WEBVTT 1 00:00:10.240 --> 00:00:17.040 Brady Hugins: Hi everyone, this is Brady Hugans. It's July 3rd. This is another Mayor Mayor presentation. 2 00:00:17.230 --> 00:00:22.320 Brady Hugins: Happy Friday. Thank you for joining. We may have some people joining in. 3 00:00:22.830 --> 00:00:24.179 Brady Hugins: During the call. 4 00:00:24.360 --> 00:00:28.170 Brady Hugins: So, and if you're viewing this on recording, happy that you're here too. 5 00:00:28.710 --> 00:00:34.659 Brady Hugins: I've been putting together some finishing touches on our… Data sovereignty course. 6 00:00:34.760 --> 00:00:40.799 Brady Hugins: And then, rounding out some of the information with some practical tools. Last week, we covered… 7 00:00:41.070 --> 00:00:49.650 Brady Hugins: Getting started with the first 15 minutes of ecosystem and tech development with Cloud Code. So, code builder. 8 00:00:50.200 --> 00:00:54.930 Brady Hugins: What are some of the basic prompts? As well as some of the… 9 00:00:55.110 --> 00:01:01.889 Brady Hugins: initial steps and setups. That can be intimidating, it's sometimes the first initial steps, so we put together a… 10 00:01:02.570 --> 00:01:09.809 Brady Hugins: 15-20 minute overview on… How to take your first 15 minutes in the new tech world. 11 00:01:10.180 --> 00:01:21.409 Brady Hugins: So today, we're going to be covering practical AI use, what is happening today. We're in the middle of 2026, so this is a formidable time to be learning about AI. 12 00:01:21.670 --> 00:01:23.679 Brady Hugins: Especially amongst a lot of the changes. 13 00:01:24.450 --> 00:01:41.670 Brady Hugins: my background comes from enterprise and many different Fortune 500 companies, that have used data in various ways, whether it's through their own email use and marketing, or their own 14 00:01:42.790 --> 00:01:56.999 Brady Hugins: Software ecosystems, so I have a number of different instances in which AI has been used, and I've seen use in many different types of styles of companies. 15 00:01:57.310 --> 00:02:09.339 Brady Hugins: So, I'm covering over 5 different types today, and what we can expect to see in the few… next few years from AI, or what… what we can expect to see things going. 16 00:02:13.800 --> 00:02:22.369 Brady Hugins: So, there's often two ways that people… I guess it could be three, but oftentimes, two different areas in which people 17 00:02:22.760 --> 00:02:29.369 Brady Hugins: position AI. They… they… they think of it either as either a toy, so something that is… 18 00:02:29.730 --> 00:02:41.380 Brady Hugins: A funny trick or some light tool, or they believe it will change everything and do everything and do more than it can present itself to be. 19 00:02:42.480 --> 00:02:46.770 Brady Hugins: When we're taking our practical steps, we're definitely looking at 20 00:02:46.940 --> 00:02:54.310 Brady Hugins: potentials of technology, but also practical potentials. So, what does that mean today? What does that mean in the next weeks, months? 21 00:02:55.640 --> 00:03:01.979 Brady Hugins: Even though it may have big impact over the years, we're still working in timelines that are very human. 22 00:03:03.900 --> 00:03:15.719 Brady Hugins: So, oftentimes, people are selling a perspective based on either what they are selling, so if it's NVIDIA, they're usually selling hardware and chips, so tokenization and… 23 00:03:15.950 --> 00:03:23.289 Brady Hugins: And then if they're… you know, OpenAI or Anthropic, they're looking at 24 00:03:23.450 --> 00:03:29.950 Brady Hugins: selling their services and connecting with their models, so there's various different ways. 25 00:03:30.080 --> 00:03:31.050 Brady Hugins: that… 26 00:03:31.300 --> 00:03:41.400 Brady Hugins: companies sell their perspective in AI and position themselves, so it's good to have your own independent view, understanding, as well as your own independent forms of 27 00:03:41.730 --> 00:03:46.160 Brady Hugins: AI that you can run locally, which means runs on your local computer. 28 00:03:46.770 --> 00:03:51.050 Brady Hugins: So… This is an operator's version today. 29 00:03:51.180 --> 00:03:59.410 Brady Hugins: So, what's useful now, what breaks, and then what's coming. So, I make these… Presentations as practical as possible. 30 00:03:59.800 --> 00:04:04.589 Brady Hugins: So that people can go about and start building the systems that they need. 31 00:04:06.130 --> 00:04:15.409 Brady Hugins: So, we don't have any demos today, although we have many different demos that we've run in the past, I use AI and agents 32 00:04:15.700 --> 00:04:22.610 Brady Hugins: Quite a bit internally, and through developing for web platforms, so externally, too, so… 33 00:04:23.300 --> 00:04:26.990 Brady Hugins: I have a wide breadth of experience when it comes to that. 34 00:04:29.390 --> 00:04:44.469 Brady Hugins: So AI is not just one thing, it's referencing parallel processing, which is a lot like… it's how the brain operates, where it's a nodal structure, where you have a neuron, and then another neuron, and then they work independently, but together. 