Boosting Small Business Intelligence with AI and Technology, with Matthew Krieger, Episode 25

1
00:00:06,589.999 --> 00:00:23,810
Hello, and welcome to Small Business Intelligence by Community Investment Corporation, the podcast where we analyze the latest news headlines and address the all-important question of what does it all mean to you as a small business owner? We'll share our insights in this regard so you can be better prepared for what might happen in the days, weeks, and months ahead.
2
00:00:24,240 --> 00:00:32,139.999
I'm Joe McCaffrey, a business advisor with CIC, and my co-host for this podcast is Louis Silva, the president and CEO of the Community Investment Corporation.
3
00:00:32,150 --> 00:00:33,110.001
It's good to see you, Louis.
4
00:00:33,230.002 --> 00:00:34,39.999
It's been a while, Joe.
5
00:00:34,169.997 --> 00:00:35,140
We missed you last time.
6
00:00:35,220 --> 00:00:35,859.997
I know it.
7
00:00:35,889.998 --> 00:00:39,839.997
I h- but I heard the podcast while I was away across the pond, and it was great.
8
00:00:39,840.003 --> 00:00:40,420.003
So good job.
9
00:00:40,870.001 --> 00:00:43,189.996
Clearly this thing runs without me, so this is awesome.
10
00:00:43,239.996 --> 00:00:43,929.997
No, it's good.
11
00:00:44,660.002 --> 00:00:45,010
We, uh...
12
00:00:45,400 --> 00:00:48,539.999
So who else we have with us in the studio today? Luke's here, as always.
13
00:00:48,590.003 --> 00:00:49,10.001
Hello, everyone.
14
00:00:49,10.001 --> 00:00:51,39.997
And, um, and we have a special guest.
15
00:00:51,39.999 --> 00:01:00,620
We have Matthew Kreger, a fellow SCORE mentor, CIC educational consultant, and a personal friend of mine is here to talk about small business intelligence, including artificial intelligence.
16
00:01:00,989.996 --> 00:01:03,159.997
Um, small business technology, sorry.
17
00:01:03,190.001 --> 00:01:04,880.003
And, uh, Matt's joining us actually on the phone.
18
00:01:04,880.003 --> 00:01:05,489.997
So welcome, Matt.
19
00:01:05,489.997 --> 00:01:07,540.002
How are you? Hey, Joe.
20
00:01:07,540.002 --> 00:01:08,139.996
Thanks a lot.
21
00:01:08,139.998 --> 00:01:08,779.999
Good to be here.
22
00:01:09,080 --> 00:01:10,639.998
So let me tell you a little bit more about Matt.
23
00:01:10,650 --> 00:01:15,620.001
Matt is the CEO of CoBra Technologies, which is a Connecticut-based manufacturing company.
24
00:01:15,999.999 --> 00:01:18,750.002
Matt has spent his entire career in tech-related roles.
25
00:01:19,29.997 --> 00:01:23,610.002
He's one of my go-to people when it comes to tech-related questions, particularly those pertaining to AI.
26
00:01:23,650 --> 00:01:26,809.998
I've been looking forward to today's episode, so let's get right into it.
27
00:01:26,819.998 --> 00:01:33,659.996
So in pr- preparation for today's discussion, I was reviewing the news articles from the last few months, and there's a lot of stuff we can talk about.
28
00:01:34,59.996 --> 00:01:40,779.992
Our clients and listeners have varying degrees of tech savviness, and they're in different stages of their entrepreneurial s- uh, journey.
29
00:01:40,779.997 --> 00:01:48,329.993
So we're gonna start off with some simple, general, hypothetical infor- informational technology questions, and then we'll, then we'll get to the, to the good stuff, to the heavy stuff.
30
00:01:48,330 --> 00:01:50,70.004
So, um, okay.
31
00:01:50,70.004 --> 00:01:52,269.995
So let's, uh, let's go to our, our first question.
32
00:01:52,270.003 --> 00:02:13,762.001
So Matt, if you were in a situation where you were starting a new business today, uh, what sort of technology equipment, hardware, software would you need to get your business going? I think all the same stuff that you have as an individual who isn't running a business, plus maybe a little bit more, right? So you're gonna have your mobile phone.
33
00:02:13,902.001 --> 00:02:16,632.001
In fact, many people start the business with a phone only.
34
00:02:17,492.001 --> 00:02:24,782.001
But you're gonna have your phone, you're gonna have some laptop, you're obviously gonna have some home networking equipment, assuming that you work at home.
35
00:02:25,992 --> 00:02:47,872.001
If you have to work with customers or do calls, you'll have a free Zoom account, you'll have a camera and a mic, and hopefully you'll have some basic accounting software like QuickBooks, and you're probably using some combination of Microsoft Office or Google Sheets.
36
00:02:48,72.002 --> 00:02:52,202.001
And nowadays you'll probably have one or more AI subscriptions.
37
00:02:52,791.999 --> 00:02:53,1.999
Yeah.
38
00:02:53,1.999 --> 00:02:54,351.998
That, that seems to be...
39
00:02:54,392.001 --> 00:02:57,72.002
These are, you know, are the essentials that every business needs.
40
00:02:57,72.002 --> 00:02:59,781.998
Doesn't matter, you know, what, uh, what, what stage you're at.
41
00:02:59,782.001 --> 00:03:00,501.999
These are the, you know...
42
00:03:00,591.998 --> 00:03:08,072
When you think about it, you don't r- really need a whole lot to get a, a new business started, and because of technology, the barriers to entry in many industries are, are pretty minimal.
43
00:03:08,072 --> 00:03:10,192.003
So, uh, okay, second question.
44
00:03:10,192.003 --> 00:03:18,11.997
We're- we now live, we've been living in the information age, and much of that information exists in digital format, which means it can be analyzed.
45
00:03:18,21.999 --> 00:03:24,141.996
For small businesses, where would you find this type of information? Like, you know, you mentioned, like, your, your accounting system.
46
00:03:24,151.996 --> 00:03:27,792
What, what else? Like, and, and what can you do with it when you find it? So let's talk about that.
47
00:03:29,628.001 --> 00:03:32,458.001
I mean, it's gonna sit in your, uh...
48
00:03:32,468.001 --> 00:03:37,278.001
I think, I think if we're honest about it, I think much of our business information sits in our email.
49
00:03:38,128.001 --> 00:03:38,458.001
Okay, good.
50
00:03:38,468.001 --> 00:03:51,808
Um, it's going to sit in, uh, random spreadsheets and documents, and if you have a CRM system, a customer relationship management system.
51
00:03:52,758.001 --> 00:03:55,498.001
It'll sit in a point-of-sale system like Toast.
52
00:03:56,218 --> 00:03:58,378
It'll sit in your QuickBooks.
53
00:03:59,718 --> 00:04:04,698.001
If you use a note-taking tool like Apple Notes or Evernote, it'll sit there.
54
00:04:05,158.001 --> 00:04:15,957.999
And that challenge, by the way, of having your data in multiple places is a challenge that's been around as long as I've been doing things.
55
00:04:15,958 --> 00:04:28,748.001
The advice I would offer is when you're setting up a new business, set up as few systems as you can to do the job and use those systems for what they're good for.
56
00:04:28,958 --> 00:04:42,207.999
So when you start a business and you use email, it's great to use email for communications, but your customer data and your customer contact information should be in some relevant system.
57
00:04:43,7.999 --> 00:04:53,127.999
Whether it's a project tracking system or a CRM system, it's better, in my opinion, to put the data into fewer places with more intention for each of them.
58
00:04:53,128.002 --> 00:05:00,357.998
And then later on, we can talk about how to kinda get a single picture of that data, you know, across, across systems.
59
00:05:00,357.998 --> 00:05:07,687.998
But you will, by definition, have your data in multiple places every time you pick a new software tool to use.
60
00:05:08,67.997 --> 00:05:09,898.002
That's another spot your data is living.
61
00:05:10,687.999 --> 00:05:10,907.999
Yeah.
62
00:05:10,908.001 --> 00:05:13,417.997
One of my favorite is the, uh, the POS system.
63
00:05:13,417.997 --> 00:05:22,17.996
There's really, really good stuff in there that, uh, you know, is, is one of those, those systems that you can really look at the data, and it really tells you a lot about your business, your performance.
64
00:05:22,417.995 --> 00:05:23,597.996
Uh, you know, it's got a really...
65
00:05:23,618 --> 00:05:24,857.999
lot, lot of good stuff in there.
66
00:05:24,857.999 --> 00:05:32,558
So, uh, whether it's Toast or there's many, many systems out there, but they all have really good analytical, uh, capabilities and stuff like that.
67
00:05:32,597.996 --> 00:05:39,128.003
And so, uh, you mentioned spreadsheets and, you know, we us- we use electronic spreadsheets to analyze quantitative data.
68
00:05:39,128.003 --> 00:05:44,608.002
And now that we have artificial intelligence, we can analyze both quantitative and qualitative data.
69
00:05:44,627.994 --> 00:05:48,617.999
Like, you mentioned, like, the emails and, you know, note-taking apps and things like that.
70
00:05:48,967.991 --> 00:05:53,678.003
You know, r- really all you need to know then is how to navigate the spreadsheet or whatever other program you're using.
71
00:05:53,678.003 --> 00:06:02,160.001
So do you think the days of creating spreadsheets by s- by hand from scratch, if you will, are over? I don't think so.
72
00:06:02,180.001 --> 00:06:14,974.001
I, I think spreadsheets continue to be good at Very, very easily, um, looking at tabular data.
73
00:06:15,624.001 --> 00:06:26,384
And I also think that it's important when you think about a spreadsheet to distinguish between spreadsheet the format and spreadsheet the storage mechanism.
74
00:06:26,394.001 --> 00:06:39,814
So having data in Excel in a format or Google Sheets in a format that's a spreadsheet format is very different than actually analyzing the data in that format.
75
00:06:39,824 --> 00:07:05,523.998
So what I mean is, regardless of whether new data, you know, is created from scratch in a spreadsheet, I think probably it's, it's more important to say, "How do I analyze all of this data that's in the spreadsheet?" And in my observation, most people who are heavy users of spreadsheets are not really doing analysis on those spreadsheets.
76
00:07:06,123.999 --> 00:07:14,214.003
They are, they are doing simple addition and math on those spreadsheets, which is by far the most common use case.
77
00:07:15,33.999 --> 00:07:36,833.998
What some of the AI tools are now giving people the ability to do, and these are people without strong analytical backgrounds or technical backgrounds of spreadsheet formulas, is they're allowing people to ask and get answers to business questions and analytical questions using plain English.
78
00:07:37,103.999 --> 00:07:45,694
And that is enormously powerful because then the access to those answers comes quicker and cheaper and easier than it did before.
79
00:07:45,694.003 --> 00:07:48,783.997
So spreadsheets are not going away.
