UnHacked - Cybersecurity Made Simple for Small Businesses

Justin Shelley (00:00)
Welcome everybody to episode 73 of Unhacked. ⁓ Guys, I say week after week, know, usually in my introductions, but we recorded yesterday. So ⁓ no lies. I'm tired, but here we are. We're going to do back to back episodes. And today we've got a little bit of a barn burner. Get on your technical hats, guys. We're talking to business owners, but we are going to get in the weeds a little bit today. ⁓ But the outcome is what's most important. We are here today to talk about how to protect your business.

Bryan Lachapelle (00:20)
you

Justin Shelley (00:29)
in this crazy exploding world of all things AI. So with that, let's go ahead and start with some introductions. I am Justin Shelley, CEO of Phoenix IT Advisors. I help companies build wealth using technology and then help them protect that wealth from all the nefarious hackers out there, specifically the Russians. Mario comes in with the Chinese. I think Brian covers the rest of them, but between the three of us, we got the whole world covered.

Bryan Lachapelle (00:48)
Ha ha

Justin Shelley (00:54)
⁓ but that's not enough. So we have to go after, know, protecting from government fines and penalties and class action lawsuits and like, you know, over and over, we're just getting, ⁓ we're getting it handed to us from, from all sides. So we're here to help you with that. And I am here as always with my good friends and co-hosts, Mario and Brian in that order, guys, tell everybody who you are, what you do and who you do it for.

Mario Zaki (01:16)
Thank you, Justin. Mario Zaki, CEO of Mastek IT, located in New Jersey, right outside of Manhattan. We've been in business for 21 years now. We help small to medium-sized businesses stay safe from all those hackers that Justin just mentioned. And we specialize in helping business owners sleep better at night, knowing their businesses are safe the next morning.

Justin Shelley (01:41)
Love it, Brian.

Bryan Lachapelle (01:43)
I'm just going to say ditto. No, I'm kidding. All right. My name is Brian Lashford with B4 Networks. We're based out of Niagara area in Ontario, Canada. ⁓ And we help, specifically my team and I help business owners deal with the frustrations and problems that comes with dealing with technology. And we do that by applying all standardized ⁓ systems and processes to make things a lot easier.

Mario Zaki (01:45)
Check.

Justin Shelley (02:09)
Perfect. All right, guys, time to introduce the guests. So guys, I have been practicing and I'm not any better than when I started. So I am going to spell our guests name today. F-A-I-Z G-O-U-R-I. Did I get that part right? Beautiful. 100%. I'm really happy. I've been saying Fies and I think if you pronounce it correctly, it's slightly different from that and gowry or something. And I know I've slaughtered it.

Faiz Gouri (02:24)
Yes, did, Dustin, yeah.

Justin Shelley (02:39)
Guys, go ahead and pronounce your name properly for us.

Faiz Gouri (02:39)
No, sure.

Thanks, Justin. Yeah, you're almost there, actually. So hi, my name is Faz Gauri, and I'm an elite software engineer. ⁓ I have like over 15 years of experience. I work in the back end development in tech. Started my career in India, then came here in USA about 11 years ago, more than a decade ago. Right now, my specialty is like I work with AI and machine learning tools. Also,

My specialty is also to help organizations save money by optimizing their system and also make sure that all the company data is secure.

Justin Shelley (03:22)
I love it. Cause listen, in our world, we have to talk a lot with business owners about why it's important to spend money on technology. And sometimes it is hard to, to directly tie that to increase in revenue or decrease in costs. And so we'll have to pick your brain there. It sounds like you've got some experience and, Fais you left out the company you work for. ⁓ I'll save all the jokes I've heard about it, but at least give us the name.

Faiz Gouri (03:46)
Yeah, yeah, yeah, so I work at Microsoft. I've been here for about seven years.

Justin Shelley (03:53)
Seven years at Microsoft. ⁓ I listen, I'm, kind of a, I I'm a wannabe slash ⁓ wish I were, I guess that's the same thing. Coder developer. That's what got me into technology back when I was a kid is I loved to write code. I dreamed that I would be writing code and building everything new in the world by this point in life. ⁓ It didn't go that way. However, I will say that AI has kind of rekindled that love for me. And I dabble with code now.

Faiz Gouri (03:54)
Yeah, yeah.

Justin Shelley (04:23)
but I don't really write it. just say, Hey, AI do this for me. And then I go look through it and see how cool it is. That's my, my claim with, ⁓ writing code these days. So I love, I love talking to developers. It's like I said, it's a love of mine, ⁓ an unrealized passion. So looking forward to this, that really caught my attention. I'm thrilled to talk.

