first started getting turbos. Right. And then everything became turbo. Like you get a turbo, like hair dryer and you get a turbo. I don’t know, whatever it might be, microwave, I don’t know. But like everything has turbo in the. And the same with AI now it’s just becoming like a.
George Little: That’s actually a really good analogy because it was, you know, building onto an older tech.
Mike Smith: Right.
George Little: You have like this combustion engine. You’re trying to get more efficiency out of it by adding this piece to it. And that’s. It feels like it’s not too different than a lot of the tools now. Like you have a word processor and now there’s this thing that makes it more fast or efficient or suggestive or whatever it is. But fundamentally the product is still the same. Anyway, good tangent. So I think would love to jump into a little more detail about some of the kind of day to day pieces of these tools that you’re using. Sure. I think more on the startup side is where I’m kind of really curious. Like when you’re going into conversations with some of the founders or some of the technologists at these companies, like what are the conversations around when it comes to implementing AI? Is it more high level like this is the outcome or does it always end up sort of down in the weeds?
Mike Smith: Well, I think the one thing, one second, I think the conversation is it goes back to being outcomes driven. So these big companies and small companies want to implement AI, right? They want to have a use case. But I think the challenge is finding the use case first. So I think one of the things I’ve noticed over the years is you start with the use case first and you say, what are you trying, what are you trying to do here? Because you don’t always need AI. But I think the other piece of it is that a lot of folks end up getting shiny object syndrome here with AI something I’m. I know the technology is powerful. I know I can do all these things. I know I have these conversations with myself and other knowledge workers. Figuring out how it fits in is so hard to do because so much of what we’re doing as knowledge workers is not repeatable stuff. So I know that AI can schedule my tweets and my LinkedIn posts for me. Cool. That’s saving me a small amount of time. Yeah, I know that it can. Whenever I play with mid journey, I have Albert Einstein riding an elephant. Cool. Great. I’m not a content marketer, so it’s less valuable. I think the, the more you get away from consumer facing, the harder it is to integrate it into different use cases.
George Little: Why do you think that is? Like, what’s the big. What’s the big roadblock there?
Dean Hanson: Is it that the solutions don’t necessarily exist? Is it the solutions aren’t robust? Is it that they require too much customization, manual intervention that actually makes it redundant?
Mike Smith: I think that is a big piece of it, the customization and the manual intervention. I would love a perfect world for me as I put my work into an LLM. It does everything. I walk away, I come back and I pick up my paycheck. Like that’s ideal, right? But that happening is so far away because I think of it a lot like an intern. I want to grow and coach people. I want to get intern work. But I’ve had. Years ago I had an intern who was a great person, but I spent so much time training them up that by the time they were gone, like I spent so much time into developing them that my work fell behind. So I think the same is true with AI that I would love to shove all my stuff into AI, do no work, get a paycheck, but it would take me so long to oversee it. And the same is true of hiring virtual assistants. The onboarding to these tools and the management is really, really challenging. And then you add on the privacy layer. Again, I wouldn’t put client data into a thing, but I can’t. So that’s an option two, training it and telling it what to do. I think it’s kind of like navigating. It’s not really hard to get on the road and drive. It is hard to figure out where you’re going in a world where you don’t have Google Maps.
Dean Hanson: So in terms of. Yeah, as George mentioned, in terms of the, I guess the tactics that you sort of do use or recommend or see effectively used within, within startups. Is there anything sort of that. That is, that you see any commonalities, any. Anything that you would recommend and you say, look, you might want to just do, you know, X, Y or Z, just use this like, AI solution to give you that little bit of a boost as opposed to it being a fundamental change. But Is it something that you would say as a commonality, that you’d say, you might want to use this?
Mike Smith: For me, like a very low lift thing is changing the tone of things. I find myself in consulting. We’re always putting out fires. And even in startup land, I tend to be a little brusque with like, let’s get to business. It’s helped me be a team player. Rah, rah, write a better version of it, be more coaching. It’s nice to help with tone when sometimes I’m just like, here’s the exact thing I want to do. I don’t care about anything else. It helps me celebrate our team, have a better narrative. So it’s really good. And then something my fiance put me onto is like asking it to ask me questions back to challenge my thinking to saying, you know, build a marketing plan and ask me questions to clarify my thinking. Because then it pushes me. I push back and back and forth.
Dean Hanson: Hmm, that’s interesting. We’ve. We’ve heard that before, certainly around challenging sales people’s assumptions on their Personas that they’re trying to sell to. So I think this about a certain Persona AI, you know, like, is that correct, Will? Is that how a Persona would react? Have I got this correct or should I think about them in a different way?
Mike Smith: Yeah, yeah. The Personas is super interesting because again, from the consulting lens, we tend to parachute into these situations where we have, like, I believe consultants fundamentally get paid to move the ball forward. In the last three years, I’ve been put into bank integrations, software transformations, backup planning, process optimization, automation, like, all of these things that I’m a pretty good business person, but I had slim to no knowledge before jumping in. This current engagement I’ve been on for six, seven months is the first time I’ve been able to use an LLM to get me up to speed on knowledge versus Google. And is there a huge difference? I mean, I think I’m getting to speed faster because I think it does a better job of summarizing. And I can say, what are the three key things to know about backing a bank software with Google, it’s like, here’s an article. SEO, clickbait stuff. I think it gets to the heart of stuff better.
George Little: Yeah, yeah.