80
00:07:48,853.999 --> 00:07:56,33.996
I don't think creating spreadsheets from scratch is going away, but I think the intelligence we can get from them is much greater than before.
81
00:07:56,383.997 --> 00:07:59,944
Yeah, that to me is the most ex- exciting part of all this, is that you can...
82
00:07:59,963.999 --> 00:08:08,944
It just opens up whole new avenues of, of opportunities to, uh, to find the story that's embedded within the numbers and things, which not a lot of people are good at doing.
83
00:08:08,944 --> 00:08:17,133.998
So you could really leverage technology to help you, uh, you know, tell me what this means, you know? You can ask questions like that and really find out a lot of really cool stuff.
84
00:08:17,133.998 --> 00:08:21,824.005
So, so let's talk about AI because that's, uh, that's a big part of the equation nowadays.
85
00:08:22,204.005 --> 00:08:38,874.004
You know, we've, we've been talking about AI for a number of years now, but, uh, is it still like a shiny new toy or have we moved to the point of ubiquity where it's like it's everywhere, everybody's using it? What, what do you think on that, Matt? Where are we at in the process? I think it's, I think it's all of those things and more.
86
00:08:38,963.999 --> 00:08:45,733.992
So, um, there is no shortage of shiny new toy options every day.
87
00:08:46,643.992 --> 00:08:50,754.003
Um, there's no shortage of gratuitous uses of technology...
88
00:08:50,783.991 --> 00:08:51,903.997
Of, sorry, of AI.
89
00:08:52,783.991 --> 00:09:09,250.001
Um, we have certainly moved to a point where the individual productivity benefits, whether it's to the consumer or to the business owner, the individual productivity benefits are, I think, pretty clear.
90
00:09:10,340.001 --> 00:09:19,680.001
Where I think there is substantial ambiguity is, uh, top-line opportunity for small businesses.
91
00:09:19,680.001 --> 00:09:36,780.001
Certainly, we've seen confusion and debate in big enterprise about, um, realizing top-line opportunity, but it is ubiquitous and, uh, the, the technology is ubiquitous.
92
00:09:36,780.001 --> 00:09:40,299.999
Whether the use is ubiquitous or not i-is a different story.
93
00:09:40,299.999 --> 00:09:55,280.001
But I would look at that in, in the following way As the technology evolves in the beginning, and we're still, in my opinion, in the beginning, you tend to think about AI as a set of tools.
94
00:09:55,310.001 --> 00:10:02,990.001
You tend to think about, uh, using a ChatGPT or using a Claude for a specific reason.
95
00:10:03,000 --> 00:10:04,540
You say, "I'm using AI."
96
00:10:04,620.001 --> 00:10:12,160.001
And what's happening is over time, that technology is being embedded in everything we do.
97
00:10:12,170 --> 00:10:17,060
So Google Workplace has tons of AI features built in.
98
00:10:17,69.999 --> 00:10:20,70.001
Microsoft Office has AI filters built in.
99
00:10:20,570.001 --> 00:10:30,189.999
At some point, I suspect, just like with the internet and mobile, you don't do internet or do mobile, you use the tools that integrate those things.
100
00:10:30,230.001 --> 00:10:31,060
They're just there.
101
00:10:31,060 --> 00:10:34,610.001
They're just part of the ether, just like technology is part of the ether.
102
00:10:34,610.001 --> 00:10:45,919.998
So at some point, pretty soon I would imagine, we're not gonna be talking about AI and not AI in terms of, in terms of, uh, business productivity.
103
00:10:45,960.001 --> 00:10:49,790.002
It's just going to be everywhere in all the tools at once.
104
00:10:50,320.001 --> 00:10:54,230.003
Now that being said, that's not necessarily a good thing.
105
00:10:54,839.998 --> 00:11:02,359.997
There's AI washing, which essentially slaps an AI label on everything, and suddenly everything is an AI tool.
106
00:11:02,399.996 --> 00:11:04,559.999
Well, that's not, that's not true either.
107
00:11:05,210.001 --> 00:11:23,950
But in summary, uh, if you're a small business person, if you're a consumer, if you're a large business person, you are going to be using AI and you will find AI in all of the products and services that you use, regardless of whether you're aware that they're there.
108
00:11:25,210.001 --> 00:11:25,919.999
Matt, it's Louis.
109
00:11:25,920 --> 00:11:34,839.996
I, I, I gotta imagine that at some point, if, if not already, you're gonna be using AI and you don't even know it, right? Um, and I think that- That's what I mean.
110
00:11:34,839.998 --> 00:11:42,769.997
Yep yeah, and that's, I think that's one of the exciting things about it becoming part of the ether, right? It's, it's a technol- it's a tool, and it's so refreshing to hear you say it as a tool.
111
00:11:43,180.002 --> 00:11:49,259.996
Um, because if you don't consider it a tool, it takes on a different meaning, if you will.
112
00:11:49,259.996 --> 00:11:58,950.001
And, and we constantly hear about AI almost becoming a life form versus a tool, and that's something we battle with all the time, especially from a, from a small business perspective.
113
00:11:58,950.001 --> 00:11:59,610.001
It's like, how do you...
114
00:12:00,229.997 --> 00:12:07,919.996
I, I guess I'd ask the question, you know, how do you dispel that notion that AI is not a, not a life form? It's certain- most certainly a tool, and, and can it...
115
00:12:08,99.992 --> 00:12:11,50.002
Is it complex and is it highly capable? Yes.
116
00:12:11,339.998 --> 00:12:15,49.994
Um, but it's more about how you leverage it as a tool than anything else.
117
00:12:15,379.991 --> 00:12:19,904.001
What's your response to that? I, I think what you said is true.
118
00:12:19,934.001 --> 00:12:44,174.001
I think that I, I've struggled when trying to think about AI in, in sort of an abstract way about w- when is it a tool and when is it something substantially more powerful than a tool? Now, in the past, before AI, the things in technology that we've called tools have been sort of more tactical.
119
00:12:44,184.001 --> 00:12:46,183.999
They've been more point solutions.
120
00:12:46,984.001 --> 00:13:00,464.001
If you have a hammer, a hammer is clearly a tool, and you use it to drive in a nail If you have a spreadsheet, it's a tool that allows you to enter data in row and column format.
121
00:13:01,484.001 --> 00:13:14,794
AI is different in the sense that it solves such, or it can solve and can be used in such a general way.
122
00:13:15,634 --> 00:13:22,454
It can solve or help you solve a set of problems that are incredibly general.
123
00:13:23,294 --> 00:13:33,844
And when something can be applied in such a general way, it kind of loses its definition as a tool and becomes something more powerful.
124
00:13:34,554 --> 00:13:57,494
Now, I'm not hyping AI, but what I am saying is that over time, I would imagine we're gonna stop looking at AI as a tool and start looking at AI as a set of capabilities and something that's sort of embedded in, in what we do.
125
00:13:57,524.001 --> 00:14:08,914.003
And what I mean is, I get back to my previous statement, we're gonna not say, "I'm using AI to do such and such," but rather AI is just going to be there.
126
00:14:10,424.003 --> 00:14:24,903.997
Now, you talked about the y- you know, or, uh, it was you or Joe who mentioned sort of, you know, AI maybe being omnipresent or, or, uh, uh, you know, taking on some of the future-- a future life form.
127
00:14:25,433.997 --> 00:14:34,273.997
I don't know whether that will happen or not either, but again, that speaks to the generality of the, of the technology.
128
00:14:34,333.999 --> 00:14:54,23.996
If the technology can eventually solve such a broad array of problems as the marketing material from the big AI companies will tell us, it ceases to be a tool and starts to really become something much more integrated in, into what we do.
129
00:14:54,344.001 --> 00:15:02,184.002
And I don't know where the direction is going, but certainly for small businesses today, we are more looking at it from, from the tool perspective.
130
00:15:02,424.003 --> 00:15:12,343.993
I still think, and I still urge people, and I'll stop here, is, is we should be focusing on the capabilities and not on the tool.
131
00:15:12,353.996 --> 00:15:19,883.999
The question is always what do we want to accomplish and not what tool are we grabbing to figure out how to use it or where to use it.
132
00:15:20,853.996 --> 00:15:29,883.99
So w- with that in mind, I mean, let's talk about some of the f- the functions that entrepreneurs could and, and should be using AI to do for them.
133
00:15:29,913.993 --> 00:15:38,823.995
I mean, we'll talk about two, on an AI assist basis where you're still making the decisions and in control and stuff, and some of the things that you can do on an automated basis.
134
00:15:38,833.991 --> 00:15:41,823.997
So let's talk about some examples of how, how it can be applied, Matt.
135
00:15:41,833.991 --> 00:16:05,106
What do you, what do you-- what comes to mind when you think of that? I, I like to focus on w- AI for things that, uh, I could do faster and better with it, or, uh, make me more money, or make me more profit, or do things that I couldn't do without it.
136
00:16:06,246 --> 00:16:12,076
What I don't need to do is use AI to solve problems that already have solutions.
137
00:16:13,256 --> 00:16:29,240.001
So when we talk about this idea you mentioned, Joe, of, of AI assist, right? Versus, versus automation, um- You know, one great example is, is, is writing.
138
00:16:30,360.001 --> 00:17:03,810.001
Now, I'm a big fan of not generating content with AI, but rather letting my content and my ideas sort of lead, but then using AI for the things it's good at, which is to, um, help me target the writing, help me personalize the writing toward a given audience, uh, help me emphasize points, help me make points more succinctly, help me, um, take a concept and explain it in different ways.
139
00:17:03,810.001 --> 00:17:07,779.999
That's sort of a canonical use of, of AI.
140
00:17:08,150.001 --> 00:17:21,819.998
Another AI kind of assist, uh, you know, situation is where you are interacting with data and you wanna use AI to answer questions about that data.
141
00:17:21,819.999 --> 00:17:23,890.001
That's not an automated process.
142
00:17:23,900.001 --> 00:17:33,99.997
That's not necessarily a, a repeatable process, but it's one where AI, you know, offers enormous, enormous value.
143
00:17:33,759.997 --> 00:17:48,9.996
Um, another answer, another, uh, way that I'm using, and I see my clients using AI in enormously, uh, powerful ways, is as, as a thought partner.
144
00:17:48,679.996 --> 00:18:10,219.999
So AI, uh, has the ability to, um, consume lots of data and conduct what can be thoughtful, for use of an odd term when we're talking about AI, but thoughtful conversations around a given topic.
145
00:18:10,369.998 --> 00:18:24,309.999
So using AI as a thought partner to talk about your business, to talk about your options, whether it's personal or business or whatever it is, um, can be, can be enormously powerful.
146
00:18:24,310 --> 00:18:26,689.995
It doesn't get tired of you, it doesn't get annoyed.