Faiz Gouri (04:32)
secret you know yeah

Sure, just wanna say like

90 % of coders are doing exactly the same thing. ⁓

Justin Shelley (04:47)
I know, I know, but I wrap myself out.

you know, it is what it is.

Mario Zaki (04:52)
I started

coding in college and in the very beginning I was very into it and then programming just drove me nuts when I put like a comma instead of a period and I stared at it for like three days and I couldn't figure it out. And at the end of the day I'm like, you know what? Fuck this, I can't deal with this shit anymore. And I switched majors just because of a little comma that I had inserted.

Faiz Gouri (05:08)
Yeah

Hahaha!

Justin Shelley (05:14)
⁓

Bryan Lachapelle (05:17)
Yeah.

Justin Shelley (05:19)
So...

Faiz Gouri (05:19)
Yeah, yeah.

Bryan Lachapelle (05:20)
And the irony is now you can take that entire code, set it to AI, and it'd be like, here's your comma.

Justin Shelley (05:25)
Right, right.

Faiz Gouri (05:25)
Hahaha!

Mario Zaki (05:25)
Exactly.

Justin Shelley (05:26)
And my story is the exact opposite. I was actually going to school for electrical engineering and was required to take a computer science, you know, CS 101 class. And my professor pulled me a science like, dude, what the hell are you doing in electrical engineering? You need to switch your major to computer science. So I did. ⁓ That's what we'll save the rest of that story because it doesn't end the way we would all expect. Anyways, Fais, let's talk about, you know what, when we did our little

Bryan Lachapelle (05:35)
Mm, yep.

Faiz Gouri (05:41)
Thanks.

Awesome.

you

Justin Shelley (05:54)
⁓ strategy session we'll call it. ⁓ You mentioned MCP, which Brian correctly identified as a Microsoft Certification, ⁓ Microsoft Certified Partner. Tell us a little bit more about the MCP program and what it is.

Faiz Gouri (05:56)
Okay.

Bryan Lachapelle (06:04)
You

Faiz Gouri (06:09)
It's slightly more popular than Microsoft Certified Professional. So in current norm, know, MCV. Okay, professional, Yeah, so.

Justin Shelley (06:13)
professional not partner. I did that wrong. It was wrong all the way around. So anyways, tell us what MCP

Mario Zaki (06:20)
Yeah.

You got the Microsoft part right?

Justin Shelley (06:21)
really means. Right, right.

Bryan Lachapelle (06:24)
Yeah.

Faiz Gouri (06:24)
Yeah, so it's a, the full form is model context protocol. So it's basically a, you can say as I think Justin, you yourself mentioned, it's kind of USB drive that plugs AI into your own data. So if I have to tell you in a very lemur terms, if you talk to AI, like cloud, Gemini, chat GPT, ⁓ it does not have your information. You can ask it.

anything about the whole world that this AI is trained on. So it knows about everything who is the president of United States, although sometimes it gives you wrong information also. But you can ask it everything, but not something that your personal data, because obviously you didn't provide the data. So what if a small organization or in fact a big organization want to leverage this AI for their

Justin Shelley (07:02)
Yep. Yep.

Faiz Gouri (07:17)
customers, know, for example, let's take an example of hospital, know, a small hospital or a pharmacy. Let's take example of hospital only. ⁓ I had my, you know, annual visit of the doctor's visit and I can see in my chart, I can see all my results, those are like all foreign language to me, you know, what does the hemoglobin 162, 180 mean? Although it gives me range, you know, hey, this is the healthy range and this is your data.

But what if it can speak to me in natural language and I can ask questions like, hey, ⁓ cholesterol is 150 this time. What is the normal range? And can you pull the data for last year also? So right now it doesn't have this information. this is what MCP does. If you plug AI to your own data, in this example, it's a hospital which has my data, my health data.

Then I can ask all these questions on my data in natural language. So this is what MCP does. You don't need to write fancy code or anything. You don't need to create your own chat GPT, basically. So you can use the existing chat GPT or the cloud or the existing AI agent to hook it up to your own database. And then you can ask questions to your own database. So yeah, this is what MCP is.

Bryan Lachapelle (08:39)
⁓

So if I'm understanding you correctly, ⁓ if I ran a ⁓ manufacturing plant and I had all my data in there, could connect it to my data and say, how many cars did we produce last week and how many did we produce a week before? based on the data, how fast were we able to get certain components installed or whatever? Whatever is in the database, can ask it questions in natural language and it will extract the data and do all the calculations and everything for you.