35 00:04:44.740 --> 00:04:53.819 Brady Hugins: So, there's different capabilities that that kind of processing is capable of. Large language models are one of the primary ones. 36 00:04:55.520 --> 00:05:02.729 Brady Hugins: So we're gonna go over 5 different types of capabilities, and then what those would look like in the near-term future. So… 37 00:05:02.910 --> 00:05:07.660 Brady Hugins: Confu… it's sometimes confusing, but this is here to give some clarity on that. 38 00:05:08.290 --> 00:05:12.959 Brady Hugins: So, let's start with the first one. Usually, what most people are aware of. 39 00:05:13.650 --> 00:05:20.230 Brady Hugins: And what they've had most experience with, which is either chat or… Claude. 40 00:05:20.850 --> 00:05:22.730 Brady Hugins: So it's text in, text out. 41 00:05:23.360 --> 00:05:27.490 Brady Hugins: Large language model, referred to as LLMs. 42 00:05:27.740 --> 00:05:32.839 Brady Hugins: So it's good at drafting, summarizing, if you provided a large context of information. 43 00:05:32.950 --> 00:05:44.690 Brady Hugins: It can pull the structure, clean up any of the messy text, infer what you mean by it, and then answer and develop a document or a system, several documents around that. 44 00:05:46.200 --> 00:05:48.589 Brady Hugins: So, it breaks at… 45 00:05:49.090 --> 00:06:01.840 Brady Hugins: hallucinations, so AI fills in the gaps, so there's factual and contextual information that needs to be referred to, reviewed, researched, and then checked. 46 00:06:02.300 --> 00:06:04.869 Brady Hugins: Oftentimes, so there's versioning? 47 00:06:05.670 --> 00:06:09.940 Brady Hugins: That's important to review and revise in documentation. 48 00:06:10.190 --> 00:06:19.409 Brady Hugins: So, sometimes context is, as you would go and ask someone to create a document for you, you would need to have a revision 49 00:06:19.630 --> 00:06:25.300 Brady Hugins: And to go and read and review, because there might be certain points that are missed, or certain points of… 50 00:06:25.950 --> 00:06:31.629 Brady Hugins: It could be an expert saying, no, that that is not in my expertise. 51 00:06:31.880 --> 00:06:42.410 Brady Hugins: In my experience, so there might be a double check, like, it could be a lawyer, it could be as simple as a plumber going over the specs, or an engineer checking out specifications on a new build. 52 00:06:44.350 --> 00:06:48.210 Brady Hugins: So… it… 53 00:06:48.880 --> 00:06:58.770 Brady Hugins: doesn't always have contextual information, it only has the contextual information that it's been built on, and then also what you give it in order to answer it, so… 54 00:06:59.910 --> 00:07:06.690 Brady Hugins: Having… it's really good at… Giving summaries of information, transcripts. 55 00:07:07.450 --> 00:07:09.699 Brady Hugins: Action items that are pulled from it. 56 00:07:10.260 --> 00:07:17.039 Brady Hugins: And some… some of those are the real use of it. It can be… used to… 57 00:07:18.610 --> 00:07:31.220 Brady Hugins: gather large amounts of information together, and then consolidate them into a readable format. Edible… editable, that you would change it, but, definitely… 58 00:07:31.400 --> 00:07:32.830 Brady Hugins: Very good, Ed. 59 00:07:33.640 --> 00:07:39.150 Brady Hugins: Providing that work. We use one for transcribers here internally. There's a lot of people that has 60 00:07:39.870 --> 00:07:51.679 Brady Hugins: Transcribers, so it records my voice, it's listening to my voice, and then it's… it's… recording… the transcription… 61 00:07:52.000 --> 00:07:57.650 Brady Hugins: And then it's referring to the transcription when it's building out and summarizing the information. 62 00:08:00.110 --> 00:08:03.879 Brady Hugins: Okay, coding agents, builders. So, these are… 63 00:08:05.400 --> 00:08:10.230 Brady Hugins: Sometimes people are not very well aware of these, and these are what's… 64 00:08:10.340 --> 00:08:23.960 Brady Hugins: creating a lot of the splash in the AI and software development economy that's happening right now in 2026. And 2025, and in the introduction, I think in late 2024, 65 00:08:24.420 --> 00:08:32.740 Brady Hugins: So, it's a… what coding agents and builders are, is they're AI that writes and edits software. 66 00:08:33.309 --> 00:08:46.829 Brady Hugins: So, the reason why it does it so well is because it works in its native format with data development and coding, so it can refer to and look at information very quickly, rather than 67 00:08:47.280 --> 00:09:02.429 Brady Hugins: pull up an image document, read the image document after conversion, so this is very quick. It's able to read the data and information natively, and then create that information. Plus, it understands the context and the… 68 00:09:02.600 --> 00:09:05.319 Brady Hugins: Hard rules around software development. 69 00:09:06.810 --> 00:09:12.200 Brady Hugins: So it's good for understanding how to build tools, Fixing bugs. 70 00:09:12.400 --> 00:09:16.609 Brady Hugins: Wiring APIs, which are data connectors between different databases. 71 00:09:17.250 --> 00:09:29.420 Brady Hugins: And what people often refer to as the boring 80%, or what's really a lot of the behind the scenes. When we run a Zoom webinar and meeting, there's quite a bit of code that's being… that has been built. 