147
00:18:27,119.998 --> 00:18:41,270.002
Um, but, but using it as, as a thought partner, whether it's a financial thought partner, a strategic planning thought partner, all those kind of things are enormously powerful uses of AI, and they're not automated AI use cases.
148
00:18:41,270.002 --> 00:18:44,224.9935
They are interactive sort of AI-assisted use cases.
149
00:18:44,224.9935 --> 00:18:46,70.003
Well, that's exactly what we use it for.
150
00:18:46,70.003 --> 00:18:46,680.003
Another option- Hold on.
151
00:18:46,680.003 --> 00:18:48,390.002
Let me interrupt you briefly here, Matt.
152
00:18:48,610.003 --> 00:18:52,470.005
That's what we use it for when, on our economic, um, analysis report that we do.
153
00:18:52,470.005 --> 00:19:04,749.997
We take all these news articles, like six hundred fifty, seven hundred news articles, and I ask AI a bunch of questions like, "What's all, what's all this mean? What does it mean for small businesses, et cetera?" You know, that's the, uh, the report that we put out each month.
154
00:19:04,750.003 --> 00:19:16,290.002
And that's exactly what we do is, uh, you know, we kinda have a little bit of a conversation with it, if you will, say, you know, "What do you think about all this stuff?" So that's a, that's another example of, of how we here at CIC use it for that purpose.
155
00:19:17,684.001 --> 00:19:18,664.001
A- absolutely.
156
00:19:18,714.001 --> 00:19:25,894.001
Uh, and, and that's synthesis, and that was gonna be the, the second y- you know, area that I think is so enormously powerful.
157
00:19:26,104.001 --> 00:19:40,314.001
If you, if you go back to, you know, four years ago and you wanted to do research or you wanted to get answers to questions, you would go to Google and you would search for things.
158
00:19:40,314.001 --> 00:19:46,664
But Google and search engines didn't give you answers to questions.
159
00:19:46,664.001 --> 00:19:52,404
What they did was they gave you resources, and you had to formulate the answers.
160
00:19:52,934.001 --> 00:20:03,674.001
And where Google is of course the best is, you know, find me the best product reviews for product XYZ, and you get results back.
161
00:20:04,403.998 --> 00:20:12,960.001
But the second you need To actually synthesize information around multiple sources, you were doing that manually.
162
00:20:13,700.001 --> 00:20:18,130
And that's where AI-based synthesis is incredibly powerful.
163
00:20:18,590.001 --> 00:20:31,410
So one of the use cases today that I, uh, use AI for the most, and so many people do, is, of course, meeting transcriptions, right? So you have a meeting, you get a transcription.
164
00:20:32,160.001 --> 00:20:38,680.001
AI is wonderful because for me, I'm taking less notes, I'm listening better, and I end up with a transcript.
165
00:20:39,180 --> 00:20:39,740.001
That's easy.
166
00:20:39,740.001 --> 00:20:40,730.001
That's low-hanging fruit.
167
00:20:40,770 --> 00:20:45,769.999
But what I'm doing now is just like the work I do with, with CIC.
168
00:20:46,360 --> 00:20:59,249.999
I have meeting transcripts and other spreadsheets and artifacts and documents and whatever other, uh, data there is from my interactions, Joe, with you and the, and the, and the CIC, uh, TA clients.
169
00:21:00,39.999 --> 00:21:07,520.001
Um, I now have this corpus of transcripts and other data that I can ask questions across.
170
00:21:08,189.998 --> 00:21:14,449.998
I can ask information without me having to-- I can ask questions without me having to mine that information manually.
171
00:21:14,879.997 --> 00:21:16,229.996
Show me the themes.
172
00:21:16,540 --> 00:21:18,190.003
Show me the common elements.
173
00:21:18,419.998 --> 00:21:21,659.999
Help me think about such and such in a way that I haven't thought about it.
174
00:21:22,349.998 --> 00:21:26,190.003
Uh, based on my objectives, what may I be missing? All of those things.
175
00:21:26,540 --> 00:21:39,409.997
And again, this powerful capability all comes from the idea that with AI, you don't focus on the mechanics, but you focus on what your intent is.
176
00:21:39,889.999 --> 00:21:45,880.003
With AI, your intent is whatever the goal is, you know, you want to achieve.
177
00:21:46,180 --> 00:21:48,959.997
I want the answers to questions is my intent.
178
00:21:49,309.997 --> 00:21:56,229.997
I can focus on synthesizing the answer rather than focusing on the mechanics of generating that answer.
179
00:21:56,260.002 --> 00:22:00,280
And to me, that's where, uh, AI really plays in.
180
00:22:00,830.001 --> 00:22:03,559.996
I'll tell you one other part of your question.
181
00:22:03,559.997 --> 00:22:23,620.001
You said, "When should AI consider to be used on an automated basis?" I think that processes or workflows that are repeatable, processes that are well-defined Processes that are actually messy are good for AI.
182
00:22:23,650.001 --> 00:22:29,50.001
And what I mean is you could be a well-defined process but messy, meaning the data can be a little bit messy.
183
00:22:29,440.001 --> 00:22:32,550.001
The data perhaps is unstructured.
184
00:22:32,550.001 --> 00:22:39,100
It could be a call transcript, it could be email communications, it could be, uh, things like that.
185
00:22:39,110 --> 00:22:54,710.001
So when you look at customer service, uh, responses, the ability of AI to automate a customer service workflow is enormous because AI can synthesize data in a way that provides the user answers to their question.
186
00:22:54,710.001 --> 00:22:59,369.999
Today, chatbots are a thousand times better than they were five years ago.
187
00:22:59,600.002 --> 00:23:07,899.999
They're actually useful in getting you answers and automating those processes, as opposed to before when they were simple, you know, hit one for this and two for that.
188
00:23:08,319.999 --> 00:23:14,260
So that's an example of, of automation, um, that is a good use case.
189
00:23:14,290.001 --> 00:23:19,660.001
All right, so here's, here's an extreme example perhaps of a, uh, of, of automation.
190
00:23:19,660.001 --> 00:23:21,450
This is an article I pulled out of The New York Times.
191
00:23:21,450 --> 00:23:22,90.001
I'll read it to you.
192
00:23:22,479.999 --> 00:23:26,489.997
"A pastor says his digital twin can bring more people to God.
193
00:23:27,9.998 --> 00:23:31,889.999
To meet the growing needs of his church, Justin Lester created an AI duplicate of himself.
194
00:23:31,929.996 --> 00:23:34,370
Every day, he sees more people using it.
195
00:23:34,399.997 --> 00:23:43,629.996
Le- Lester cloned his voice and trained his digital twin using almost two million words of his sermons, podcast, books, text me- text messages, and emails.
196
00:23:43,710.003 --> 00:23:49,859.997
For the last several months, it's been publicly available on his website and Facebook page for online chats or phone calls.
197
00:23:50,210.003 --> 00:23:58,460
It has learned to mimic his bent toward optimism, his obsession with the Green Bay Packers, and his casual slang, but it did more than just regurgitate what he had already said.
198
00:23:58,869.999 --> 00:24:06,109.995
It mapped his thinking onto new conversations to provide advice around the clock for people he'd never met who had questions he'd never been asked."
199
00:24:06,719.997 --> 00:24:08,199.997
That seems pretty freaking cool.
200
00:24:11,212.001 --> 00:24:12,832.001
I think, I think that's happening.
201
00:24:12,832.001 --> 00:24:23,262.001
I don't know whether this, this gentleman ha- uh, whether his digital twin is actually a visual avatar, uh, or, or just a, a sort of text, you know, chatbot persona.
202
00:24:23,742.001 --> 00:24:59,152
But we are seeing more and more this idea of, um, taking, taking people's personas, particularly people that are, uh, either high p- high profile or have domain expertise, and mapping them toward digital avatars with lifelike, you know, um, uh, interfaces and the persona of that individual, putting them in front of the perspective or the possible clients, and the clients are having, you know, some powerful interactions.
203
00:24:59,152 --> 00:25:09,912
Now, regardless of how you feel about this, because certainly it is a topic, you know, that is well contested because the human side of things is, I think, very important to maintain.
204
00:25:10,682.001 --> 00:25:24,712.002
I've worked with several clients who are doing counseling and, uh, they're doing virtual counseling, but they also want to extend that virtual counseling beyond the availability of the human counselor.
205
00:25:24,961.996 --> 00:25:36,541.996
And so they are providing, um, avatar-based chatbots which mirror the personality of the counselor or have other sort of domain-specific knowledge.
206
00:25:37,052 --> 00:25:43,261.997
And, you know, people are finding comfort in seeing that visual interface and in having those conversations.
207
00:25:43,891.996 --> 00:25:44,852.001
So this is happening.
208
00:25:44,861.998 --> 00:25:48,961.995
You look at companies like HeyGen, which is the, which is the avatar company.
209
00:25:49,452.002 --> 00:25:51,311.996
Uh, I think they're having some success.
210
00:25:51,312.002 --> 00:25:58,712.001
I think what you're saying, Joe, is going to be a growing use case- Mm-hmm particularly for, um, younger people.
211
00:25:59,391.996 --> 00:26:08,51.996
Now, w- uh, in fairness, we can't talk about all this great stuff without talking about the other side of the equation, which is, you know, some of the not so great things about AI.
212
00:26:08,511.999 --> 00:26:09,291.999
People, you know...
213
00:26:09,291.999 --> 00:26:15,092
I, every month I capture articles about AI's taking people's jobs, the environmental impact.
214
00:26:15,431.997 --> 00:26:22,561.992
Uh, this past month particularly, we can talk about this, you know, talk about rogue AI systems, you know, operating out of bounds and stuff.
215
00:26:22,562.002 --> 00:26:25,361.999
What can we do to mitigate some of those concerns? 'Cause they're real.
216
00:26:25,422.003 --> 00:26:26,711.996
I mean, these are real issues.
217
00:26:27,924.001 --> 00:26:29,454.001
Yeah, they're, they're very real.
218
00:26:29,474.001 --> 00:26:42,104.001
I think that the rules around these things don't change over time with respect to trusting any source or using technology.
219
00:26:42,104.001 --> 00:26:53,254
The rules are pretty basic, and while the technology evolves and the risks become higher, our own individual responsibility kind of doesn't change.
220
00:26:53,773.999 --> 00:27:00,683.999
You, you, you wanna be careful about who you give your data to and why you give your data.
221
00:27:01,154.001 --> 00:27:05,294.001
You, you wanna have trust with the people you give your data to.
222
00:27:05,803.998 --> 00:27:20,974
You wanna be thoughtful, um, about, about what is the role of AI in your business and what part of your business do you want to, uh, absolutely, you know, create a red line between and, and remain human.
223
00:27:21,463.998 --> 00:27:30,764.001
Um, there, there are businesses where, you know, the business owner will intentionally say, "This part of my business will always remain human."