Mario Zaki (08:41)
But correct me.

Faiz Gouri (08:47)
Mm-hmm.

Correct. Absolutely. Yes.

Exactly. Yes.

Exactly. You don't

need to be a database expert or you don't need to be like API expert to, know, ask the question to your data. You know, when a CEO or person like HR professional can just go and ask, you know, those questions what you mentioned, how many cars did we sell? What are the profit and all? So yeah.

Bryan Lachapelle (09:21)
Yeah.

Right, how many people did we hire last month or last year? And how many of those people made it through?

Faiz Gouri (09:26)
Yeah, maybe in this time, I think the more appropriate

question would be how many people did we fire last year?

Bryan Lachapelle (09:34)
Yeah. Yeah.

Justin Shelley (09:35)
Jesus, no kidding,

Mario Zaki (09:35)
He

Justin Shelley (09:36)
right? God.

Mario Zaki (09:39)
But now, see, in Brian's example, he has access to the data right away. In your example, you would require to download your data and be constantly downloading it, you know, after every exam or something like that. Wouldn't it be better if the, like, the hospital, you know, portal has this built in, you just log in and they kind of have their own ⁓ platform that says, hey,

Faiz Gouri (09:42)
Mm-hmm.

Mm-hmm.

Correct. So. ⁓

Mario Zaki (10:09)
You know, what was my, you know, a one C the last, you know, couple tests, am I getting better? What should I do differently and stuff like that.

Faiz Gouri (10:17)
Yeah, no, that's a good question. So it has to be built on the server side. It has to be built by the data owner. So data owner is the hospital. I'm just a user. So hospital has to build this functionality. They need to build the MCP server on their own hospital database, all the database of all the patients, you know. And then, you know, a user or a patient can use this data and leverage AI to ask questions and get some health tips and some recommendations.

So yeah, I'm just a user. I'm not building this thing as a patient, right? Just a hospital, you know, people who own the data.

Justin Shelley (10:54)
So let's talk about security, cause that's, you know, kind of what we're here for. ⁓ this is, this is what comes up over and over where AI is concerned as you start feeding it information, it ingests that, and then it spits it out to the whole world. ⁓ what talk about guardrails, talk about how MCP keeps the data contained so that only the right people can access it.

Faiz Gouri (10:56)
Mm-hmm. Yeah. Yeah.

Yeah, so MCP usually ⁓ keep your data secure. ⁓ most like ⁓ just going to be a technical, but you the best way to use MCP is through APIs, you know, on your data, you know, you have APIs to get, hey, for example, in a branch example, hey, there are some API in your car manufacturing company that returns the number of cars sold, you you can add, hey, get cars sold or something last month, you know, so you pass some parameters.

But MCP connect those API and then fetch this data and return in the many very meaningful, you know, natural language. Hey, we saw like 50 cars last year or last month. So there are a bunch of, you know, ways that we can kind of, you know, or the hackers kind of, you know, infiltrate and misuse this information. One of the most common is like kind of prompt engineering, you know, you must have been problem. So if they pass some...

malicious prompt, for example. Okay, so one more thing, MCP not just return your data or ask, you can ask question, through MCP, you can create AI agent that you can work or do a task also, let AI do the task on your behalf. It's basically like updating your database also. For example, as a car manufacturing part, let's say the... ⁓

the showroom owner sitting there, they can say, sold 50 cars, you put insert into our database? Things like that, just update database also, it can do certain tasks also. So the malicious prompt engineering, through a hacker through malicious prompt, they can ask, hey, can you delete fast data from database or can you update fast hemoglobin level to 200?

Something like this, we have like MCP has to be, you usually MCP is smart enough to tackle all this time, but we have to be very careful about, you know, who we are authorizing to do what. So authorization is very important in MCP. Like it must have some, again, some technical term, but like role-based access. So like we have to decide roles, like who can perform certain type of actions, like manipulating the data.

⁓ writing to the database or things like that. ⁓ And who can access what data? For example, especially in terms of when we are working with very sensitive data, like patient data in a hospital, who can access FAS data? It has to be either the provider who's seeing me or just the patient himself or herself. So yeah, those are the things we have to ⁓ manage through authorization, have proper... ⁓

Azure-based authorization or any authorization the MCP can use. And also like rule-based access, like who can perform what. So one of the, I say, Yes, let's have the MCP service or the server level, yeah. Have to provide all this information, yeah. Make sure that, know, like prompt is one of the very common example where it can get very serious. you can...

