Introduction: Putting AI to Work in Your Business — with Jenny Moshea

Jenny Moshea joined EXP Technical for a practical look at AI strategy for small business. The talk covered agentic AI, custom GPTs, real small‑business implementations, and security considerations in tools like Microsoft Copilot. Watch the full webinar on YouTube and read the transcript below.

▶ Watch the full webinar on YouTube:

Key Takeaways

• AI strategy for SMBs
• Agentic AI vs. conversational AI: conversational tools respond to prompts; agentic AI plans and acts toward goals with a human in the loop.
• Start where the pain is: target repetitive, copy‑paste work and the ‘spreadsheet mess in the middle’ to unlock quick wins for small businesses.
• Use paid tiers + guardrails: prefer paid models, disable data sharing, and keep sensitive data out of free tools; Microsoft Copilot in a governed tenant can reduce risk when configured properly.
• Personalize your stack: leverage custom GPTs for focused roles (e.g., AI CFO/COO) and teach voice/tone with your own samples.
• Human/Expert‑in‑the‑Loop matters: keep approvers in workflows; verify sources to avoid hallucinations.
• 90‑day move: create your own virtual assistant GPT and start collaborating daily.
• Local help when you need it: EXP Technical provides IT support in Seattle and guidance on secure AI adoption (see EXP Academy resources).

 

 

Welcome, Agenda, and How to Participate

This section sets the stage for the webinar, introduces the speakers, and explains how attendees can participate and access the recording.

Kelly Paletta: [EXP] Hey, good afternoon everyone. I’m Kelly Paletta, Director of Sales and Marketing here at EXP Technical. Welcome to our webinar.

Today we’re going to be discussing putting AI to work in your business. Our guest today is Jenny Moshea. I’ll introduce her in just a minute.

But beforehand, I have a couple of brief announcements before we get started. Let me see. I’ve got a little presentation that gives us an outline here. So, bear with me here. We’ll jump to this now. If I’ve done it right, you should see an overview here.

Just wanted to give a little bit of an overview and an agenda so people understand or can see where we’re headed here. There is an agenda for today’s discussion.

If you’ve attended these events in the past, you know that sometimes they are very structured and follow—there’s a lot of PowerPoint slides and prepared material. Other times they’re more conversational. I think this one’s going to be a little bit in between.

Jenny has some prepared material and we’ll go back and forth with a lot of Q&A as well. And what that means—the good news to you is that you can participate as well.

If you have questions, you can ask them via the chat or via the Q&A feature in Zoom. They should be at the bottom of your screen here. The difference between the two is, if you submit a question via Q&A, it’s somewhat private. It will just go to myself as the moderator and Jenny. Whereas if you submit a question via chat, everyone in attendance can see it.

We’ve got about 100 people registered for this event, so be careful what you share in the chat. The Q&A is a little bit more private, a little bit more confidential.

One question that comes up every time we host one of these events is, is this event being recorded? Yes, in fact, it is. Usually takes me about a week before I make that available. I like to do a little bit of post-processing of the transcript, but you will get an email from me with links to video and to a transcript.

We’ll post it at EXP Academy, on our website, in our blog, on YouTube. So, it will be available and I’m sure there are going to be those moments that you want to rewind and revisit. And hopefully you want to share it with your friends and your colleagues as well. So, yes, indeed, that will be available to you.

And then our structure is that I’ll introduce Jenny. We’ve got a couple of poll questions lined up. Don’t know if we’ll get to those or not, but Jenny has some material prepared and that’s the heart of our conversation, which is putting AI to work in your business.

So, probably a good time for me to introduce Jenny to you all. Some of you may know her because she works or used to work in the same industry as a lot of our clients. And I even have heard of at least one person in attendance knows both of us somehow. So, there are probably others as well.

But if you don’t know Jenny, she is currently an executive AI adviser for her firm Kinetiq Shift. I think she has some new information about her career path that was just publicly announced an hour ago. She in the past has been an advisor to some corporations in the region that are household names. I will leave them nameless for now.

But where I first started following her was when I realized or when I noticed that she was a CIO at a prominent construction firm here in the Pacific Northwest, one that changes the skyline and the landscape of our cities here. And for that reason, her role as a CIO and in the industry where we have a lot of clients, it seemed like I should be following her on LinkedIn. And so that’s Jenny—I don’t know if you know that, but that’s where I first came in contact with you.

But so, having said all of that, Jenny, welcome to our webinar. Thank you so much for being here.

Jenny Moshea: Oh, it’s a thrill to be here, Kelly. And I know we’ve been talking about this back and forth for several months now, and so it’s fun to actually see it come to life.