224
00:27:31,134 --> 00:27:36,433.997
Customer service, for example, may be a core capability, and that should remain human.
225
00:27:36,434.003 --> 00:27:43,534
So I think being very intentional on how you use AI is gonna help you s- uh, avoid some of those things.
226
00:27:44,163.999 --> 00:27:49,840.001
I think that, um Um, education is important.
227
00:27:49,900.001 --> 00:27:52,940.001
I think legislation is a, is a tricky one.
228
00:27:53,520.001 --> 00:28:06,0.001
Uh, I'm not a big, big fan of putting the government, uh, as, as a, as a, a layer in between AI models and people.
229
00:28:06,0.001 --> 00:28:10,400.001
I've heard some people talk of, you know, let's say an FDA for AI.
230
00:28:10,400.001 --> 00:28:25,0.001
I think there's a role, uh, for that sort of regulatory structure, but it has to be very, very carefully thought through because the technology cycles are so much more, um, compressed.
231
00:28:26,10.001 --> 00:28:41,500.001
The challenge, I think, Joe, is the individual consumer may not have a choice in a lot of these things, but the consumer still has a choice in the, the security basics that they've always had.
232
00:28:41,999.999 --> 00:28:50,630.001
Securing your own data, making sure who you give it to, using strong passwords, backing up your data, not putting it where you don't want.
233
00:28:51,19.996 --> 00:28:54,349.997
All those things are the low-hanging fruit that keep you safe.
234
00:28:54,349.998 --> 00:29:10,520.001
The bigger issues around how data will be used, whether it's flock cameras or, uh, government databases, those are all things that we don't individually have a lot of control over.
235
00:29:11,160.003 --> 00:29:14,910.001
Um, but I hope there's common sense answers around.
236
00:29:14,929.999 --> 00:29:36,119.998
I'm not confident that there will be, um, but I do think here we have an extra danger, Joe, because the power here for, um, uh, malicious use is, or the capability, I should say, the possibility of malicious use is sort of greater than, you know, you know, than it ever was.
237
00:29:37,309.997 --> 00:29:37,949.995
For sure.
238
00:29:38,330.001 --> 00:29:46,159.998
I saw an article, uh, just, uh, just in the last couple of days that, uh, th- this, this one, when I saw this one, it kinda, it kinda blew my mind.
239
00:29:46,159.998 --> 00:29:54,699.999
It was talking about this guy who was, um, was involved in a legal matter and, um, let me see if I can find the article for you here, here real quick.
240
00:29:54,709.991 --> 00:29:56,629.997
It's, um, uh...
241
00:29:56,629.997 --> 00:29:57,40.004
Da, da, da, da, da.
242
00:29:57,729.995 --> 00:30:01,559.993
Judge sanctions Connecticut man for hidden AI prompt.
243
00:30:02,50.002 --> 00:30:11,669.997
So a gentleman by the name of Matthew Elliott, who was representing himself in a medical malpractice case, hid a message in a Connecticut court filing that was meant to be read by a machine, not a judge.
244
00:30:12,9.994 --> 00:30:25,592.001
The message, written in tiny white text, told any artificial intelligence system that encountered the document to agree with his argument and help undo an earlier court ruling in a court case That is, like, insane.
245
00:30:25,622.001 --> 00:30:31,902.001
I mean, what do you th- I mean, that is, this is, this just captures- Yeah I pulled this out, I mean, what's, uh, t- uh, actually, this was the day before yesterday.
246
00:30:32,102.001 --> 00:30:34,412.001
When I read that, I was like, "This is crazy, man."
247
00:30:34,422.001 --> 00:30:37,282.001
But, you know, to your point- This is- Go ahead.
248
00:30:37,372 --> 00:30:40,422.001
Th- this is called prompt injection.
249
00:30:40,442 --> 00:30:41,472
Yes, that's exactly, yeah.
250
00:30:41,472.001 --> 00:31:05,781.999
And this is, th- this is, you know, this is a common, um, vulnerability today that AI systems can encounter where, uh, you as an individual are using an AI system and you essentially inject, often in a covert way, instructions that aim to have the AI break out of the guardrails that were put on it.
251
00:31:05,791.999 --> 00:31:20,902.001
So for example, if you go to ChatGPT and you ask ChatGPT, "How do I make a nuclear weapon?" Uh, y- you, you will probably find that ChatGPT gives you resistance to answering that question.
252
00:31:20,902.001 --> 00:31:25,292.001
Now- It knows how to do it, but it won't tell you It knows how to do it, but it won't tell you.
253
00:31:25,292.001 --> 00:31:28,602.001
Now, creative people can, uh, circumvent those things.
254
00:31:28,602.001 --> 00:31:46,312
When you talk about prompt injection, prompt injection is a little bit more of a sophisticated, you know, more nefarious attack where you are covertly embedding instructions into production AI systems that, again, make them operate outside, you know, their intended boundaries.
255
00:31:46,622.001 --> 00:31:54,952.001
This, uh, is, I, I think, going to be addressed in several ways, although it will continually be a leapfrog situation.
256
00:31:54,952.001 --> 00:32:00,451.999
But you're gonna have the AI model developers putting guardrail systems on.
257
00:32:00,451.999 --> 00:32:05,972.001
You're gonna have third-party software looking for that content and trying to detect that content.
258
00:32:06,521.998 --> 00:32:19,731.999
Um, of course, and this isn't directly related, but when you look at, uh, any AI output, you know, keeping yourself out of trouble, one of the best ways to do that is to not blindly trust that output.
259
00:32:20,082 --> 00:32:21,841.999
But, but that's always the case.
260
00:32:21,841.999 --> 00:32:26,291.998
We should never blindly trust, uh, anything from any source.
261
00:32:26,291.999 --> 00:32:28,372.001
And so that's, that's not new here.
262
00:32:28,372.001 --> 00:32:35,681.997
What's new is that I think the stakes are higher and the, uh, the possibility for abuse is higher.
263
00:32:35,691.996 --> 00:32:41,752.002
Not to mention the fact that, um, y- you know, AI can be...
264
00:32:41,792.003 --> 00:32:46,771.996
Uh, we talked about rogue AI, and maybe we'll get into that a little bit, a little bit later.
265
00:32:46,771.998 --> 00:32:56,851.999
But the idea of AI, quote-unquote, thinking means that the AI systems can and will be able to operate outside of human control.
266
00:32:56,852 --> 00:33:01,371.999
So the guardrails that we put around these things, I think, are gonna be more important over time.
267
00:33:01,911.998 --> 00:33:05,491.999
So I think it's really a matter of, like I always say to people, trust but verify.
268
00:33:05,491.999 --> 00:33:12,981.998
You know, like for instance, you know, our economic, uh, summary reports that we put out from all the articles, I read every word of it before we, we publish it.
269
00:33:12,991.999 --> 00:33:21,851.999
So, you know, even though I trust the, the, the, the, the chat, you know, you know, ChatGPT of what it gave me, I still read it just to make sure that everything is, uh, is good.
270
00:33:21,852 --> 00:33:24,802.002
So in the end, I mean, I think there is some trust in AI.
271
00:33:24,802.002 --> 00:33:25,61.997
It's...
272
00:33:25,61.999 --> 00:33:25,691.997
But should it...
273
00:33:25,731.997 --> 00:33:33,31.992
What about people using AI? Can we trust people using it? That, that's probably more the issue than, than the AI system itself.
274
00:33:33,032 --> 00:33:41,18.001
Would you agree? Yes, I would agree.
275
00:33:41,58.001 --> 00:33:50,628
I think that people are people, and they're gonna fall back to the, you know, c- uh, pa- path of least resistance.
276
00:33:51,218.001 --> 00:33:58,78.001
And when you fall back to that, you will do less checking, you will do less validation, you will become more dependent on things.
277
00:33:58,088 --> 00:34:27,658.001
I will tell you from experience that when I am using AI as a thought partner or when I am using AI to build, uh, a technical, you know, artifact, the, uh, sort of allure of outsourcing your brainpower to the AI is very strong, and you need to resist that and really be part of the process in an active way.
278
00:34:28,208.001 --> 00:34:37,688
When you go back again to, like, the hammer and the nail example, it's, it's sort of hard for a hammer to do too much work for you.
279
00:34:37,927.998 --> 00:34:39,308.001
It does one thing.
280
00:34:39,508 --> 00:34:42,368
You have to swing the hammer, and it puts in the nail.
281
00:34:42,927.998 --> 00:34:54,387.999
And with AI, you can fall into this trap of essentially outsourcing your own brain and your own judgment and your own critical thinking to the AI.
282
00:34:54,807.996 --> 00:35:06,518.001
If that's what you mean by p- trusting people to do those things, that's where I think we're gonna have a hard time because when an answer is so easy to come by, you're gonna take the shortcut.
283
00:35:06,817.998 --> 00:35:07,87.998
Yep.
284
00:35:07,167.996 --> 00:35:11,518.002
And it's not clear to me what the answer to that problem is.
285
00:35:11,548.001 --> 00:35:19,207.999
Certainly, kids today in school are not learning, uh, h- y- you know, how to avoid taking the easy path.
286
00:35:19,567.998 --> 00:35:19,887.998
Right.
287
00:35:20,078 --> 00:35:20,627.997
That's true.
288
00:35:21,228.001 --> 00:35:23,547.998
Um, okay, let's, uh, let's shift gears a little bit here.
289
00:35:23,548.001 --> 00:35:33,287.996
So one of the biggest changes impacting small business has been the shift from search engine optimization to generative engine optimization.
290
00:35:33,618.001 --> 00:35:34,928.002
So what do small business...
291
00:35:34,957.997 --> 00:35:41,288
And I'll, I'll, I'll ask Luke also to, to join in the conversation on this question, uh, because this is his area of expertise.
292
00:35:41,288 --> 00:35:51,608.001
But what do small businesses need to do in order to remain visible on search engines in the, in the, in today and in the future? I'm gonna punt to Luke on that, Joe.
293
00:35:51,618.001 --> 00:35:51,718.001
Okay, that's fine.
294
00:35:51,728.001 --> 00:35:54,768.001
I, I can't add any value in that one.
295
00:35:55,158.001 --> 00:36:00,468.001
What do you say, Luke? Yeah, I mean, uh, this is a, an area that continues to evolve for sure.
296
00:36:00,498.001 --> 00:36:02,028
We talk about it all the time.
297
00:36:02,758 --> 00:36:24,924.001
I mean, the headline here is that the GEO, um, the generative optimization that is being done by Chatbots from, um, GPT to Claude to Gemini, um, in our opinion, is actually making it easier for people to find the stuff that they want.
298
00:36:26,304.001 --> 00:36:37,454
So SEO, search engine optimization, based on our experience, and we're going back to 1999, let's talk about Netscape and so many other engines that have come and gone.