Bryan Lachapelle (14:19)
And that's done at the MCP level.

Faiz Gouri (14:34)
you can ask is hey can you delete this data or maybe you know manipulate something you know can you set the price of the car sold to 2000 rather than maybe 50,000 so yeah have to be very careful about those security concerns

Justin Shelley (14:52)
that that's handled in the setup when you when you connect these all this stuff. Does it does it also prevent the AI engine, the LLM or whatever to from training on and having that become part of the model and future releases? Because I know that's something that comes up quite a bit as well.

Bryan Lachapelle (14:52)
Fair enough.

Faiz Gouri (15:09)
That's very good question. Yes. So MCP has been very ⁓ vocal about it. It's not going to use your data to train its own AI model. Yeah. It's clear guidance over there. So unless they are lying. yeah.

Justin Shelley (15:25)
Unless they're lying. Yeah, we who knows? ⁓

Bryan Lachapelle (15:27)
Who does that? Nobody's ever done that

before.

Mario Zaki (15:32)
So, I mean, the thing is, I feel like at some point there's going to be a vulnerability somewhere. You know, like we see it everywhere, you know, there's going to be a vulnerability somewhere and this is ⁓ if or when that happens, what is the end result?

Faiz Gouri (15:55)
So if some security breach happens, that's you're asking, right? You know, through MCP. As I mentioned, I think most of the very known issues are very well taken care of the MCP by itself. But also when you enable MCP to your own data, you need to take certain precautions to make sure that doesn't happen. But if it does, then ⁓ it's again the same thing as, you know, somebody got access to your database and,

delete something. So it's just like that. It has nothing special with AI, but something similar to when hackers are very smart. When they can access the US government data, they can access everything. So ⁓ it just that you have to be precautious. Another approach is keep logging everything. how it helps is if you have logs or traces or everything that's going on in your system.

you can monitor those things and then, ⁓ again, using AI, I don't know to use it, using AI to kind of, ⁓ in the real time, scan these ⁓ logs and then proactively alert you if something fishy happens. So that's another approach.

Justin Shelley (17:11)
Okay. So now this is, warned everybody about this, but now we've got to take this down out of the technical weeds and, ⁓ speak in terms of, know, just talk to me, a business owner of any random industry. And what I just heard is a whole bunch of shit that I don't need to know about, right? Like, so why, why does it matter to me that, that this exists? ⁓ and, and just say, let's just say I have a, I don't know. I'm a vet. I'm a veterinarian.

Bryan Lachapelle (17:28)
Yeah

Justin Shelley (17:40)
Right. I've got this little vet clinic. there's 10 people that work for me. Is there any practical application for this in my world?

Faiz Gouri (17:49)
For sure, yeah, it's for everyone. So let's take the veterinary example. So if you have the veterinary clinic, ⁓ you can have, you definitely will have the data of, you know, all the pets that, you know, that comes to you, all the, I would call them maybe patient or all the pets, but ⁓ let's say the users, you know, the one of the very good example for business owner is that their customer can talk to some, it's actually AI, not a person, but you know, they can,

talk to the system in natural language without worrying about ⁓ all the system things and all the stuff. They can just talk to natural language. They can get all the technical answers or all these jargons that they may use. For example, for the vet clinic, if I own a cat, ⁓ I can see all the history of the vaccination for my cat. And I can ask in the natural language, when the net

next vaccination is due without you know waiting for the clinic to open, veterinary clinic to open without you know like so it works 24 7 basically you know I don't need to wait for their call.

Justin Shelley (18:56)
So how would I,

again, not being a technical individual, I want this in my practice. It sounds great. What's my first step? How do I even go about setting this up?

Faiz Gouri (19:02)
Hmm.

I have to be little technical to say this actually. first I would be maybe hire a new, yes, hire a consultant. Yes, first I would be the hire a consultant.

Justin Shelley (19:11)
So I need a consultant.

Okay. Okay. ⁓

tell me, tell me this, ⁓ is, were you involved in the development of this protocol?

Faiz Gouri (19:25)
no i was not i wish i were yeah

Justin Shelley (19:27)
Okay,

just something that you're very familiar with and see the value in it. Okay.

Faiz Gouri (19:29)
Yes. Yes.

Justin Shelley (19:34)
Yeah, interesting. I tell you this, ⁓ I have always loved the world of technology because it constantly changes and evolves. ⁓ but holy hell has that. I mean, it, it almost stagnated for a while, right? It used to be, waiting for the next processor to come out. my God. The I five just came out. We're so excited. Or back in the day, we go back farther. The MMX, you guys remember the MMX processors, ⁓ the MMX technology, I guess it was just like, ⁓

Bryan Lachapelle (19:58)
my God.

