Should you company be using AI? (with Bobby Jacobs)

[00:00:00] Welcome to Gone Phishing, a show diving into the cybersecurity threats that surround our highly connected lives. Every human is different. Every person has unique vulnerabilities that expose them to potentially successful social engineering. On this show we'll discuss human vulnerability and how it relates to unique individuals.

[00:00:18] I'm Connor Swalm, CEO of Thin Security, and welcome to Gone Phishing. Everyone, welcome back to another episode of Gone Phishing. I'm your host Connor CEO at Thin and I am joined by Head of Growth at Thread, Bobby Jacobs, Bobby, how are you?

[00:00:35] Doing great Connor, good to be here, excited to go fishing. Excited to go fishing now you've gone fishing as a as a says. Farron, that's right, we've gone fishing, you know, we're prior to spend a lot of time on the high seas so it's appropriate to fish.

[00:00:52] That is right, for folks who have no idea what Bobby's talking about. If you catch him or any of his coworkers at an event, they're usually dressed up in pirate gear and they're very noticeable so it's pretty cool.

[00:01:03] For folks who don't know you, we don't know Thread, could you just give us a little bit of your background and also a little bit about Thread? Sure, yeah absolutely.

[00:01:12] Love helping MSPs utilize new technology. I got to work with Lion Guard for a number of years before I was with Thread, had worked with some hardware distributors and start of accelerators and such it before that.

[00:01:28] Thread is a super fun company like a lot of vendors in the space we come out of a pretty large MSPR cells, our founders, Matt and Mark were head of data and Chief of Staff.

[00:01:41] At the MSP. And so, yeah, we're a service experience platform where all about helping people meet their customers where they live provide great experiences.

[00:01:52] Little newer on the space. I think one thing I'm really excited to maybe not get it ton into today but at least get to collaborate on is were pretty similar and far as age and size and growth and I love talking about what's going on with thin and

[00:02:09] I think we've both been on a bit of a similar ride. So it's cool to be here. I remember seeing you at your first conference with Thread. Yeah, after having a lot of fun. Yeah, data kind of DC. That's right.

[00:02:26] So Thread leverages AI and some capacity and a topic that you brought up here that's super relevant and on the top of everyone's mind. Is sure anyone should you. I think you as a company should you as a company be using AI.

[00:02:43] I dare to say everyone's used to actually be to at this point, right? Like I have really enjoyed seeing people that I thought would never use to actually be to make vacation.

[00:02:52] I tend to raise and like do math and things and I'm like man, like you're using it. And I'm not going to say who these people are to surprise me, but it shows like obviously everyone knows about it is using it.

[00:03:04] I think the question should you be using in your company. That's that's a bigger question mark, right?

[00:03:10] Because it's not like oh am I going to do my vacation or I tend to raise and am I going to give access to data is it kind of change how my employees work is the quality going to be as high or things going to be slower or we have to pay for it.

[00:03:21] Or it's the data hosted right there's so many different pieces that come into it. Now, but I think the short of it is like should you be making an effort to use it in some way or you're going to fall behind.

[00:03:33] If you don't or should you let everyone else kind of bring your head to get to the wall. Well, I don't know what do you think, Connor? Guilty is charged we use it personally actually we have a company.

[00:03:45] It's like the way we leverage AI today could technically be we allow our employees subscription to touch a PTA or open AI. And a lot of us is like we don't tell people to use it.

[00:03:57] I'm just a lot of us naturally do and you mentioned like oh you're doing math with it. I actually did that the other day. I was trying to solve this math problem about like the way you could combine numbers on the lock.

[00:04:10] And I couldn't quite figure out how to check that my math was correct. I had an answer as I don't know if this is right. Literally went to CHF2PT.

[00:04:18] Right me some Python that does x, y and z and calculates all the ways you can combine these numbers in these ways. Two seconds later spit it out put it into my computer and it's like.