299
00:36:38,304 --> 00:36:57,643.999
The, the way that search engine optimization worked or still works, in our opinion, is somewhat flawed because it's based on the data, the copy, um, the way that sites are structured, and there is an opportunity to fool those engines.
300
00:36:58,263.999 --> 00:37:08,703.998
I think that, and we see this with the, um, the GEO, the bots are smart, and they do a lot of the research.
301
00:37:08,704.001 --> 00:37:14,13.998
They're looking at some of the things that SEO typically doesn't take into consideration.
302
00:37:14,313.998 --> 00:37:30,264.003
People are also talking to their bots, so they're driving, and they're asking ChatGPT a couple questions about how to find the best daycare near them, and they want something that has a high reputation score, and they want to know how many years they've been in business.
303
00:37:30,283.999 --> 00:37:42,334.003
And so they're, they're using their, um, you know, their own experience, and they're communicating with the chatbots in a way that ultimately, in our experience now, is yielding better search results.
304
00:37:42,803.996 --> 00:38:17,163.997
And so the way that, um, we see this affecting clients is, and the way that we are coming at website development, storytelling, uh, thought leadership, blog creation, podcasting, is that for clients who are really thoughtfully connected to the details, who are curating meaningful information, who are telling authentic stories on their website and through their emails and through social, all these things are coming back to the way that the chatbots, through GEO, are elevating findings for people.
305
00:38:17,843.997 --> 00:38:26,563.995
And so we're really excited about it- Yeah and, and ultimately feel like there is a level of validation attached to some of the stuff that we, we've been doing.
306
00:38:26,633.998 --> 00:38:32,614.002
Um, and that, that's I think where we see a big differentiation in terms of SEO, which obviously is critical.
307
00:38:32,614.002 --> 00:38:35,383.998
You still have to be mindful of Google's algorithm.
308
00:38:35,383.998 --> 00:38:45,464.002
Um, but GEO has allowed us to lean into the stuff that we're good at from a marketing storytelling perspective and see, uh, better results- Mm-hmm um, in terms of search findings.
309
00:38:45,824 --> 00:38:59,253.992
Yeah, and I think we can use CIC as a test case, right? I mean- Great example um, our archaic website versus what we have today has been mind-blowingly different- Blowingly different from a, from a v- traffic perspective on the web.
310
00:38:59,283.992 --> 00:39:05,913.994
I mean, it's, it's pretty incredible, and, uh, obviously I credit you, Luke, for, for the, the expertise in that.
311
00:39:06,313.991 --> 00:39:16,472.001
But what can you attribute To sort of the success we've seen in terms of the highly vis- visible data and storytelling that we're doing and how that's leading to more lead generation.
312
00:39:17,202.001 --> 00:39:18,262.001
That's a great question.
313
00:39:18,262.001 --> 00:39:20,712.001
I mean, I am constantly blown away by the CIC site.
314
00:39:21,292.001 --> 00:39:21,782.001
It's crazy.
315
00:39:23,472 --> 00:39:25,992.001
We went from zero to hero on that.
316
00:39:26,32.001 --> 00:39:26,322.001
Yeah.
317
00:39:26,322.001 --> 00:39:29,352
Um, I think it, you know, this is a really interesting example.
318
00:39:29,352 --> 00:39:49,66.001
We're still picking apart the information about what happened when we went from the, what was it? 2016, 2018 version of a WordPress website to the 2025 Arguably 2026 Wix model that was, uh, just better structured.
319
00:39:49,416.001 --> 00:39:55,326.001
We did a lot of work, obviously, on the content curation, which speaks to what I had mentioned in terms of the storytelling.
320
00:39:55,336 --> 00:39:58,756
We talk about the service, uh, funding vehicles in a different way.
321
00:39:58,796 --> 00:40:27,376.002
And the truth here is that the, the blog editorial content, the podcasting, the stuff that Joe's doing in terms of our ChatGPT analysis, like all, all of these things, uh, including the LinkedIn elevation, the storytelling there, the emails that we send, the 504 rates, and the SBA outbound internal link, uh, external inbound link.
322
00:40:27,765.999 --> 00:40:31,306.002
Sounds technical, but the SBA obviously lists us for people who...
323
00:40:31,326 --> 00:40:34,285.999
CIC for people who are looking for those loan vehicles.
324
00:40:34,835.998 --> 00:40:47,525.999
Um, those-- all of those things together working in concert with how people are using the chat engines, and we can see on the back end of our website how many people are using ChatGPT, Claude, and Gemini to find CIC.
325
00:40:47,866.001 --> 00:40:52,135.999
And that number, uh, from a month over month increase is off the charts.
326
00:40:52,146 --> 00:41:06,256.001
So we know that it's working, and we know that it's working because of how we built the site and how we constantly keep it up to date and how we work to tell the story not only of what CIC does, but more importantly, what CIC does for clients.
327
00:41:07,006 --> 00:41:08,165.996
And I think that's the thing.
328
00:41:08,165.996 --> 00:41:19,95.995
You can pat your own back, you can thump your own chest, but the, the chatbots, like humans, in some capacity are intelligent, and they know that you can say whatever you want about yourself.
329
00:41:19,125.996 --> 00:41:25,325.999
But the real meaningful content, the real validation, the real reviews, they come from what other people are saying about you.
330
00:41:25,425.999 --> 00:41:25,665.999
Right.
331
00:41:25,665.999 --> 00:41:28,756.002
So that external link coming from SBA is a huge endorsement.
332
00:41:29,105.999 --> 00:41:37,446.002
Not to mention all the other things that we do, podcast guests, LinkedIn case studies, new loan closing highlights.
333
00:41:38,25.997 --> 00:41:38,345.995
I don't know.
334
00:41:38,345.997 --> 00:41:49,525.999
I don't wanna say that we've, like, cracked the code, but, um, we're getting better at it, and we love the opportunity, uh, to see it working like it is for CIC.
335
00:41:49,925.999 --> 00:41:50,565.998
Yeah, I mean- Blown away.
336
00:41:50,715.999 --> 00:41:51,265.995
For sure.
337
00:41:51,265.995 --> 00:41:53,425.999
I mean, I'm certainly blown away every single day.
338
00:41:53,425.999 --> 00:42:00,136.002
The, just the level of inbound inquiries that we've got, um, is, is, to use the term mind-blowing again.
339
00:42:00,175.999 --> 00:42:07,936.001
Um, you know, we're only in the marketplace, but so much, and so you have to r-recognize and realize that, hey, you have this whole digital persona and image you have to keep up, and that's...
340
00:42:08,465.992 --> 00:42:19,915.991
That to me has been far more powerful this last fiscal year in terms of leaning into that and leveraging the tool, uh, to, to, to use what Matt said, uh, has been completely, um, defining for our business.
341
00:42:19,925.999 --> 00:42:38,315.991
And, um, and I think to kinda circle back what, what Matt's commentary was on, on leveraging the tool, right? I think as you're considering SEO and GEO, you know, there, there are considerations to make there, but whatever it is that you decide to do, make sure you're using the right tool to emphasize the overall impact that you're looking to get.
342
00:42:38,315.998 --> 00:42:49,142.001
And I think that we-we've leaned into that GEO piece Through the, you know, the, the, the, um, the maximization of our new website and, and the storytelling and all the, all the other things that we're doing.
343
00:42:49,142.001 --> 00:42:50,522.001
So it's working for us.
344
00:42:50,552.001 --> 00:43:02,502.001
Um, you know, if there's a takeaway for a small business, then I think that's a huge one, is if you're gonna, if you're gonna If you're curious about AI and how it's going to be impactful for me from a business generation perspective, use the tool that's gonna get you there.
345
00:43:03,62.001 --> 00:43:03,312.001
Yeah.
346
00:43:03,312.001 --> 00:43:08,142
I mean, so this is one of those things, like if you haven't updated your website in 10 years, that's a good place to start.
347
00:43:08,142 --> 00:43:08,802.001
Now's, now's the time.
348
00:43:09,002 --> 00:43:10,052
Right? Now's the time.
349
00:43:10,62.001 --> 00:43:12,332.001
Like here's, here's one thing I pulled out of Entrepreneur Magazine.
350
00:43:12,332.001 --> 00:43:15,242.001
46% of businesses are invisible to AI.
351
00:43:15,961.999 --> 00:43:21,612.002
So that's a- Geez that's a pretty startling number, and those are all the small businesses who probably haven't updated their website in 10 years.
352
00:43:21,972.001 --> 00:43:31,822
You know, the companies that win with the, win in the next phase of search will be those that create original expert-driven content that both customers and AI systems trust.
353
00:43:31,832.002 --> 00:43:40,71.999
So this is about validity and, you know, legitimacy, and it, that, that's ultimately what they're look- they're looking for, and that's what they're judging nowadays.
354
00:43:40,81.998 --> 00:43:42,81.998
Is it not? Yeah, absolutely.
355
00:43:42,452.001 --> 00:43:46,241.996
And they, and they're judging it, you know, the I, I, I'll just say this last thing.
356
00:43:46,521.996 --> 00:43:54,11.996
Search engine optimization seemed like a very flat way to judge the depth qu- and quality of content.
357
00:43:54,12.002 --> 00:43:56,631.996
It was based on an algorithm, which is all fine and good.
358
00:43:56,661.996 --> 00:43:58,521.997
So is a lot of the GEO stuff.
359
00:43:58,531.999 --> 00:44:05,342.002
However, the way that the bots are, the chatbots are looking at it is much more multidimensional.
360
00:44:05,661.996 --> 00:44:18,352.002
So the way that they aggregate content and come up with their results and their scoring and their findings, um, and recommendations is much more similar to how a friend would do that work for you.
361
00:44:18,421.998 --> 00:44:35,590
Or so, you know, Lewis, back to what you are saying, back to what, uh, Matt pointed out, I mean, the tool is the tool, but The way that it is being used is incredibly powerful for small businesses, um, 46% of whom are invisible to AI.
362
00:44:35,600.001 --> 00:44:41,900
Like, there's, there is an opportunity here to say, uh, for us on the mar- on the marketing side, we're saying like, "Huh, finally.
363
00:44:41,930.001 --> 00:44:42,600
Okay, cool.
364
00:44:42,780.001 --> 00:44:43,940.001
We're doing all this stuff.
365
00:44:43,940.001 --> 00:44:45,130.001
It's better sourced now.
366
00:44:45,140 --> 00:44:46,330.001
We're getting better feedback.
367
00:44:46,330.001 --> 00:44:47,520.001
We're getting more traffic."
368
00:44:47,740 --> 00:44:49,060
And this isn't just for CIC.
369
00:44:49,060 --> 00:44:51,500.001
We see this across the board with clients who have new websites.
370
00:44:51,740 --> 00:44:55,560.001
Like, it is time, um, and it works better.