Justin Shelley (20:02)
We're always waiting for new developments. And, you know, I got to a point where I'm just like, you know, at what point is it enough? And this was before it became enough. It was like, we've gotten to where the computers do everything we need them to do. ⁓ and then sadly that prediction became true. And I honestly started getting a little bit bored with technology because it was just the same thing. ⁓ good Lord has AI. ⁓ my God. It has rekindled all of that. Both.

Mario Zaki (20:24)
Now you can't even keep up?

Justin Shelley (20:29)
So this is, this is another example. And honestly, I, maybe I'm correct me if I'm oversimplifying this, but it seems like this is another approach to what a lot of people are trying to solve. Is that, is that fair?

Faiz Gouri (20:47)
not solve but actually leverage. So MCP solved a lot of problems that came with AI. So when like chat GPT or the LLM were introduced like maybe four years ago, one of the major challenges for the organization was how they can use this tool into their own system. It knows everything about the world, but what if like, for example, like veterinary clinic, how I can use the same tool so that my users can...

Justin Shelley (20:51)
Okay.

Right, yeah.

Faiz Gouri (21:17)
chat in natural language or even, know, now the AI is so smart, you can speak to it, you know, and then you can get the very natural language answer, you know, through speaking also, typing your answers rather than they need to worry about all the, you know, for example, hospital in my chart, I need to understand what the, you know, cholesterol level, healthy cholesterol level is, what's unhealthy, what is the history of my cholesterol level, you know, five years down the line, can you see the trend analysis? All the questions you can ask through MCP, you know.

because I can say, you draw the chart of my cholesterol level, my weight gain, and you can kind of relate this data. It's smart enough to do that. Can you relate all my health data in the last five years and provide me some recommendation? So it's smart enough to do that. It will provide me some recommendation. Hey, you gained some 50 pounds of weight. Maybe you need to cool down on your red meat. So it's very smart enough to do that. So yeah.

Justin Shelley (22:03)
Gotcha.

Hey now, hey, let's

not talk like that. We're keeping this show G rated. Keep that down. Good Lord.

Bryan Lachapelle (22:14)
Whoa, whoa, ho ho ho.

Mario Zaki (22:17)
Hehehehehe

Faiz Gouri (22:17)
you ⁓

Bryan Lachapelle (22:20)
Yeah.

Faiz Gouri (22:21)
You

Bryan Lachapelle (22:23)
That's like a swear word around here.

Faiz Gouri (22:24)
Yeah ⁓

Justin Shelley (22:24)
⁓

Mario Zaki (22:25)
Hahaha

Justin Shelley (22:25)
it's worse than a swear word. I like swear words, but I won't talk like that. ⁓ well, listen, this is, ⁓ I'll be honest. ⁓ I didn't know anything about this before you and I first chatted and I am now wildly intrigued. I won't lie. I've been kind of scrolling through the website as we talked and I'm seeing all kinds of code that I can nerd out on. and, and I am looking for ways that I can take this back to my clients and make their worlds better. So.

Bryan Lachapelle (22:29)
You

Justin Shelley (22:55)
⁓ I'm, I'm pretty excited about this. Honestly, guys, want to, I want to ask your take on it as well. Mario, Brian, think this is relatively new for you as well. ⁓ talk about it, Mario, you go first and then Brian, what, ⁓ what do think about all this?

Mario Zaki (23:10)
I mean, just like everything AI, this is a game changer. It really elevates the ability that you're able to do with information. It even provides you with stuff that you don't even know that was there. You may end up looking at something and never even realize that something is pretty concerning, or even seeing a trend, something rising that maybe not being

Faiz Gouri (23:15)
it is.

Mario Zaki (23:38)
layer, you know, like at a major point, you know, staying in like the medical field. ⁓ I'm sure AI will kind of bring it to your attention. You're like, hey, by the way, your last couple tests, you know, I've noticed this. ⁓ You may want to get further testing done or, you know, other examples like, hey, we're seeing a trend.

in your database, you you're manufacturing, hey, you know, this car is selling a lot better than the other ones. You know, you're losing money on this. So it's definitely going to point out stuff that nobody may have, you know, brought it, thought about it for a while. So you can, you can be on top of stuff a lot sooner.

Faiz Gouri (24:22)
That's a point, yeah.