72 00:09:29.800 --> 00:09:36.250 Brady Hugins: In order for this… these webinars to go about and be presented in the way that… 73 00:09:36.490 --> 00:09:45.439 Brady Hugins: they are now. And so, there's quite a bit of coding to make sure that that happens with, like, latency and video and audio and all the… 74 00:09:45.820 --> 00:09:49.940 Brady Hugins: gadgets and gizmos we get for recording and working with these, so… 75 00:09:51.450 --> 00:10:03.360 Brady Hugins: Let's see, it breaks, around big, ambiguous systems, so… It's… Needs specifications, so doing some… 76 00:10:03.590 --> 00:10:09.440 Brady Hugins: Pre-work, or pre-planning, and having an idea of hard deliverables that you want. 77 00:10:09.560 --> 00:10:27.199 Brady Hugins: And driving towards revisions and designs that move towards those deliverables are really important, so having human clarity about direction and about the project, because it can go off in different areas and develop in different ways without the guidance, so… 78 00:10:27.410 --> 00:10:29.130 Brady Hugins: Always want to be reading. 79 00:10:29.530 --> 00:10:31.330 Brady Hugins: What it's building, even though 80 00:10:31.800 --> 00:10:45.880 Brady Hugins: it's too challenging for humans to write at that level. So, it's used… it's basically watching and supervising as it's being built, and then guiding the code builder as it's building. 81 00:10:47.310 --> 00:10:49.799 Brady Hugins: There's quite a bit of planning, 82 00:10:50.120 --> 00:10:58.240 Brady Hugins: that needs to be done and is encouraged. Real use, we use code builders very often. I'm one of the heaviest, probably, users of… 83 00:10:59.900 --> 00:11:03.799 Brady Hugins: Claude Code, I'm probably in the top tier. 84 00:11:03.910 --> 00:11:05.470 Brady Hugins: I've… I've… 85 00:11:05.600 --> 00:11:13.069 Brady Hugins: spent billions of tokens, which is a lot of tokens. I code almost every day, so… 86 00:11:13.960 --> 00:11:20.900 Brady Hugins: It's a, I use a lot of it in order to build the ecosystems, the dashboards, the automations. 87 00:11:21.380 --> 00:11:27.500 Brady Hugins: And anything… Setting up the systems in order to make it easier for human review. 88 00:11:28.010 --> 00:11:29.150 Brady Hugins: So… 89 00:11:29.390 --> 00:11:37.509 Brady Hugins: I'm in my second year of business, so I've built a lot of the foundational elements, which is, like, the administration and the 90 00:11:37.710 --> 00:11:48.680 Brady Hugins: Corporation, and kind of… and getting up… set up, the business up and going, and this next year has been a lot more focusing on tech development, data development, messaging. 91 00:11:48.800 --> 00:11:54.229 Brady Hugins: And then production, and making sure that we're hitting certain production goals at certain points. 92 00:11:54.790 --> 00:12:02.099 Brady Hugins: In order to do that, there's quite a bit of automation and behind-the-scenes work that needs to be done and at least looked at. 93 00:12:02.390 --> 00:12:12.519 Brady Hugins: in order for all the information to flow well. Like, when an RSVP comes in for an event, or an email comes in asking about something. 94 00:12:13.540 --> 00:12:16.629 Brady Hugins: Sometimes we want a system to be able to either listen. 95 00:12:16.890 --> 00:12:18.880 Brady Hugins: To listen and to respond back? 96 00:12:19.380 --> 00:12:35.710 Brady Hugins: As part of our review or dashboarding. So there's some various reasons why you would want to have someone look at your messages coming in rather than just write back. A lot of people expect it's like a chat, AI is like a chat, and it's not always set up like that. 97 00:12:35.850 --> 00:12:36.830 Brady Hugins: It's… 98 00:12:36.970 --> 00:12:45.169 Brady Hugins: That's a more advanced component to be able to chat, refer to information, then send back, which a lot of people expect, and which is a great use of 99 00:12:45.780 --> 00:12:51.740 Brady Hugins: AI… But coding agents are different, they dive into the code directly, so… 100 00:12:52.540 --> 00:13:04.859 Brady Hugins: it's very adaptable to people, so for new coders, I would recommend them look into these coding agents, especially because they're going to be using… have the ability to ask questions, and… 101 00:13:06.130 --> 00:13:12.209 Brady Hugins: Write in plain English. I want to do this, I want to create this type of dashboard, this type of page. 102 00:13:14.800 --> 00:13:16.270 Brady Hugins: Alright, number 3. 103 00:13:17.430 --> 00:13:19.020 Brady Hugins: vision, reading. 104 00:13:19.180 --> 00:13:26.469 Brady Hugins: So there's understanding, reading, and then generating. So there's certain steps… the AI can do multiple steps, but… 105 00:13:27.420 --> 00:13:34.870 Brady Hugins: It's often designed in different processes or phases, so this one is about reading images and PDFs. 106 00:13:35.400 --> 00:13:41.170 Brady Hugins: So it's good at reading emails, receipts, forms… describing… 107 00:13:41.330 --> 00:13:47.000 Brady Hugins: Images, grading images, pulling, data from screenshots. 108 00:13:47.650 --> 00:13:52.549 Brady Hugins: Reading accounting receipts or, like, a sign-up sheet. 109 00:13:52.710 --> 00:13:55.520 Brady Hugins: It's good at doing those things. 110 00:13:55.910 --> 00:14:07.910 Brady Hugins: even task lists. I write down notes and tasks, and I can take a picture of my notebook, and then upload many of those for a little bit of clarification, but it can understand 111 00:14:08.120 --> 00:14:11.770 Brady Hugins: what… my needs are, so… 112 00:14:12.360 --> 00:14:27.739 Brady Hugins: I can go off and write in the park, or go off for a walk, or sit by the river, and go and take notes about what I need to get done, and then go and take a picture and send it to my AI-enabled ecosystem, and then it will go and 113 00:14:28.030 --> 00:14:32.840 Brady Hugins: Start on the work, or position some of the information so it's… 114 00:14:34.030 --> 00:14:37.470 Brady Hugins: it's more available to the system, right? It knows what… 115 00:14:37.650 --> 00:14:44.530 Brady Hugins: date… what information… what… what info I'm sharing when I'm sharing the information. 