371
00:44:55,620 --> 00:44:57,700
So let's do this.
372
00:44:59,690.001 --> 00:45:00,829.999
You know, sticking- Hard sell.
373
00:45:01,919.999 --> 00:45:02,960.001
That's not a hard sell, believe me.
374
00:45:03,319.998 --> 00:45:04,299.999
The proof is in the pudding.
375
00:45:05,440.001 --> 00:45:11,599.998
Um, you know, sticking with sort of, you know, the small business impact, Matt, you know, question for you.
376
00:45:12,449.998 --> 00:45:17,779.999
You know, we, we think about AI and how it's becoming so transformative for small businesses.
377
00:45:18,89.999 --> 00:45:54,466.001
When you think about, you know, a small business's workforce and gearing up and prepping for, um, the use of AI or s- or it's already in the business, if you will, and you're teaching your workforce on how to leverage it, what are some tips you can give to some of these small businesses that are grappling with the notion that A- AI is either going to take over their business or it's gonna be so, uh, transformative that, um, it's gonna eliminate workforce, et cetera, et cetera? I think, I think again, I think that all, all of the concerns, some of the concerns you just mentioned, they're all true at once.
378
00:45:55,86.001 --> 00:46:14,976.001
I think that if you are an individual, uh, administrative worker in a business, you should be thinking about your own competitiveness in the marketplace, just like you probably should have for the rema-- you know, for the rest of your career.
379
00:46:14,976.001 --> 00:46:29,466.001
You shouldn't only be thinking about being competitive, uh, w-when, when there's not such an existential threat to your job Although it makes it easier not to, but you have to be think about, you know, think about what makes you more competitive.
380
00:46:29,466.001 --> 00:46:40,136
So certainly getting on the curve and, and, and understanding how to help you help yourself do your job better with AI, I think is, is very important.
381
00:46:40,165.998 --> 00:46:48,16.001
Ultimately, you're gonna be judged on your output, but you also have to be, um, very conscious of what the alternative is.
382
00:46:48,45.999 --> 00:47:15,46.003
When I used to work in big corporate and I used to work in a department that provided, uh, global technology infrastructure services to the rest of our, uh, corporate offices around the world, we would always look at ourselves internally and say, "What do we cost? And what does the competitive alternative in the market look like if we were to be outsourced?" And we would benchmark ourselves against that.
383
00:47:15,56.001 --> 00:47:27,765.997
And I believe that entrepreneurs and, uh, individual workers should sort of do the same thing, which is they should think about their own opportunity and risk as a function of what the replacement cost is.
384
00:47:28,175.996 --> 00:47:37,676.001
And AI, uh, is both a magnifier of an individual's potential, but it's also representing an increased risk of them being replaced.
385
00:47:38,25.995 --> 00:47:41,625.999
So I think you should think about things on, on both sides.
386
00:47:41,626.001 --> 00:47:51,478.001
If you were a workforce, member of a workforce today and you were worried about your job being eliminated You probably are worrying about the right thing.
387
00:47:51,728.001 --> 00:47:56,748.001
You should be thinking about that, and some of these questions are things that people don't have answers to.
388
00:47:57,398 --> 00:48:18,948.001
If you are a business owner, I would urge you to be thinking ultimately about the same things as always: how do I create a good product or service that gets me the revenue and profitability that I want? I would be using AI to augment what you do in areas where you need it.
389
00:48:19,238 --> 00:48:31,768
I think the wrong question for small business people to ask is, "How do I use AI to make my business better?" I think that's sort of putting the cart before the horse.
390
00:48:31,858 --> 00:48:55,317.999
I think what you wanna do is understand what you do, understand what your roles are, understand where your gaps are, understand if there's opportunities for revenue or savings that maybe you haven't been able to achieve before, and then map in an informed way where this new technology could help you.
391
00:48:55,737.998 --> 00:49:13,938.001
I think people need to remember, and it, it's sort of hard to say, but it's the truth, you know, the objective of a business is not to employ people The objective of a business is to achieve a business and or a financial outcome.
392
00:49:14,358.001 --> 00:49:24,638
And so as an entrepreneur, you're gonna do what it takes to create that product or service at the lowest cost structure in the best way for your clients.
393
00:49:24,648 --> 00:49:40,958.001
So as a business, I'm gonna say, "How do I use AI to fill in some of the subject matter expertise that I don't have? Uh, I'm gonna use AI to help me generate revenue where I didn't have access to that revenue source before.
394
00:49:41,488 --> 00:49:52,557.999
I'm gonna ask AI, how do I keep my cost structure the same and do more? I'm gonna ask AI how I reduce my cost structure and do the same."
395
00:49:53,138.002 --> 00:49:56,748.001
Um, I'm gonna, I'm gonna look at all of those things.
396
00:49:56,767.999 --> 00:50:05,598
I think there is enormous uncertainty, and for the right reasons, about the workforce impact of all these things.
397
00:50:06,98.001 --> 00:50:08,588
I think, you know, when we look at, um...
398
00:50:08,798.001 --> 00:50:15,908.002
And, and, and tell me, Louis, if I'm, if I'm sort of deviating into a, a second, you know, area, and if I should continue or not.
399
00:50:16,17.999 --> 00:50:18,618.001
Yeah, please do- Um 'cause you're, you're hitting all the marks.
400
00:50:19,308 --> 00:50:19,338
Okay.
401
00:50:19,517.999 --> 00:50:29,678
So, you know, so I think th- this question of is AI gonna take all our jobs is sort of not a nuanced enough question.
402
00:50:30,217.996 --> 00:50:47,428.001
I think that we're learning that AI is probably gonna take jobs where we didn't think it would, like, uh, uh, h- you know, higher, uh, echelon professions like legal and, and medical, right? I see a potential impact.
403
00:50:48,47.998 --> 00:51:06,594.001
We, we obviously are-- I think we're gonna see an impact on, uh, on things like trucking Um, we, we are seeing opportunities that seem to be free of AI risk in the trades and things like that.
404
00:51:07,14.001 --> 00:51:16,274.001
We are clearly seeing AI impact, uh, more junior professionals in technology and financial services.
405
00:51:16,304 --> 00:51:20,884
I think, I think that's a real risk, you know, for, for those individuals.
406
00:51:21,384 --> 00:51:31,234.002
But ultimately, the question of what is the workforce gonna look like in five, ten, fifteen years with AI in the mix, nobody knows.
407
00:51:31,514.001 --> 00:51:34,484.001
I do think all things are gonna happen at once.
408
00:51:34,944.001 --> 00:51:37,914.001
I think there will be some new jobs.
409
00:51:38,214.002 --> 00:51:43,754.001
I think in some segments, uh, things could be very, very bad for people.
410
00:51:44,433.999 --> 00:51:58,563.998
I think that, um, you know, if you, if you look at the Industrial Revolution when, you know, the fear was that these machines are gonna take everybody's jobs and ultimately manufacturing employment, you know, grew higher than it ever was.
411
00:51:58,564 --> 00:52:14,683.997
Well, that's true, but that took decades, and in the meantime, there was a lot of carnage in the labor market, and that is something that is, is very, is very, very unknown.
412
00:52:15,403.996 --> 00:52:28,204
I think that when you look at business owners, whether it's small or big business, if you have an opportunity to save costs, whether it's on labor or other processes, you're gonna take advantage of it.
413
00:52:28,233.998 --> 00:52:41,662.001
Now, where we are with AI today, we see stories in the news of companies going all in on AI, you know, laying off- Not unsizeable percentages of their workforce.
414
00:52:41,662.001 --> 00:52:48,332.001
Of course, again, the messaging really never is that they're laying off people for AI.
415
00:52:48,362.001 --> 00:52:50,122.001
Maybe it's true, maybe it's not.
416
00:52:50,582.001 --> 00:52:56,532.002
But in some cases, we see overreactions, right? These cycles, they're very, they're very hype-driven.
417
00:52:56,762 --> 00:52:58,002
They're very news article.
418
00:52:58,002 --> 00:52:59,402
They're very clickbait-driven.
419
00:52:59,402 --> 00:53:04,092
So we see companies that maybe are, like Amazon, uh, Web Services even.
420
00:53:04,102 --> 00:53:17,802.002
Amazon Web Services, which is, you know, I think the most massive, uh, hyperscaler and in-cloud infrastructure provider, um, had a big layoff, which I think was under the name of AI.
421
00:53:17,802.002 --> 00:53:19,141.998
It went a little too far.
422
00:53:19,141.999 --> 00:53:23,72.001
But they had outages because of it and actually started hiring back people.
423
00:53:23,312.001 --> 00:53:27,181.999
We're sort of seeing the same thing here as we saw when outsourcing became the big range.
424
00:53:27,222.001 --> 00:53:28,542
Let's outsource everything.
425
00:53:28,882.001 --> 00:53:31,652.001
But then you spend more money and more cost managing that.
426
00:53:32,121.999 --> 00:53:40,11.999
So all of these things are true at once, and, you know, where it will all converge, I think is, is very unclear.
427
00:53:40,62.001 --> 00:53:48,142.002
All we can worry about is what's right for our business, and all a consumer or an employee can worry about is how they can best be prepared.
428
00:53:48,152.002 --> 00:53:54,531.999
And the only way I know how to be prepared today is, is to try to educate myself, you know, on AI and where it's going.
429
00:53:54,542.001 --> 00:54:04,382.002
That being said, if I had a kid leaving high school and going to college, I would feel less confident about any advice I could give them than I ever would have before.
430
00:54:05,211.999 --> 00:54:06,701.996
Yeah, no, all great points.
431
00:54:06,731.996 --> 00:54:17,261.998
Um, Matt, uh, you know, the last comment in terms of, you know, if you're, if you're an employee, you know, learning how to use it, I, I think resonates with me, um, because I hear it all the time.
432
00:54:17,292.001 --> 00:54:23,952
I hear it, um, you know, from, from our own workforce in terms of, you know, how we get comfortable with AI, and I hear it from small business owners as well.
433
00:54:24,512.002 --> 00:54:32,462.002
You know, it's not just about asking Chat or Claude or these, these AI, um, uh, bots, you know, a question.
434
00:54:32,921.998 --> 00:54:35,772
A lot of it is the prompt and making sure you're asking the right question.
435
00:54:36,211.999 --> 00:55:26,234
Um, what can you say to, you know, employees who are learning and, and trying to leverage AI and trying to get comfortable with it in terms of making sure you're asking the right questions and, and putting something out there that is going to get you the respon- response you're looking for? I, I think there's kind of like two personas that we're talking about here and, and, and they're kind of extremes, right? So the one persona is the individual who is, um, uh, hasn't tried the technology yet, doesn't know where to start, and is sort of averse to, y- you know, u- using the technology either for fear of technology, you know, like has been around since long before AI or, or maybe they just feel that they don't know, you know, what some of the opportunities are.