Bryan Lachapelle (24:24)
Well, it sounds to me like this is a bridge between data and AI and being able to access your data. So I'll give you an example. I have a lot of metrics in my business. I track a lot of metrics in my business, but it's all gotta be programmed by a programmer. Somebody has got to develop an integration with our database, figure out how to extract that data and produce a report and or gauges on a wall, which we have a lot of those. But it sounds to me, if I'm correctly understanding this,

If I connected my data, I can then just ask questions of the data, of the AI, and it will go fetch the information out of the database and do the natural language processing to figure out what it is I'm asking, figure out how to extract that data, and then present it in a way that makes sense. And then using the rest of the AI model, allow me to analyze that data and do things with it. So it sounds to me like instead of me going, hey,

Faiz Gouri (25:17)
This room here.

Bryan Lachapelle (25:20)
know, Brandon, who's one of my programmers, developed me a, ⁓ I need this data. I need you to create me ⁓ a gauge that shows me this data. I can just connect the AI through MCP into a database and say to the AI, hey, I'm curious how many tickets I got last month. And it would be like, you got this many. how many of those were alerts and how many of those were client initiated? And it would be like, well, here you go. Here's the information. Without having to have a programmer in between programming it out. So as a business owner, yeah.

Faiz Gouri (25:43)
Yeah.

Mario Zaki (25:47)
So what are you gonna do with Brandon?

Faiz Gouri (25:49)
Yeah.

Bryan Lachapelle (25:51)
As a business owner, it sounds to me like I could use this internally to start analyzing data in my business. But more importantly, if I have customers, I can then expose this to my customers and my customers can now go to an AI and also ask it how many tickets did I specifically put in for my company? And it would go to my database live, grab the information and present it to them in a way that makes sense for them. So I guess from my perspective, like if I were a business,

MCP is the ability for me to get data either for myself and or for my clients. ⁓ And in a way that doesn't require me to have to constantly be changing the ⁓ reports and how to extract that data. The AI would figure it out on its own.

Faiz Gouri (26:36)
Absolutely, yeah. And I'll take the Mario's analogy, know, like ⁓ for the car owners, know, or the car showroom owner, they can ask all the questions in natural language, how many cars did we sell? And AI is very smart enough to give you the pattern also, hey, which model did we sell the most? ⁓ And you can, hey, can you give me ⁓ why those, you know, why could those cars not sold?

it has all the data about your cars. You hey, maybe those cars would not, for example, let's say Tesla's example, know, like, we sold most of the Model Y because, know, it's ⁓ it's SUV, it's within the range of, know, like it's very affordable car. It's, you know, ⁓ it qualified for the EV tax credit also. Why didn't we sell Model X? Because it's the same thing as Model Y, just a little bit better, but you know, it's like double the price. It also doesn't qualify for the EV tax credit. So things like that, you know.

So it can give you all the answer, all the analytics also. You can use all the analytics through natural language without hiring a special person doing the stuff, which will take a person to do it in maybe a week. You can do it in seconds.

Bryan Lachapelle (27:46)
Yeah, I mean, probably even start analyzing trending data like, you know, what repairs do I have to make to my cars? And when were they mostly manufactured? If all the repairs are because somebody is on a Friday and they're manufactured on a Friday, now we've got a Friday problem, So interesting. So I think there's a lot of applications ⁓ from a perspective of a business owner. If you're trying to get data out to your customers, this is a good way to do it. And if you're trying to get data for yourself to be able to analyze different trends and information in your data, ⁓

Faiz Gouri (27:55)
Okay.

Absolutely. Yep.

Bryan Lachapelle (28:16)
Holy crap, I mean, this stuff is invaluable. How do we get started?

Faiz Gouri (28:18)
step.

Justin Shelley (28:18)
So talk to me,

talk, yeah, exactly. So I want to talk about limitations, if any. And I do a lot in healthcare. So let me, let's say there's, you know, you've got web-based applications, you've got server-based, some of them run SQL server, but let, for example, Athena, we've got Metinformatics, we've got Dentrix. Are all of these products accessible? Or are there limitations on what we can connect to and what we cannot connect to?

Faiz Gouri (28:30)
Mm-hmm. Mm-hmm.

So the responsibility is on the data owner. What data you want to connect to the AI. basically, in the beginning, you mentioned a very good analogy about the USB drive. So MCP is just the USB drive on your data. So ⁓ you can hook it up with any data you want, but you have to be very careful which data you want to hook it up to. And also, need to make sure that you create some rule-based access. So for the health care provider, you know, ⁓

Justin Shelley (28:50)
Okay.