116 00:14:44.950 --> 00:14:47.810 Brady Hugins: So it's very contextual, it can grab some of those contacts. 117 00:14:49.580 --> 00:14:51.480 Brady Hugins: Real use, auto grading. 118 00:14:51.720 --> 00:14:57.699 Brady Hugins: We use it for products and looking at images. We're in the early phase of this, so it's… 119 00:14:58.390 --> 00:15:09.359 Brady Hugins: A lot in development is really testing and getting everything up and running at the beginning, go through certain products, and then making sure it's working with that product, and then spreading out to the entire ecosystem through 120 00:15:09.580 --> 00:15:19.830 Brady Hugins: a phasing structure, so… We… I've been developing a content system so that all the content is scored. 121 00:15:20.480 --> 00:15:32.150 Brady Hugins: Appropriately, whether it's art, or a need for an art, like, we might need, like, an email header, or an email footer, or it might be, like, 122 00:15:32.410 --> 00:15:35.480 Brady Hugins: A web thumbnail for the event page. 123 00:15:36.650 --> 00:15:44.519 Brady Hugins: So, there's different needs, and in order to… Assess whether you're… 124 00:15:44.830 --> 00:15:48.310 Brady Hugins: Audit… your web page is audited, and… 125 00:15:48.530 --> 00:15:56.890 Brady Hugins: Is performing that way, you would need to run a scan of either that image or the whole page in order to grade 126 00:15:57.180 --> 00:16:00.400 Brady Hugins: And give some sort of feedback grade. 127 00:16:00.630 --> 00:16:11.709 Brady Hugins: for the system, and so we use AI in order to read and audit, oftentimes. Even your own information or your own deliverables, to make sure they're going out 128 00:16:12.930 --> 00:16:16.939 Brady Hugins: In the appropriate way, with the appropriate context to the right people. 129 00:16:18.950 --> 00:16:24.330 Brady Hugins: And then fail-safes and redos for anything that's fallen through. 130 00:16:24.770 --> 00:16:28.799 Brady Hugins: Backlogging, it's called, in software development. 131 00:16:35.120 --> 00:16:39.559 Brady Hugins: Okay… As far as transcriptions, recording to text. 132 00:16:40.790 --> 00:16:51.429 Brady Hugins: So, it's good at near-human transcription. It's not perfect, but it can be a 98, 98, 99%. I have a little bit of an Arizona mumble, and so… 133 00:16:52.300 --> 00:16:57.560 Brady Hugins: Let's see, that's… my T's… And, 134 00:16:59.560 --> 00:17:04.539 Brady Hugins: you know, B's and R's can be sometimes hard to pick up. 135 00:17:09.800 --> 00:17:10.869 Brady Hugins: So… 136 00:17:12.380 --> 00:17:20.160 Brady Hugins: it's… it's good. You can use it reliably in the sense that it's going to get the gist, but there may be some small editing 137 00:17:20.940 --> 00:17:34.040 Brady Hugins: I've noticed that the names don't always come out right, so even though you may pronounce a name or spell a name, is that the transcription almost always puts its own interpretation in, so… 138 00:17:35.530 --> 00:17:37.020 Brady Hugins: You know, 139 00:17:38.000 --> 00:17:56.370 Brady Hugins: so it's not always… not always spelled right. Like, SoulShine is one of our products, and it's spelled S-O-L-S-H-I-N-E, and it often comes out in transcriptions as S-O-U-L-S-H-I-N-E, or something else entirely. Because it's a unique branded name. 140 00:17:56.950 --> 00:17:59.120 Brady Hugins: It's not that heard that often. 141 00:17:59.460 --> 00:18:05.030 Brady Hugins: SoulShine, I don't know if there's any other… Products that match that. 142 00:18:07.430 --> 00:18:14.420 Brady Hugins: So it breaks are heavy crosstalk, so if there's multiple people talking on an audio, it may not be able to discern and tell that. 143 00:18:14.990 --> 00:18:22.449 Brady Hugins: Rare names, like I was saying before, and emotional nuances, or… yeah, moods and modes. 144 00:18:22.890 --> 00:18:34.360 Brady Hugins: it can pick up the transcription, but not necessarily the tone, or the intention, you know, the aggression, or the softness, you know, it's… 145 00:18:34.890 --> 00:18:39.770 Brady Hugins: I can't really pick up those, so… a voice note. 146 00:18:40.580 --> 00:18:42.370 Brady Hugins: And a voice analysis. 147 00:18:42.880 --> 00:18:51.929 Brady Hugins: is helpful in order to dynamically create a richer profile. You'll find that it'll pick up more than you expect, in some ways, too. 148 00:18:52.210 --> 00:19:02.720 Brady Hugins: As well. So, it definitely covers a lot of the… Databasing need in order to… Cover the information that 149 00:19:02.850 --> 00:19:05.109 Brady Hugins: One would expect for a meeting. 150 00:19:09.570 --> 00:19:15.579 Brady Hugins: Image generation. This is… has been made popular also through video generation as well. 151 00:19:16.390 --> 00:19:21.869 Brady Hugins: Which is really just a… Layering and animation style of multiple images. 