[00:04:30] All right I got the right answer yeah I just checked it that would have taken me hours to write on my own from scratch. That's awesome man.

[00:04:37] Yeah I think there's so much low hanging fruit that like a little bit of it is just thinking about what are ways that I could use this like maybe it's not you wouldn't inherently think like write me some Python to do.

[00:04:46] So a math equation maybe you'd say like write a blog post or I think a lot of the more creative things I runically or kind of the lowest hanging fruit for AI right now. But yeah I mean when we talk about like should use it for your business.

[00:04:59] I know Microsoft had a webinar about co pilot yesterday that a bunch of people were talking about and we're on. It's interesting to me that like co pilot is has so much interest and I think it's just because.

[00:05:14] It's hard to know how to use AI in your business and if Microsoft has a solution.

[00:05:19] That's probably an easy way to start like I don't think that people necessarily think co pilots the best one I think it's integrated right and I think the other is that it's like kind of safe because Microsoft's doing it. Yeah I mean that's definitely the belief.

[00:05:36] I think for using it for your business there's two types of productivity there's individuals and then there's like does your tool does your service offering does your product offering leverage it in some capacity.

[00:05:47] And I'll draw upon a ton of experience that Josh my co founder actually has in this because his previous startup that he was working at was a malware analysis startup that leveraged machine learning and this was way before open AI or any kind of the available LLM says like.

[00:06:02] They would get the data sets they would train their own machine learning models they built it all from scratch and he's like this is all the stuff that I learned on what not to do with the last startup.

[00:06:14] And it was all things like the really novel thing about AI isn't necessarily the problems that it can solve today but it's the scale it lets you solve the problems at.

[00:06:25] So an example that is really poignant for us is like email tree caution somebody reports an email and says I think this is fishing what do you do with it it's like well currently what do you do.

[00:06:36] You have a sock team whether that's internal or external or you know it's internal IT folk and they review it and they say thumbs up or thumbs down and they're probably leveraging a whole bunch of tools but like what if we could replace that sock team member or most of that sock team members time with.

[00:06:52] Hey, we've already done all the research leveraged AI to do it look you got 10,000 emails sent to you this month but we did it in one eighth the time.

[00:07:00] Yeah and I think that keys into a really like when we're at the question should I use a I am my business.

[00:07:07] I think it needs to be a really specific use cases initially because with anything if it's like I'm just going to try to do this more like if I give you feedback and I'm like Connor you should speak more clearly.

[00:07:16] Like what does that mean like how do I speak more clearly but if I'm like hey when you say this word you actually don't pronounce the de on it right then it's like okay like now I can say that word differently when I talk and I think it's the same thing with AI if I'm like hey team go like use copilot on a daily basis like okay like should I just ask it a lot of things like.

[00:07:35] Should I just click buttons but if it's like hey let's go use it for internet response and the specific situations or like thread we write your time entries for you you click a button and write the time entry you don't like do any setup there's no like training on data.

[00:07:51] And I think that's another big thing that can be overwhelming is like I want to use these AI tools that give me these awesome ROI's but I have to go.

[00:08:00] Maybe to the point where I'm like coding and building things from scratch and setting up my own Azure Open AI service instances and connecting them with APIs.

[00:08:09] Or maybe it's just like I have to go set up hundreds of rules and that's how the tools help me do it.

[00:08:13] I think is much as you can find tools that have done the back in prompting and configuration for you and just give you that quick ROI even if it's simpler right like we could do.

[00:08:25] More intense things than write time entries for you and we do some other cool things but that was our first release so that's like hey click a button and it's going to solve this really annoying thing that no one likes doing.

[00:08:35] I feel like that's where a lot of people can start using AI and I'm not saying go use thread and write our time entries you know I would love that but whatever that like low hanging fruit is like find the really easy ways to use AI at first.