Faiz Gouri (29:13)
healthcare industry, they need to make sure that no person can use this data, you know, which is he or she is not authorized to. For example, I as a patient can only access my data, not any other data. Yeah.

Justin Shelley (29:26)
Right.

So I, and I get that. but I'm asking like, can I, are there programs or other applications or other data sets out there? Because I, you say the owner of it, but really who owns it? It's my information that I put in, but it's somebody else's system. How do I connect to their system? If I want to go ahead and use my data that lives in their world.

Faiz Gouri (29:30)
Yep.

Mm-hmm.

So generally, yeah, as a user, can't do that. It has to be done by the data owner or the system. Yeah. Yeah. Yeah.

Justin Shelley (29:57)
through a consultant or whatever is it, we, and we'll talk a little technical now. Are we talking,

we just, does it just tie in through API? Do you have to have direct connection and access to a SQL database? ⁓ Where are, where in the world does it not penetrate? Does that make any sense? how?

Mario Zaki (30:14)
I think he said it needs an API. So the API.

Faiz Gouri (30:17)
Yes.

Justin Shelley (30:17)
Is everything through API?

Faiz Gouri (30:19)
Yes, generally through APIs, yes. Most of the time, yes, through API. So if you have access to the APIs, yes, yeah.

Justin Shelley (30:22)
Okay, good. Which, which most things have.

Yeah. Most products have some sort of an API. ⁓ will it, will it, is there a way to interface directly with a locally hosted SQL server, for example?

Faiz Gouri (30:36)
No, no. If you don't have access to the locally host, like those server, yeah, no.

Yeah, the access level is streaming same, yeah.

Justin Shelley (30:45)
Well, I've got some, ⁓

I've got some homework. That's all I know. Cause I've got questions and I can go on forever. It will get technical. ⁓ I think I'm going to be hooking AI up to this website, which by the way is called model context protocol.io. And I'm going to have AI teach me everything it knows, ⁓ or can learn about this product.

Bryan Lachapelle (30:50)
Yeah.

Faiz Gouri (31:05)
So the biggest actually

challenge that it solves is, ⁓ as I mentioned, four or five years ago when the chat GPT was introduced, so did the other like LLM tools like Gemini and Cloud from other companies. The biggest problem was I as a user, I can upload certain files and documents. Let's say I have my own PDF or some Word document. I can upload it there and I can ask question on my data. But there was a limit, you mentioned, you remember if there's a token limit.

Justin Shelley (31:25)
Right.

Faiz Gouri (31:34)
You can only use 2000 tokens or something like that, right? So you cannot upload, let's say, one terabyte of SQL Server and then ask question to the chat GPT. So this is what MCP does. Now you can connect to terabytes of data to it, and then you can ask questions in natural language. That's the biggest gap. know, it fills. Yeah. In a very simple way, there were options to do that earlier. Like you have to train AI on your own data, which will take like...

Justin Shelley (31:52)
Okay. I mean that.

Yeah.

Faiz Gouri (32:03)
some time to build the system but you know now it's like plug and play

Bryan Lachapelle (32:09)
and dynamic because the database is changing live and so, Excellent.

Faiz Gouri (32:10)
End dynamic. Yes, exactly. Yeah.

Justin Shelley (32:11)
Yeah.

you just ruined my weekend. Thank you for that. Cause, now I'm going to be sitting here playing with this all weekend. ⁓ I guys, I, I have more than enough to blow my brain. So I'm going to, we're going to kind of wrap up. go around the room and you know, any final thoughts, key takeaways, anything else you want to say, this is your chance. And then, we're going to wrap up so I can go play. ⁓ Brian, we'll start with you and then Mario and then Fais will go ahead and have you bring it home for us.

Faiz Gouri (32:17)
Hahaha

Bryan Lachapelle (32:44)
Sure, I guess from my perspective, what I just wanted to mention is that this is ⁓ like the purpose of this webinar or this podcast was more to introduce clients to the possibility or business owners to the possibility of how to access your data in different ways. MCP gives us the ability to ⁓ obviously with the help of a consultant, ⁓ put in place a system that allows you to integrate your database with an AI so that you or your clients can access that data. And so...

reach out to one of us and maybe we'll have the answers by the time you reach out to us on how we can integrate your data with AI.

Justin Shelley (33:19)
you

Yeah, I'm gonna have those answers by Monday. Mario, what do you got?

Faiz Gouri (33:26)
.

Bryan Lachapelle (33:26)
Yeah.

Mario Zaki (33:27)
I mean,

there is a lot to uncover in this one. I know we tried to really not go too technical. can, for the average, besides using a consultant, where does the average person start with? Where is ⁓ there a direction that they see what they can do? Or is there certain people that specialize in doing this?