152 00:19:24.530 --> 00:19:41.680 Brady Hugins: So, Midjourney, we use Midjourney. There's Runway, Descript, many different media editing and generating tools. Midjourney does editing as well, where you can upload an image or use one of the images there, and then cut out certain areas, and then have it 153 00:19:43.100 --> 00:19:47.040 Brady Hugins: Repopulate that area that you've cut out. 154 00:19:49.020 --> 00:19:58.999 Brady Hugins: It helps keep consistency, but also editability, in, like, particular editability physically on the, image. 155 00:20:01.090 --> 00:20:10.970 Brady Hugins: So there's… It's challenging to create, like, exact logos. Text and image can be… Jarbled and confusing. 156 00:20:11.450 --> 00:20:15.539 Brady Hugins: So… and sometimes consistency across it, so… 157 00:20:15.710 --> 00:20:26.190 Brady Hugins: Photoreal, like, we don't use photoreal AI generation, we only use, like, a watercolor or different stylizations that it's clear that it's stylized. The intent… 158 00:20:26.790 --> 00:20:36.800 Brady Hugins: Is to have very… real pictures and videos, but also stylized with different AI imagery, or… mood… 159 00:20:36.910 --> 00:20:40.820 Brady Hugins: You know, it's there to create the mood and enhance the experience. 160 00:20:42.730 --> 00:20:50.150 Brady Hugins: So, oftentimes, all sorts of media uses multiple forms of media, including video and audio. 161 00:20:50.470 --> 00:20:52.769 Brady Hugins: In real… in real life, and… 162 00:20:53.400 --> 00:21:05.570 Brady Hugins: what I recommend people is to use a hybrid model, which is humans and AI, as according to preference, either light or heavy use, but also have humans in the mix. 163 00:21:05.720 --> 00:21:09.110 Brady Hugins: Humans there for editing and review. 164 00:21:11.120 --> 00:21:12.579 Brady Hugins: As a part of it. 165 00:21:12.710 --> 00:21:14.930 Brady Hugins: So, match the capability of the job. 166 00:21:16.780 --> 00:21:17.870 Brady Hugins: So… 167 00:21:20.110 --> 00:21:29.369 Brady Hugins: Most of the AI failed us stories are over-investing into a particular sort of technology that doesn't fit the need. 168 00:21:31.210 --> 00:21:40.099 Brady Hugins: And I see this all the time at enterprise level. It's usually more lean at a smaller company, but that's not always the case. 169 00:21:40.330 --> 00:21:44.520 Brady Hugins: And then under… understanding how it all works. There's quite a bit of… 170 00:21:44.670 --> 00:21:49.499 Brady Hugins: Database needs that are needed in order to use several different types of 171 00:21:50.370 --> 00:22:03.550 Brady Hugins: tools and AI tools, so a lot of the infrastructure development and the admin development that I've been creating has been a lot of the behind-the-scenes data work and databasing in order to utilize these tools more appropriately and fully. 172 00:22:03.740 --> 00:22:07.250 Brady Hugins: Takes more time, but the quality is higher over time. 173 00:22:11.010 --> 00:22:18.790 Brady Hugins: So how do businesses actually use AI today? What are practical ways that they use this, and will continue to use this in the future? 174 00:22:19.400 --> 00:22:32.210 Brady Hugins: Draft and review. So, AI drafts are a great way to either respond back to prospects, to write drafts and documentation for operational guidelines or business guidelines, sales and marketing meetings. 175 00:22:32.580 --> 00:22:45.339 Brady Hugins: Copy, replies, summaries. It's the lowest risk. There's all sorts of information that's important for people to have on file. We record these meetings, and I send it out to people that's appropriate for my team. 176 00:22:46.290 --> 00:22:50.990 Brady Hugins: And then you get instant time back, where you don't have to work on those things. 177 00:22:51.620 --> 00:22:58.640 Brady Hugins: You can focus more on the creative, or on the review process, about what type of Art. 178 00:22:58.880 --> 00:23:02.210 Brady Hugins: What type of subjects are important for you to create? 179 00:23:02.690 --> 00:23:05.830 Brady Hugins: What direction of the business is important to go? 180 00:23:06.190 --> 00:23:11.420 Brady Hugins: Do you need better fulfillment on a certain product or marketing campaign? 181 00:23:14.880 --> 00:23:22.719 Brady Hugins: Extracting and routing. This is some of the behind-the-scenes infrastructure work I was talking about, where you pull a structure from emails, forms, calls. 182 00:23:23.560 --> 00:23:27.420 Brady Hugins: And you send it somewhere, Where it's contextually… 183 00:23:27.760 --> 00:23:34.109 Brady Hugins: Moving either a process or information, so it flags certain information so that 184 00:23:34.410 --> 00:23:37.740 Brady Hugins: You're either aware of something happening. 185 00:23:38.010 --> 00:23:43.070 Brady Hugins: Like, if someone fills out a form, sometimes it's really important to get back on a very quick basis. 186 00:23:43.360 --> 00:23:47.999 Brady Hugins: One, to confirm that they've signed up the form. Two. 187 00:23:48.030 --> 00:24:07.599 Brady Hugins: there's obvious… obviously, usually a need that they're trying to solve, or at least gather information about when they fill out the form. So, responding back to them at an appropriate time, and then putting them on a follow-up list is oftentimes a high use of AI. So, sales, marketing. 188 00:24:08.600 --> 00:24:09.679 Brady Hugins: New leads. 