Dean Hanson: And I use. Since. Since the call with the. Since the episode we had with Stephanie a few weeks ago, I’ve been using Perplexity AI for exactly that purpose instead of using Google. If I’ve ever got any questions to ask which require, you know, which I’m not just Looking for a specific website. If I want an answer and I want to summarize Danza, then Perplexity for me is just a fantastic tool because it also gives you the sources. It says look, you know, it mentions the sources, it’s pull, it summarizes it, it suggests follow up questions that will allow you to dig deeper into the subject if you like. And for those sort of questions, I will never use Google again. Like at least not standard Google because so limited.
George Little: Yeah, totally. There’s a tool that I’ve been using as well with ARC browser. They have a mobile version called yeah, ARC is great and they have this ARC search. And by the way, just as an aside, like what a fantastic way to integrate AI tools. Like they, the first way that they did their voice assistant was that you pick up the phone and call it or if you’re asking a question, it will call you and it looks like a call coming in on your iPhone. You can like talk to the thing, which is great, but it’s very clever. But it also does this sort of perplexity style kind of readout. But in more Gen Z it has like emojis and like call outs and things all the way down. Yeah, exactly. But it’s like, it’s interesting because it’s, it cites it all. And to your point, Dean, like it’s perfect for a question that’s I need to explain a quick scenario and then I have a question about it, right. So like I have these ducks that are always doing random stuff and so I’m like, I have these, these four ducks of this type and they’re, one of them is doing this like what’s going on? And it will give me a sighted piece of information where if I was doing that on Google, right. I’d have to read through someone’s blog about how much they love their ducks and so on to see if they even have this problem and all that. And I’d be fed ads and you.
Dean Hanson: Know, and you end up trolling forums, you know, and going through and it becomes a nightmare.
George Little: Right, right, Completely. Yeah. Cool. So I have one other question for you, Mike. So I think you have a unique perspective in the sense that you’re in these sort of two different scales of companies, right. Given all of the kind of trajectory that you’ve had since we met 10, 12 years ago, whatever it was, to kind of now in the development that’s happened in AI. What, what’s a piece of advice do you have for someone who wants to get into either of the two types of roles that you found yourself in.
Mike Smith: I believe that AI is not going to take away everyone’s jobs. I think the people who will lose jobs to AI, they’re going to lose. Let me say this, AI is not going to take away jobs or it will small amounts. People who know how to use AI will take away jobs from people who don’t know how to use AI. And I think that’s the thing I’ve been thinking a lot about is I’ve really struggled to weave AI into my workflows. It’s messy. I can send the email or I can draft the email in ChatGPT and then send the email and then extra step always makes things harder. So I think the guidance is figure out how to use these tools even at a very, very basic level. You don’t need to have it generate code and build a startup and then run a server and all these like really complex chatgpt things or LLM things. Have it draft an email, have it write some blog posts, generate an Instagram post. I think the mistake that I’ve made with outsourcing things to team members or VAs is starting with what’s my biggest problem? Like right now I’m planning a wedding. I’m not going to outsource my wedding planning.
George Little: Because comparison.
Mike Smith: But that’s, that’s a big hairy thing that takes 30, 40, 50, 100 hours. It’s unrealistic to take your biggest problem and outsource it to AI, an intern, a va. The lowest hanging fruit is I’m not posting enough on LinkedIn. I can have someone draft LinkedIn post. So find the like the minimal thing that you can experiment with and then expose yourself to as many tools as possible. I’m in the middle of this 100 days of AI course. It’s free, it’s really nice. And what I like is that it’s just, it’s a bite sized email every day. I’m sometimes a entrepreneur in the sense that I buy these courses and I’m like I’m going to study this new technology from 4am to 8am Like I have so many courses I bought and I haven’t followed through because I don’t have four hours in a day.
George Little: Right, right.
Mike Smith: But I have 30 minutes to get a link go. Oh, this is what Claude does. Oh, this is a new tool. Like just getting more reps and exposure to tools I think is important.
Dean Hanson: Yeah, yeah, I think that’s such great advice. I think it’s just, just get started with it. Right. Just do something simple with it. Like, this is what I’ll say to a lot of people who I know outside of, you know, this space is who they talk to me about AI and they don’t really understand it. I’m like, just, you know, go to chat GPT, go to the website, go to Perplexity, ask it questions, see what comes back. And there you go, you’ve started using AI, Then, you know, take it on the next steps. Find out. Find something that you’re struggling with and try and, you know, see if it can help. You never know. It’s just familiarization.
Mike Smith: It’s so uncomfortable to force yourself to do that. Truly.
George Little: Totally. I mean, especially when the thing is like, spitting out something that you weren’t expecting. Like, I found playing with other tools, I’m like, I’m expecting some certain outcome, but I feel like I’m like, I don’t know. It literally is like I’m asking someone else to do this thing for me and then I have to review it. It feels almost unnatural. But in any event, Mike, it has been great having you on the show. Really glad that you were able to come on. For folks that would like to follow you online, where can they. Where can they find you?
Mike Smith: You can find me on LinkedIn. I’m embarrassed and let you know I know my LinkedIn URL, which is LinkedIn.inmasmith1 but you can find me on Twitter MrMikeSmith. You can find me on Instagram smitpicks and I look forward to connecting.
George Little: Cool. Dean, any final, final words for the end?
Dean Hanson: No. No. That was really, really great show. Thanks very much. It was very insightful.
Mike Smith: Thanks for having me, guys.
George Little: Yeah. All right, I’ll read us out. Thanks so much for listening to the or watching the AI Impacts podcast. You can find us on YouTube or wherever you get your podcasts. If you want to send us a note, play with our chatbot, which works sometimes head over to Aiimpacts Show. Otherwise you can follow us on LinkedIn and all the other social platforms. Thanks so much.