436
00:55:26,234 --> 00:55:30,403.999
For that individual, um, you, you just have to try something.
437
00:55:30,434 --> 00:55:34,704
Talk to it, type to it, pretend it's a human being, have a conversation.
438
00:55:35,123.998 --> 00:55:39,283.998
Um, you know, j- just, just dip your toes in the water.
439
00:55:39,384.001 --> 00:55:46,454.002
The, the barrier to entry to actually start using AI to learning it is literally zero.
440
00:55:46,494.001 --> 00:55:47,943.999
It's like less than zero.
441
00:55:47,954 --> 00:55:50,534.001
You go on, you have a free account, you do it.
442
00:55:50,564.001 --> 00:56:03,264.002
But I think you're probably asking more about the sort of administrative person, the office worker, the worker who, uh, has that computer in front of them all the time and wants to make best use of AI.
443
00:56:05,713.999 --> 00:56:06,433.999
Yeah, spot on.
444
00:56:06,433.999 --> 00:56:14,703.997
I think that, yeah, I think that, I think that y- y- and, and Joe, just re- repeat the question one more time 'cause I just wanna- Yeah be as focused as I can.
445
00:56:14,804.003 --> 00:56:32,203.998
Um, you know, when you think about using AI in terms of a certain question you might have, there's one thing to just ask the question, but it, there's an, it's the other- Got it to prompt the proper question- Okay to give, to get you a, a real and true response that's meaningful.
446
00:56:33,224.001 --> 00:56:33,584.002
Got it.
447
00:56:34,54.003 --> 00:57:00,423.996
So, so in, in any interaction today, whether it's with a human being, uh, or whether it's with Google if you're doing research, or whether you're talking to a home improvement contractor or your doctor, the richness and the quality and the amount of information and thoughtfulness in the question is directly proportional to the quality of the answer.
448
00:57:01,323.999 --> 00:57:01,413.999
Always.
449
00:57:01,424.005 --> 00:57:06,204.002
Let's look at a very basic example of, uh, a user calling the help desk.
450
00:57:07,193.994 --> 00:57:10,753.994
A bad example of a help desk request is, "Hello, help desk.
451
00:57:10,783.991 --> 00:57:12,513.995
My computer doesn't work."
452
00:57:13,794 --> 00:57:22,533.998
What does that mean, right? And, and, and so, you know, that, that becomes a, a rabbit hole figuring out what that means.
453
00:57:23,173.99 --> 00:57:28,443.998
A good example of a help desk request is, "I have booted up my Windows computer.
454
00:57:28,893.991 --> 00:57:30,663.999
I'm in Microsoft Excel.
455
00:57:31,213.999 --> 00:57:36,923.994
I'm trying to accomplish XYZ, and the error message I am getting is as follows.
456
00:57:37,743.997 --> 00:57:46,574.001
In turn, what I have tried to solve the problem is A, B, and C, and none of them worked, but here's what the output was I got from them."
457
00:57:47,772.001 --> 00:57:53,622.001
When you go to AI, the equivalent of that with an AI chatbot is context.
458
00:57:54,332.001 --> 00:57:56,82.001
Context is everything.
459
00:57:56,812.001 --> 00:58:11,232.001
The more background, the more information of a problem, the more information of your intent and expectations around a solution and how you want that answer, the better the answer will be.
460
00:58:12,212.001 --> 00:58:16,792.001
There are a couple of different techniques for achieving that.
461
00:58:17,672 --> 00:58:24,962
Most people, in my opinion, tend not to think about all of that upfront.
462
00:58:25,672 --> 00:58:27,052
So they will sort of...
463
00:58:27,81.999 --> 00:58:34,432
You can say to the AI and have an iterative conversation around, uh, "Hey, AI, here's what I'm trying to accomplish."
464
00:58:34,432.002 --> 00:58:36,42.001
The AI says, "Okay, tell me more."
465
00:58:36,262 --> 00:58:37,182
You tell it more.
466
00:58:37,381.999 --> 00:58:39,761.999
You say, "I want this type of information."
467
00:58:39,802.001 --> 00:58:42,492.001
"Okay, what else?" "I want this type of output.
468
00:58:42,752.001 --> 00:58:44,251.998
I want it in this format.
469
00:58:44,431.998 --> 00:58:46,102
I want you to look at this data.
470
00:58:46,401.999 --> 00:58:47,991.999
I'm trying to accomplish this.
471
00:58:48,782 --> 00:58:50,212
It's for this purpose."
472
00:58:50,232.001 --> 00:59:02,41.999
You have sort of an interactive, iterative conversation like you would with a human being, and sort of more information comes out, uh, over time so the AI can be more well-informed to solve the problem.
473
00:59:02,41.999 --> 00:59:21,951.997
And one of the best techniques that I recommend to people when they're using that approach is to say to the AI, "Given what you understand that I am trying to do, ask me questions that I should give you answers to or think about in order to help you help me solve this problem better."
474
00:59:23,188.001 --> 00:59:26,78.001
So that's a, that's a powerful way to get a high-quality answer.
475
00:59:26,828.001 --> 00:59:40,708.001
If you, on the other hand, are more of a, uh, sort of what I call a systems thinker or you like to think about everything up front, some of the best AI prompts out there are actually some of the longest prompts.
476
00:59:41,298 --> 00:59:46,558.001
They look like, uh, uh, product request documents or RFPs.
477
00:59:47,198.001 --> 00:59:48,908
It says, "Here's what I want.
478
00:59:49,178.001 --> 00:59:50,758.001
Here's what I'm trying to accomplish.
479
00:59:50,948.001 --> 00:59:52,238.001
Here are my constraints.
480
00:59:52,558 --> 00:59:55,118
Here are the formats I want you to communicate me in.
481
00:59:55,718 --> 00:59:59,917.999
Here is the, um, the things I want you to exclude.
482
01:00:00,148 --> 01:00:01,698
Here are the things I want you to focus on."
483
01:00:01,708.001 --> 01:00:08,138.002
You give it all that at once and then get one big answer that, that you can iterate on.
484
01:00:08,148 --> 01:00:11,598
But I think the most important thing is give the context.
485
01:00:12,477.998 --> 01:00:18,788.001
And this is sort of happening a little more automatically to an extent.
486
01:00:18,848.001 --> 01:00:41,927.996
If you work today with Claude or Google Gemini or ChatGPT and you have the memory settings turned on, which in general are turned on by default, the AI will learn more about you over time, and it will be able to fill in the gaps so that you don't have to repeat those things in, in every prompt.
487
01:00:41,927.999 --> 01:01:02,484.001
But a quality prompt gets you a quality result, and don't forget to ask the AI for its help in the questions you should think about in order to get the best answers from it So going back to the beginning of, of computers, I'm thinking of the old adage, garbage in, garbage out, right? It's still applicable today.
488
01:01:02,934.001 --> 01:01:05,404.001
This is what, you know, prompt engineering is all about.
489
01:01:05,404.001 --> 01:01:10,164
I mean, it's like you gotta put the right stuff in if you wanna expect the best stuff to come out.
490
01:01:10,164 --> 01:01:13,734
Is that, is that still the case? It's always the case.
491
01:01:13,814.001 --> 01:01:14,214.001
Always.
492
01:01:14,214.001 --> 01:01:14,564
Yeah.
493
01:01:14,874 --> 01:01:15,224
Yeah.
494
01:01:15,784 --> 01:01:18,824.001
Um, one final question on, uh, on AI for you, Matt.
495
01:01:18,824.001 --> 01:01:23,354
So let's talk a minute a bit about data privacy, and particularly as it relates to AI.
496
01:01:23,354 --> 01:01:30,704
So I always advise people who work with AI agents, I said, um, anonymize anything that they put into the, uh, into the program.
497
01:01:31,24.002 --> 01:01:45,153.999
There was an article recently in the last few days about, uh, OpenAI, which is the parent company of ChatGPT, saying that they won't retain business data, and I think this was in response to either Anthropic, one of the other companies who, who were, who were retaining business data.
498
01:01:45,523.998 --> 01:01:59,573.997
So the question I, I have is, can we trust this to be true? Like, are they act- Like, can I actually give it sensitive information, or should I still be more, you know, circumspect about that? I, I think this question is also, is also not new.
499
01:02:00,214 --> 01:02:01,994
Uh, and it's not new with AI.
500
01:02:02,63.999 --> 01:02:06,33.999
I think that you have to decide who to trust.
501
01:02:07,284 --> 01:02:16,134.003
You have to rely on reputation, whether it's reputation of the companies building the systems or the reputation of the systems themselves.
502
01:02:16,993.999 --> 01:02:30,284.002
You have to consider what data is absolutely private and sacrosanct to you, or what your legal or ethical or statutory responsibility might be in terms of sharing that data.
503
01:02:31,263.996 --> 01:02:45,342.001
If you think about today, we all have data on the cloud We all rely on the terms of service and the privacy policies of those companies to maintain the trust that we put in them.
504
01:02:46,182.001 --> 01:02:52,872
And sometimes that trust is violated directly because the company operates outside its terms of service.
505
01:02:53,362 --> 01:02:56,582.001
Sometimes that trust is violated because there's an attack.
506
01:02:57,452 --> 01:03:08,822.001
Um, often people don't take, you know, good care in sharing that data, and they are creating, uh, vulnerabilities for themselves by sharing the wrong, you know, the wrong things.
507
01:03:08,822.001 --> 01:03:09,481.999
But ultimately...
508
01:03:09,482.001 --> 01:03:22,422.001
And, and on top of that, by the way, um, you know, the, the data that you put into these systems is often in the domain of the legal system in terms of, uh, in terms of warrants and court orders and things like that.
509
01:03:23,252.001 --> 01:03:27,262
Given that, there are, you know, some new risks.
510
01:03:27,662.001 --> 01:03:33,381.999
Um, data that you put into AI is a little more than just data.
511
01:03:33,502.001 --> 01:03:37,822.001
It becomes rich information that can be synthesized into other things.
512
01:03:37,831.999 --> 01:03:42,651.999
So the ability for that data to be misused and abused is greater than ever.
513
01:03:43,511.996 --> 01:04:10,821.996
Um, I think you have to look at all the data you put into AI and say, "What is the relative risk of this data?" If I'm a therapist or I'm a, uh, a health provider, uh, I should be fired if I put my patient's PII or health data or, uh, other non-anonymized, non-aggregated personal data into the AI.
514
01:04:10,852 --> 01:04:17,861.999
Ultimately, there will be point solutions that have privacy and trust relationships that allow that data to go there.
515
01:04:17,862.002 --> 01:04:19,711.997
Those things are being actively worked on.
516
01:04:20,401.995 --> 01:04:24,351.997
Um, so you wanna be very, very careful about that.
517
01:04:24,761.996 --> 01:04:32,791.996
There is a fear that I hear people talk about, which is that suddenly if you put your data into AI, that it will be available in the public domain and the internet.