189 00:24:11.510 --> 00:24:19.149 Brady Hugins: and building tools. Building the tools is what we've done internally. Sometimes it's faster or better to… 190 00:24:19.670 --> 00:24:27.469 Brady Hugins: To build the tool, rather than rely on an external tool, or to… 191 00:24:28.510 --> 00:24:32.890 Brady Hugins: wait for another tool to be updated, so it's… 192 00:24:33.030 --> 00:24:38.769 Brady Hugins: It can be a compounded effect, whereas if you can go and build your own tool and use it, you're definitely… 193 00:24:39.000 --> 00:24:44.629 Brady Hugins: Developing at a much… Faster pace, where you can adapt and change. 194 00:24:45.220 --> 00:24:54.169 Brady Hugins: I have things where I'm adding a button here, adding it for different needs, integrating different systems together, upgrading, different… 195 00:24:54.300 --> 00:24:56.830 Brady Hugins: Project or messaging system, so… 196 00:24:58.400 --> 00:25:04.580 Brady Hugins: It's a compounding effect when you build your own tools, if you have the discipline and enjoyment for that. 197 00:25:06.380 --> 00:25:11.340 Brady Hugins: And then there's costs and privacy. So, oftentimes when we rent tools or we use. 198 00:25:14.650 --> 00:25:23.269 Brady Hugins: code builders or these LLMs, there might be some hidden costs, either hardware costs or different software costs. 199 00:25:24.190 --> 00:25:26.999 Brady Hugins: But also, privacy is an issue as well. 200 00:25:27.840 --> 00:25:34.960 Brady Hugins: So, whether you're using an American model, or Chinese model, or whatever open model is available for 201 00:25:35.310 --> 00:25:37.089 Brady Hugins: Your agentic work. 202 00:25:37.800 --> 00:25:41.409 Brady Hugins: You can create it locally, where you can download a model. 203 00:25:42.310 --> 00:25:50.469 Brady Hugins: And then use it for usually lower quality, depending on the hardware and style that you have set up, but usually lower quality, but… 204 00:25:50.810 --> 00:25:57.549 Brady Hugins: higher routine tasks. We have a hybrid system where we can use remote. 205 00:25:57.770 --> 00:26:02.690 Brady Hugins: resources, but we can also use local, and then I greatly shift over 206 00:26:03.430 --> 00:26:06.990 Brady Hugins: Resources to local, and then look over quality checks. 207 00:26:08.030 --> 00:26:09.800 Brady Hugins: Approvals all the way. 208 00:26:09.940 --> 00:26:11.490 Brady Hugins: So, it's… 209 00:26:12.080 --> 00:26:17.699 Brady Hugins: I've been building out the approval systems, and I've been building out the quality systems and the quality checks, so… 210 00:26:19.280 --> 00:26:24.120 Brady Hugins: And you can run quite a bit of private, Agentic work? 211 00:26:24.400 --> 00:26:30.690 Brady Hugins: that maybe… Rudimentary, or easy for some of these… 212 00:26:31.270 --> 00:26:35.840 Brady Hugins: More expensive models that are just… more up-to-date, whereas… 213 00:26:36.290 --> 00:26:43.970 Brady Hugins: Maybe database cleanup, or some financial checks, maybe a simple, agentic model locally would work well. 214 00:26:48.520 --> 00:26:50.090 Brady Hugins: So what's overhyped? 215 00:26:50.560 --> 00:26:54.010 Brady Hugins: Fully autonomous agents running your company. Now… 216 00:26:54.460 --> 00:27:00.159 Brady Hugins: That's not really true, there's… there's… the idea of, like, AI slop is that… 217 00:27:00.650 --> 00:27:10.339 Brady Hugins: it's… there's quite a bit that's unreviewed, that people aren't reviewing, or they aren't editing, or they're just not putting a lot of effort in, and there's quite a bit of AI interesting 218 00:27:11.240 --> 00:27:17.020 Brady Hugins: Either videos or outputs that are being made that are… after… 219 00:27:17.540 --> 00:27:19.940 Brady Hugins: Quite a bit of work has been put into it. 220 00:27:20.480 --> 00:27:28.139 Brady Hugins: So… That's because of the human review and the human understanding of how to use the technology. 221 00:27:28.290 --> 00:27:29.870 Brady Hugins: And what to build out. 222 00:27:31.950 --> 00:27:38.129 Brady Hugins: It's basically AGI, Advanced… I think it's Artificial General Intelligence. 223 00:27:38.670 --> 00:27:47.330 Brady Hugins: So… like humans, They're capable of extremely extraordinary pattern. 224 00:27:48.000 --> 00:27:57.800 Brady Hugins: connection and seeing things, but the judgments can sometimes… well, judgment of which direction, or the quality of work, or… 225 00:27:58.500 --> 00:28:06.749 Brady Hugins: humans can actually catch quite a bit of problems with AI, because they can see when things are 226 00:28:08.270 --> 00:28:09.530 Brady Hugins: Out of order. 227 00:28:09.820 --> 00:28:16.979 Brady Hugins: You know, we were referred to as the Uncanny Valley, which is where a human-like robot 228 00:28:18.220 --> 00:28:29.740 Brady Hugins: When it's, like, a robot that looks like, you know, a fluffy little animal, we're okay with it anthropomorphizing it, but as soon as it looks a little bit like a human, we have this… 229 00:28:29.740 --> 00:28:38.689 Brady Hugins: uncanny valley where we get uncomfortable with something until it becomes very human-like, or that we can have very little… 230 00:28:41.450 --> 00:28:45.909 Brady Hugins: resistance to something that's very similar. So there's definitely that. 231 00:28:46.060 --> 00:28:51.719 Brady Hugins: In some of the… artificial intelligence outputs. 