[00:08:48] That makes sense it's like a. So if you think everything's an a. It's like your or whatever it is is if you think everything's an a. You'll only use a hammer yeah AI is definitely not a.

[00:09:00] We're going to build our business on the back of it I mean you could but it's like you need to have the use case in mind at some one thing I'd say to my team all the time and you probably super similar conversations at threat is like what are we actually the best in the world that.

[00:09:13] It's like can anyone tell me can we go shout it out to the world is like are we the best in the world at X and it's like the super more the more defined you can get on that the easier time you have finding customers because you know who you're looking for you know what they look like.

[00:09:28] You know what size they are you know what kind of words what kind of tools they're using and just gets way easier than you can help we do this.

[00:09:36] Can you do this other things nope if you need that please go over to this other competitor bars because that's not going to be us yeah it's a there's a lot of power to be gained and just digging down and getting more specific.

[00:09:46] Yeah I think applying that concept in a different ways like if you're an MSP and you were asked your team like.

[00:09:53] Hey how do we use a I internally like what are the best ways we use a day and if they don't know right away and you're paying for a I tools. Then you should probably spend some time figuring out what there's already probably give some guidance around like.

[00:10:06] We used to read the right time and truth or we use roots to automate the onboarding of new customers right I think again you have to get back to that like when it takes it comes in we use copilot to tell us more about it.

[00:10:19] Cypic whatever those things are like it's not just that you have the tool and it's not just what the vendor tells you the tool can do. Water we say and that we're using the tool for and actually doing that with it.

[00:10:32] What are some safety implications in your mind and I think it was a Jimmy had someone was talking about this with us like there's two types of AI safety right there's protecting yourself against others who are leveraging AI to attack you in some fashion yeah.

[00:10:46] Then there's the actual safety of the tool that's that's second one the safety of AI itself and what are some of the implications what are people not worried about you think they should be.

[00:10:57] Well I thought you're going to go the other way with it was like what are people worried about how they handle it what are people not worried about and they should be. I mean we could go with the first one too.

[00:11:07] Yeah, you know I think like my teams using AI how do I make sure that they're not giving away company secrets or that they're not putting out a lower quality product right like I think that's a big one like I'm going to go have chat GPT right my emails for me and then it's like well it's actually using the word kindly a lot which makes it feel like I'm using a virtual assistant and I don't actually have like an onshore service desk.

[00:11:34] But because the way chat GPT is using these words that are small difference but if I'm just copy and paste and like that's going to have a big impact I think that's the first thing of like guys you have to keep the quality really high like I think a lot about.

[00:11:49] I read that Jeff basis at Amazon he personally reviewed every single customer service email for like the first two years to make sure that everyone had the right culture and the the right type of response.

[00:12:03] And I think it was like when he felt like someone's at the point where he didn't need to review it anymore than he would stop but he would do it for a really long time.

[00:12:11] And I think that just talks about the importance of word choice and customer experience. And so you don't want to let those things slip you want to make sure you're still providing the really high quality personal product so maybe that's something people are thinking about as much.

[00:12:26] I think the thing that people are thinking about is. Are you training on my data, you know people what was it like Samsung I right away gave away like all the secrets for how they built their hard drives or something like right when chat GPT came out.

[00:12:40] And then it was public knowledge. Other than the real is afraid of right. I just talked for a while, Connor, I'll let you cut in. I think your your first point is actually really interesting because I've never thought about it like that is.

[00:12:57] A phrase that one of my friends, red harnesses used is the only thing automation allows us to do is bad is shitty things faster.

[00:13:04] So it's like if you don't like immediately I interpreted what you were saying as it's like the own AI is leverage if you're leveraging something bad you're going to get way more at the bad but if you're leveraging the good you're going to get way more at the good you just to make sure that that's you know that's what you're actually getting it's like.

[00:13:23] If you read the cliff notes and expect that you've gotten like the whole book read I remember one time up bring it back to something happened in the middle school I didn't read a book right I needed to.