Faiz Gouri (33:49)
I think average person, you know.

I think interestingly, average person directly go to AI and ask this question, hey, I have this system, I want to use MCP, what are the steps? I'm a non-technical person, give me steps in the Lehman terms, what step do I need to follow to allow ⁓ using MCP on my data? I own this industry, how do I proceed with it? And it will give you some very high level steps for sure.

Justin Shelley (34:02)
Yeah, right.

I mean, that's becoming the answer these days, right? Go ask AI, it'll tell you.

Faiz Gouri (34:24)
Yeah.

Mario Zaki (34:25)
Yeah, exactly.

Bryan Lachapelle (34:26)
AI,

how do I use you?

Justin Shelley (34:31)
All right, Fais, what else? What's your final sign-off, final information for us today?

Faiz Gouri (34:38)
Final information is again, it's very interesting and it bridges the gap between your data and the AI. It's a very useful tool, definitely solves ton of problems, which human would take maybe months or years to solve. But the only thing is that you have to be very precautious about the security. If you own the sensitive data, make sure that it remains secure and no bad person

can use AI to kind of access the data that they should not.

Justin Shelley (35:13)
Yeah. It's so, that is the terrifying part about the world we're in right now. said things that kind of stabilized and honestly, the world of security was starting to get a little bit, I don't know if boring is the right answer, but it's the same thing. You know, we did this podcast for a year where we were breaking down breaches. We're talking about headlines and, and it, to me, it felt like we're just repeating the same stuff. And now it's like, my God, there's a whole new world we have to figure out and fast.

Faiz Gouri (35:22)
Yeah.

Yes.

Yeah.

Justin Shelley (35:41)
Because this stuff is, it's, moving fast and unfortunately it's moving faster than we are securing it. So, ⁓ that, that I guess would be my key takeaway today is there is some really cool shit out there guys go use it, go play with it, but please, please, please, ⁓ cover your bases, get, get somebody to look over your shoulder and make sure this stuff is being done the right way and being done securely. because the more we open this stuff up, the, know, we talk about the.

Mario Zaki (35:47)
Yeah.

Justin Shelley (36:11)
threat surface or what was that? No, I'm thinking on that, but you know, the, it, just, the target on our backs just keeps getting bigger and bigger and bigger. And damn, this is a security is just becoming more and more important. So, ⁓ that's what I've got. Cover your bases always, but it's a fun, it's a crazy world. Go out and figure out how to make more money because that is what technology is doing for us now. And it got stale and it wasn't doing that for a long time, or at least it was doing it the same way.

guys, we've, we've given our final thoughts, key takeaways. It's time. you know, we're going to wrap up for this week. We're going to say a couple of short goodbyes and then, and then we'll come back. You know, we got the us Americans anyways, have a Thanksgiving holiday coming up next week. So we won't be recording next week, but we'll be releasing one cause we recorded twice this week. Anyways, unhack my business.com for today's show notes, guest links, free resources, ⁓ or just.

while you're looking at your phone, listen to the podcast, go flip over the details and you're going to have all the links to, ⁓ this cool MCP protocol that we're talking about. We'll have fires. think you wanted your LinkedIn profile on there. Is that accurate? How would you like people to contact you if they'd like to LinkedIn? Okay. So we'll put LinkedIn on there and we'll put Brian and, Mario as always. So, ⁓ go ahead and reach out to any one of us. If you have any further questions, with that final goodbyes, Brian, what do you got?

Faiz Gouri (37:20)
Yeah, LinkedIn is the most accurate way.

Bryan Lachapelle (37:37)
Well, if you're a business owner and you're looking for someone to help you through your journey on everything from AI to security to just improving your business using technology, reach out to us and we'll give you a hand.

Justin Shelley (37:46)
All right, Mario.

Mario Zaki (37:49)
Well, if you can't sleep at night because you're worried about your business, because you think AI is going to take over, give us a call and we'll help you out.

Justin Shelley (37:55)
It already did.

And Fais, what's your final say goodbye?

Faiz Gouri (38:02)
Yeah, if you have business order, please reach out to these guys. And if you have more technical questions, just connect me on LinkedIn.

Justin Shelley (38:10)
Now you got access to a senior level Microsoft engineer. You're welcome. ⁓ guys, and I am Justin. Remember listening, take action and keep your businesses on act. We'll see you next time.

Mario Zaki (38:14)
You

guys.

Bryan Lachapelle (38:21)
Cheers.