232 00:28:52.580 --> 00:29:01.899 Brady Hugins: that… It's sometimes a little bit of uncomfortableness for some people to watch, or understand. 233 00:29:03.420 --> 00:29:06.270 Brady Hugins: So, one tool to rule them all, some… 234 00:29:07.330 --> 00:29:14.189 Brady Hugins: companies say that this does everything, and it… it… I mean, there's some great tools out there, quite a bit, but there's… 235 00:29:14.430 --> 00:29:18.109 Brady Hugins: Definitely an integration of many different types of tools that can be worked. 236 00:29:18.230 --> 00:29:21.119 Brady Hugins: Together, and there's definitely advancing tools, but… 237 00:29:21.420 --> 00:29:24.010 Brady Hugins: There's no tools that are doing it all. 238 00:29:25.490 --> 00:29:28.399 Brady Hugins: But I see quite a bit of tools working together. 239 00:29:29.130 --> 00:29:30.640 Brady Hugins: Replace your whole team. 240 00:29:31.010 --> 00:29:36.520 Brady Hugins: So, it replaces tasks, not people. People make judgment calls, they… 241 00:29:36.870 --> 00:29:41.329 Brady Hugins: Make phone calls, they read reports, they read emails. 242 00:29:42.790 --> 00:29:44.320 Brady Hugins: They send emails. 243 00:29:44.720 --> 00:29:46.799 Brady Hugins: That doesn't change at all. 244 00:29:46.900 --> 00:29:48.030 Brady Hugins: So… 245 00:29:48.310 --> 00:30:01.100 Brady Hugins: what it does is it replaces quite a bit of the background work, right? It's like writing the report or checking on the report for some of the financial data, or making sure your website's up and going, or the copy is good. 246 00:30:02.700 --> 00:30:04.500 Brady Hugins: Contextual information. 247 00:30:10.330 --> 00:30:23.500 Brady Hugins: So, what does AI look over the next 6 to 18 months? Now, this isn't a direct predictor, but you can see that there's general directions that… where AI is developing. 248 00:30:24.390 --> 00:30:31.059 Brady Hugins: And based on my own… work and analysis, I view these as very likely 249 00:30:31.250 --> 00:30:36.879 Brady Hugins: to be in the next 6 to 18 months, so this is the end of 2027, 2000… 250 00:30:37.310 --> 00:30:40.750 Brady Hugins: 2026 all the way to the end of 2027. 251 00:30:41.750 --> 00:30:48.880 Brady Hugins: Agents that reliably do multi-step work with guardrails and review. This is already true with some of the more advanced 252 00:30:49.700 --> 00:30:55.770 Brady Hugins: Agents that are doing… Different views and running multiple agents at the same time. 253 00:30:55.910 --> 00:30:57.130 Brady Hugins: Where there's, like. 254 00:30:57.330 --> 00:31:03.289 Brady Hugins: An overall agent where you ask questions, and then it'll spin out sub-agents in order to plan. 255 00:31:03.420 --> 00:31:06.029 Brady Hugins: And review what's currently available. 256 00:31:07.160 --> 00:31:09.680 Brady Hugins: In order to get a better contextual picture. 257 00:31:11.970 --> 00:31:17.520 Brady Hugins: So, I'm already running agents with guardrails and review processes already. 258 00:31:18.370 --> 00:31:22.390 Brady Hugins: They can take longer, but that's just based on the hardware that's… 259 00:31:22.560 --> 00:31:25.269 Brady Hugins: Currently available. That will get faster over time. 260 00:31:27.680 --> 00:31:34.279 Brady Hugins: longer memory and context. So, models that hold a whole project, not just a page. So… 261 00:31:34.450 --> 00:31:39.690 Brady Hugins: Having proper context and longer memory… there's two reasons for this. 262 00:31:40.510 --> 00:31:47.390 Brady Hugins: Every time… a AI model looks at 263 00:31:47.560 --> 00:31:58.079 Brady Hugins: the information that you're giving it in order to create an output, it reads the context. So, imagine it going to sleep, and then having very little memory of what happened before. 264 00:31:58.330 --> 00:32:05.669 Brady Hugins: And then waking up and working on something that's a lot of how AI works, where it kind of wipes its brain, and then it starts again. 265 00:32:05.840 --> 00:32:09.380 Brady Hugins: So having the proper context, is important. 266 00:32:09.760 --> 00:32:14.609 Brady Hugins: Longer memory, so you see this in some of the Claude code, the Claude projects. 267 00:32:14.870 --> 00:32:22.370 Brady Hugins: And keeping track of Markdown and documents that it refers to at later times, so you have a better API or MCP access, and… 268 00:32:22.730 --> 00:32:24.429 Brady Hugins: To your own database? 269 00:32:24.760 --> 00:32:26.790 Brady Hugins: So, that's important. 270 00:32:28.810 --> 00:32:32.309 Brady Hugins: And then you're able to do more ecosystem development, where I can see. 271 00:32:32.570 --> 00:32:36.539 Brady Hugins: Many different pages, many different projects when you're working on 272 00:32:37.050 --> 00:32:38.620 Brady Hugins: Some of the more advanced work. 273 00:32:40.890 --> 00:32:47.220 Brady Hugins: Multiple modality by default. Again, this is sometimes what's available to some of the 274 00:32:47.530 --> 00:32:51.989 Brady Hugins: Nicer agents at the moment, and will be more available. 275 00:32:52.290 --> 00:32:57.010 Brady Hugins: Reading texts, images, videos, all together, so… 276 00:32:57.340 --> 00:33:01.209 Brady Hugins: You don't have to have different skills, it's already built in. 277 00:33:02.470 --> 00:33:03.660 Brady Hugins: Or different… 278 00:33:04.540 --> 00:33:16.639 Brady Hugins: Sub-agents, they're already built in for some of the more common use, or you can add on pretty easily more abilities, so multi-modality by default, so you can ask more generalized questions. 