[00:13:32] I just googled summary of the book okay what I know is that there were two books with the same title and I got the summary for the wrong book.

[00:13:40] So I'm writing this book before I hand it in and the teacher was like essentially I was caught red hand it because it was very clear I didn't read the book because. I was literally summarizing this whole wrong thing.

[00:13:54] And it though I don't even remember what happened whatever but I should remember it didn't turn out in your favor.

[00:13:59] I remember learning from that as like wow is like if I'm going to take a shortcut I had probably not take that short of the shortcut and I should just do a little tiny bit.

[00:14:10] That's that hilarious but I don't I don't know if I agree with with your friend statement that automation.

[00:14:18] Let's you do shitty stuff faster like obviously that's an extreme like I get easy to poke holes in statements like that and it's probably mostly true but again going back to our time entries like.

[00:14:29] I we make time entries better like typically time entries are not very good if I'm going to go summarize what I did for the day I'm probably just going to say like how to call the corner did this did that did a demo and I'm going to move on.

[00:14:43] But if I have AI like read my notes and look at my calendar it's probably going to explain a little about how we did a podcast and some of the topics we talked about.

[00:14:52] And so maybe it's just because automation is with large language models and open AI and everything it's getting so much better that that statements a little dated. But I do think it's possible to do better work faster you just have to be very smart about it.

[00:15:09] To give you some more context on that statement because it still applies to this conversation is it was if you're it specifically.

[00:15:16] If you're understanding of cybersecurity starts with vendor education and not a set of first principles of security education you're not you're only going to do the tools that you purchase are only going to allow you to do the wrong thing more often it's not going to get you more secure.

[00:15:31] I make sense I knew that that was too easy to put on there. No, not not.

[00:15:36] No, but that makes a ton of sense and we actually put that quote up me and him did a presentation and it was just like what you guys if you're saying this publicly is like I said wrong like it tell us.

[00:15:48] But yeah, that's that's a really great way to approach AI is like this is leverage is like. This isn't turn your brain off and get free productivity this is now you have new problems to solve which is.

[00:16:01] If we're doing a lot more are we doing a lot more of the right are we doing a lot more of the wrong how can we cut off the wrong do more the right.

[00:16:09] Totally but hey before we wrap up one more just really concrete point there that I would say is like know what it means.

[00:16:16] Who know what language model your vendors are using if it's open AI or whatever that is and then the know what the magic words are to know that they're not training on your data with open AI the vendors should be using Azure open AI service that means that it separated that means that data is not being processed.

[00:16:34] There's different words like that there's different ways that you separate the data from the entrain I think that's the thing that people are asking about but they don't know how to ask the right questions.

[00:16:45] That, you know, that's why people believe prompt engineering is going to be a whole industry is because what really matters.

[00:16:52] Well, especially with the way modern AI works today and I guarantee six months from now will be seeing we probably sing in a different tune that so fast it's changing it's like the quality of question you ask is the quality of answer you'll get.

[00:17:05] 100% awesome for folks who wanted to connect with you or connect with thread to learn a little bit more or just to say hey where would you suggest they do that.

[00:17:13] Yeah, get thread dot com and find me on LinkedIn Bobby Jacobs I've got a pirate hat on or Bobby at get thread dot com. Thanks Connor.

[00:17:22] Awesome folks if you've been listening or watching we will have links to Bobby and we'll have links to thread in the show notes so feel free to reach out and connect.

[00:17:30] Bobby thanks so much for joining me today was a blast chatting as it always is can't wait to see him the next episode.

[00:17:38] As always everyone thanks so much for tuning in to gone fishing if you want to find out more about high quality security awareness training campaigns how to launch them in ways that actually engage employees to change their habits then check us out fin security at finsec.io that's ph i n s e c dot i o or click all of the wonderful links in our show notes.

[00:18:01] Thanks for fishing with me today and we'll see you next time.