518
01:04:33,91.998 --> 01:04:34,472
This is really not true.
519
01:04:34,881.999 --> 01:04:59,894.001
Um, you do, in I think all cases, want to go into the options in the various tools and turn off this, this option that allows them to take your data and use it in training for future models Um, there are options if you are incredibly privacy-focused and more solutions coming out where you can do AI processing offline.
520
01:05:00,304.001 --> 01:05:07,304.001
I know that Apple and Google and all the big AI companies are working on AI models.
521
01:05:07,314 --> 01:05:13,364.001
Sl- they're, uh, uh, small language models that can work on device, and so the data never leaves the device.
522
01:05:14,124.001 --> 01:05:17,374.001
That gives you a, a degree of privacy as well.
523
01:05:17,894 --> 01:05:29,384
But my suspicion is that things will continue as normal, which is that people will always opt for convenience over security and privacy.
524
01:05:29,384.001 --> 01:05:33,554
This was, is, and I believe will always be the case.
525
01:05:33,954.001 --> 01:05:46,3.998
Um, it's why banks have some of the worst password security requirements of any institutions, because people value convenience over security.
526
01:05:46,004 --> 01:05:52,84.001
So it comes down to the individual, uh, you know, making the right choices for themselves.
527
01:05:52,123.999 --> 01:06:12,194
One more thing, and that is that, uh, I do think people should have an extra level of caution today when working with the big AI providers because there has been enough evidence that there's, you know, sometimes some nonsense going on.
528
01:06:12,574 --> 01:06:14,3.999
This is a moving target.
529
01:06:14,53.996 --> 01:06:15,214
Things are evolving.
530
01:06:15,434.001 --> 01:06:18,994
The legislation is a work in progress.
531
01:06:19,14.003 --> 01:06:20,803.997
There's not a lot of court precedent.
532
01:06:21,443.995 --> 01:06:32,793.995
I was reading about some precedent, uh, the other week, where individuals putting data into, into AI could actually waive attorney-client privilege.
533
01:06:33,334.002 --> 01:06:36,303.997
Uh, so, so that needs to be thought about.
534
01:06:36,813.998 --> 01:06:38,863.996
But this is, this is a moving target.
535
01:06:38,904.002 --> 01:06:53,154.001
I would think, though, today, that so much of what you're trying to use AI to accomplish could be done without providing that, you know, super sensitive, super private data to the, to the AI.
536
01:06:53,184.001 --> 01:06:55,154.002
But the basics are always the same.
537
01:06:55,413.996 --> 01:07:01,953.997
When you leave your house, you lock the door, you shut the windows, you don't leave the keys all around.
538
01:07:02,224.002 --> 01:07:03,793.991
It's, it's the same thing here.
539
01:07:03,793.994 --> 01:07:19,602.001
Most of the attacks are gonna be done because you as an individual are sloppy, not because there's some sophisticated attack Yeah, it's kinda like, um, you know, the old, uh, the way that it would, way it would work in the future if you wanna remain anonymous about something.
540
01:07:19,602.001 --> 01:07:23,62.001
You'd say, "Oh, I have a friend who is thinking about doing this, that, or the other thing."
541
01:07:23,332.001 --> 01:07:24,592.001
So I guess the same is true here.
542
01:07:24,592.001 --> 01:07:27,712.001
I mean, m- maintaining some anonymity is probably not a bad idea.
543
01:07:28,312.001 --> 01:07:37,792
You know, so, uh, is, would, would you agree that's probably a good, a good, for at least for the time being, is a good way to move forward? Well, the anonymity is in, is in the data you provided.
544
01:07:37,792 --> 01:07:51,531.999
Obviously, the AI systems know who you are, and- Right, but if I say it's for a client- and they will compute- or something, they might not associate it with me, but, you know, it's like that, that's kinda what I'm, what I'm getting at is, is, you know, I don't put my name in there- I- you know, or my Social Security number or anything.
545
01:07:51,602.001 --> 01:08:02,576.001
Yeah, no, I, and, uh, I think, I think regardless of who you say it's for Is, I think, less important than the content of the data you provide.
546
01:08:02,616.001 --> 01:08:10,006
So you, you don't wanna provide data that even in isolation could be connected to somebody.
547
01:08:10,326.001 --> 01:08:10,446.001
Mm.
548
01:08:10,496 --> 01:08:14,496
I-- There's nothing wrong with putting real data in many cases.
549
01:08:14,496.001 --> 01:08:17,666.001
It's just that you don't want that data connected to somebody.
550
01:08:17,686 --> 01:08:25,256
So saying that it's for a client, you know, it's like, it's like going to your wife and saying, "Hey, I have this friend," but she knows it's really you.
551
01:08:25,556.001 --> 01:08:25,796.001
Right.
552
01:08:26,746.001 --> 01:08:30,546
Right? I think, I think it's the same, it's the same thing.
553
01:08:30,556.001 --> 01:08:34,525.999
The AI knows probably when you're, you're full of it.
554
01:08:34,626 --> 01:08:39,956
Ultimately, it's the content that you're going to want to make sure is...
555
01:08:40,76.001 --> 01:08:54,726
Now, that being said, there's also, you know, uh, temporary chats and things in AI that are sort of like a incognito browser session where the extent to which the session remembers what you're doing, you know, that lifespan is, is limited.
556
01:08:54,935.998 --> 01:08:59,416.001
But ultimately, you just have to think very carefully, um, about what you're putting in.
557
01:08:59,435.998 --> 01:09:12,346.002
There will be and, and there are, you know, Joe and Lewis, solutions, right? You're gonna see AIs that are dedicated to healthcare and run in HIPAA-compliant environments.
558
01:09:12,346.002 --> 01:09:29,066
You're gonna see data that is for legal professionals that, uh, are, are more, are less restrictive in the data that you put into them because in the back end, their contractual obligation is to maintain controls that are consistent with certain legal standards.
559
01:09:29,066 --> 01:09:32,235.999
But on the consumer side, really none of that exists.
560
01:09:32,255.998 --> 01:09:39,26.003
You're just relying on the terms of service of the companies you're working with, and you have to make a determination about trust.
561
01:09:39,386.003 --> 01:09:55,286.002
Apple, for example, um, you know, and there's, there's a good degree about cynicism, there's a good degree of cynicism here, but I think Apple has done over time a pretty good job of establishing them for, themselves as a privacy first provider.
562
01:09:55,286.002 --> 01:10:05,426.002
For example, if you turn on certain features, you know, your iCloud backups and your iMessage traffic is not only encrypted when it's moving, but it's encrypted at rest.
563
01:10:05,426.003 --> 01:10:19,856.001
So even if, uh, you know, what was the case when Apple got a, a court order to turn over some iPhone, you know, backup to the FBI, they couldn't even do it because even they don't have the encryption keys.
564
01:10:20,456.002 --> 01:10:29,606.001
And I think Apple is continuing down that road where they're looking to do on-device processing more and more where they don't see your data at all.
565
01:10:29,835.992 --> 01:10:35,105.997
And then if the data is in the cloud, uh, they can't see it because there's certain encryption.
566
01:10:35,115.991 --> 01:10:47,386.005
So then you as a consumer say, "Okay, I choose to trust Apple or not," or, "I choose to trust Google because, you know, I believe that they're doing things that are consistent with the way I want my data to be preserved."
567
01:10:47,635.995 --> 01:10:50,75.996
So it's all about those mechanisms and about trust.
568
01:10:51,380.001 --> 01:10:53,310.001
You know, this is, uh, this is really good stuff.
569
01:10:53,320.001 --> 01:11:02,310
I mean, we could, we could spend the whole day talking about this, and I know everybody, uh, we're, we're, we're just about out of time for today, so we're gonna have to, uh, we're gonna have to shift gears here.
570
01:11:02,310 --> 01:11:06,150.001
But Matt, this has been really, really insightful, really, uh, really helpful.
571
01:11:06,550.001 --> 01:11:10,490.001
I really appreciate you, uh, you taking the time to, uh, to spend it with us here.
572
01:11:10,510 --> 01:11:13,640.001
Anybody have any final thoughts? I think we need a part two to this discussion.
573
01:11:13,640.001 --> 01:11:16,270
Yeah, we do, 'cause there's so much more we could talk about.
574
01:11:16,270 --> 01:11:18,580
I mean, this is- Agreed this is, this is really good.
575
01:11:18,580 --> 01:11:18,960.001
I mean...
576
01:11:19,680 --> 01:11:21,560.001
All right, Matt, you're, you're officially invited back.
577
01:11:22,890.001 --> 01:11:26,100.001
Hey, hey, Joe, can I, can I make a closing comment? Yeah, go ahead.
578
01:11:27,229.999 --> 01:11:36,280
I wanna, I wanna say, I think as, I think as people think about these things, I urge people to not think about things in binaries.
579
01:11:36,310.001 --> 01:12:03,980.001
Is, you know, is AI good? Is AI bad? Can you trust AI, or can you not trust AI? Do I use AI, or do I not use AI? This is happening, and the question is, how do you use it to your advantage, and how do you use it to enhance your business, to enhance your life, if that's how you look at it? So, you know, so this, this is happening like the internet happened, like mobile happened.
580
01:12:03,990.002 --> 01:12:06,999.998
It's happening like, you know, technology moves forward.
581
01:12:07,469.997 --> 01:12:18,510.002
And I think if you adopt a pragmatic point of view that way and look at things less as binaries, ultimately it makes the choices easier for you as a consumer and a business owner.
582
01:12:19,889.996 --> 01:12:20,379.996
Terrific.
583
01:12:20,410 --> 01:12:21,79.997
Good way to end.
584
01:12:21,79.998 --> 01:12:22,980.001
So again, Matt, thank you very much.
585
01:12:22,990.002 --> 01:12:29,300.003
Uh, oftentimes our con- our conversations about small business intelligence don't end when our podcast episode is complete.
586
01:12:29,719.997 --> 01:12:34,149.997
There's still more to say about the things we discussed and the things we weren't able to get to, and that's why we're gonna have Matt back.
587
01:12:34,509.998 --> 01:12:42,999.996
This is why in addition to this podcast, CIC produces additional informational resources for our small business listeners, which you can find on our website at ciclending.com.
588
01:12:43,369.999 --> 01:12:49,880.001
Be sure to join us next time as we once again discuss the strategic intelligence that will enhance the value of your small business.
589
01:12:49,880.001 --> 01:12:50,719.996
I'm Joe McCaffrey.
590
01:12:50,799.999 --> 01:12:51,539.999
I'm Luis Silva.
591
01:12:51,539.999 --> 01:12:52,520.002
I'm Matthew Krieger.
592
01:12:52,520.002 --> 01:12:54,229.997
I'm Luke Scott Thank you for tuning in.