279 00:33:18.560 --> 00:33:34.479 Brady Hugins: And costs. Costs keep falling, and more capable local models. I've started seeing this as… isn't necessarily cost… costs keep falling, but hardware costs, as people generally know, is that there's a general fall in depreciation in hardware costs. 280 00:33:34.790 --> 00:33:36.060 Brady Hugins: Every year. 281 00:33:36.630 --> 00:33:42.070 Brady Hugins: The tech gets better and better with chips and… Different upgrades in the systems. 282 00:33:42.360 --> 00:33:52.389 Brady Hugins: So what happens is we're able to run more because of the improved models on our local hard drives and our local servers, and what works locally. 283 00:33:52.660 --> 00:33:59.790 Brady Hugins: We have more complex models that we're able to grab, so we're able to do more different types of work in complexity. 284 00:34:00.230 --> 00:34:05.429 Brady Hugins: With that… And have more… 285 00:34:06.060 --> 00:34:14.989 Brady Hugins: really data sovereignty, what I see is data sovereignty, which is where you're running your own systems, your own data, you see what's going on. 286 00:34:15.350 --> 00:34:23.929 Brady Hugins: And you have a… healthy… Ecosystem. Healthy kingdom. 287 00:34:27.580 --> 00:34:29.439 Brady Hugins: be more and more available. 288 00:34:30.060 --> 00:34:31.799 Brady Hugins: So, what to do now? 289 00:34:32.949 --> 00:34:38.900 Brady Hugins: Build the habit, so… Start by using AI, 290 00:34:40.150 --> 00:34:42.900 Brady Hugins: Exploring different tools, draft and review. 291 00:34:44.370 --> 00:34:51.339 Brady Hugins: in a project or something that you're working on, whether you're building out… I've had friends use it for building out a garage. 292 00:34:52.239 --> 00:35:00.700 Brady Hugins: I've used it for, you know, running for, applications and agricultural work. 293 00:35:03.750 --> 00:35:14.789 Brady Hugins: Also, also get your data clean. Hygiene is always an important process, so making sure that your database is working, and you're… have proper backups in place, and then you have… 294 00:35:14.920 --> 00:35:17.740 Brady Hugins: You know the locations, a lot of the… 295 00:35:18.260 --> 00:35:27.510 Brady Hugins: Texts or images or photos, whatever you have that you're wanting or working with, you want to make sure your data library is generally clean and well located. 296 00:35:28.530 --> 00:35:38.729 Brady Hugins: And then keep a human in the loop. On default, I almost always have an approval process, because I know there's going to be errors, especially at the beginning of testing. 297 00:35:39.210 --> 00:35:46.860 Brady Hugins: So I systemically add in gates and deployments, rather than try to do full builds 298 00:35:47.080 --> 00:35:49.549 Brady Hugins: Where I try to solve everything right at the beginning. 299 00:35:50.890 --> 00:35:56.520 Brady Hugins: I try to solve it system by system, as I have the time and availability. 300 00:36:00.620 --> 00:36:06.940 Brady Hugins: match each job to the right capabilities. So, we talked about 5 today. So, when you have… 301 00:36:07.240 --> 00:36:08.949 Brady Hugins: Your own business needs. 302 00:36:10.150 --> 00:36:18.549 Brady Hugins: Whether you're… You know, operating in a truck every day, or you're… Maybe you work from home. 303 00:36:18.770 --> 00:36:20.490 Brady Hugins: And you're a knowledge worker. 304 00:36:21.490 --> 00:36:23.240 Brady Hugins: We're your salesperson. 305 00:36:23.550 --> 00:36:25.630 Brady Hugins: We're in marketing, and so… 306 00:36:25.820 --> 00:36:35.629 Brady Hugins: know what… basically what needs that you have as a job, and start using some of the tools in order to develop them. You'll be surprised by how much outputs you can create with some of them. 307 00:36:38.930 --> 00:36:41.180 Brady Hugins: So go deeper when you're ready? 308 00:36:41.490 --> 00:36:46.710 Brady Hugins: So… If you are… want some more… 309 00:36:47.410 --> 00:37:03.769 Brady Hugins: We have plenty of free material available, on MirrorMirror, with lots of workflows to look through on our website. We also have our weekly series here, where we're inviting new people, or anyone that wants to review some of the basics. 310 00:37:04.240 --> 00:37:12.219 Brady Hugins: Of data sovereignty and building out their own ecosystem, technical… technology ecosystem. 311 00:37:13.190 --> 00:37:16.430 Brady Hugins: You could be an independent or a business operator. 312 00:37:17.690 --> 00:37:26.259 Brady Hugins: Happy to talk, so you can email me at brady at roseinthegrove.com, or check us out at mirrormirror.rosandthegrove.com. 313 00:37:26.920 --> 00:37:34.310 Brady Hugins: We teach what we build here, so you can rest assured that we've already tried and tested these 314 00:37:35.330 --> 00:37:43.949 Brady Hugins: Directions before we've used them, and made our own mistakes and errors, and then checked in on those and fixed them as we can. 315 00:37:47.260 --> 00:37:53.049 Brady Hugins: Good. Well, thank you for time today, appreciate you being here, and we will… 316 00:37:53.990 --> 00:37:58.479 Brady Hugins: Catch you at the next webinar, hopefully, and thank you. 317 00:37:58.730 --> 00:38:02.119 Brady Hugins: Okay, I'm Brady with Mirror Mirror Systems. Thank you.