EXP: Yeah, I know. It feels like forever, but here we are. So, I wondered, do you have anything else to share? I kind of gave a bullet point list of your bio, and I teased that there’s a new announcement. Do you have any background or any other information that you can—

Jenny Moshea: Well, I continue to, you know, as a former CIO, I get tapped a lot as an adviser, especially in the AI space now for the AEC industry and beyond. And so, actually, I have joined Rapid Canvas as an AI growth advisor. What’s different about Rapid Canvas is it’s an agentic AI platform plus data science team. So we can talk more about that later, but I love, you know, I’ve just joined that partnership. I’ve been collaborating with them for many months and so it’s exciting for me to bring it to life and be able to help support enterprises both above 100 employees and below.

EXP: Awesome. Well, and I know you have some information prepared, but before you do, I wanted to ask a couple of quick poll questions of our audience.

Jenny Moshea: Yes.

EXP: Initially, I wanted to ask a question of our audience members about their individual familiarity and comfort level with AI. There’s four questions there or four levels: whether you use AI daily, regularly, if you consider yourself competent, or this is new.

This kind of goes back to that comic that I shared in the first slide, which, by the way, I got from one of our attendees and I credited it in the slide, but it seems relevant that there are a lot of people these days that are like, ‘Well, I need to be using AI. I feel like I’m behind. There’s this new technology that I don’t know what it can do. And also, I don’t know how I’m going to apply it.’ And that’s going to be part of our conversation today.

So, if I’m looking at the poll answers, I think I can end the poll here now. And I have one other question, then I’ll set you loose, Jenny.

Jenny Moshea: I have a question, too. So, this is good.

EXP: Okay. So, let’s see if I end the poll. I’m going to share results, but from what I see, there’s a broad cross-section. You know, it’s almost a quarter, a quarter, a quarter, but it’s between people that never use AI versus people that use it regularly and often.

Let’s see. I wanted to share one other poll question, and this is where I see that a lot. Those answers reflect a lot of other surveys I’ve seen. So, and you’re all right in there.

Jenny Moshea: Yeah.

EXP: And it’s something that I’m not surprised. I expected that from this audience because I talk to them every day and I hear that when I speak to them.

So, the next question speaks to—I’m asking people to answer on their organization’s adoption of AI. You know, whether it’s something that is fully implemented as an organization “we’re using it every day,” or “we haven’t really started, we’re just exploring.”

I suspect a lot of people in this session are early in their AI journey, and it looks like the results are speaking to that too. I’m going to end the poll and I’ll give you three, two, last chance to respond. One. Okay, I’m going to end and share that.

One more. Nope. Okay, so we’ll share the results here. And what I see is that nearly everybody is in the category of haven’t started yet or just experimenting with small pilots or individual tools.

So, it seems like people in attendance today are early in their journey in AI adoption, and I wanted to get that out of the way so that it might influence our conversation as we go.

I’ve been speaking a lot. Jenny, I know you came to this event with a little bit of information prepared. Perhaps I should yield the stage to you. Do you want to pull up some of the information that you had prepared for our audience here?

Jenny Moshea: A little first though, I’d love to ask if people can somehow respond to this question, but I would love to know—sure, go ahead, just toss in—how many employees you have at your organization. Is it you, one as a solopreneur, or do you have hundreds? I would love to just get some idea of numbers because we can fly at different altitudes throughout this conversation and that will help. So, I don’t know if you’ll see in the chat, maybe we can do chat.

EXP: Yes.

Jenny Moshea: So, I see it in the Q&A and perhaps wherever you want to toss it. So, I see one response that says 114 employees, 25 employees. One person says, ‘I’m the only person in my organization using AI.’ Let’s see.

EXP: This is what I expected.

Jenny Moshea: Okay. A lot—25, under 10, 4, 50, 36, 17. I don’t see—oh, Chad, 500. Yeah, you can see the responses, too. So, there’s a broad cross-section there. I don’t see any that are over a thousand. I don’t see many that are over a hundred, but a lot that are in the 30 or so range.

EXP: Okay. Well, let me—yeah, I think that’s perfect. And that’s what I expected. So, this is great. This is where it’s fun. It’s fun. And there’s a lot you can do to just accelerate with AI.

Why AI Adoption Is Accelerating + Starter Tips

Jenny Moshea shares examples of rapid AI adoption, practical starter tips for using AI securely, and advice on personalizing your AI experience.

EXP: Do we want to jump in, or what were you—I think because I know you had some information prepared that you wanted to share, that kind of opens up people’s minds to things that might be available. I have a long list of questions too that I want to run past, but I don’t want to step over the information that you came here prepared to present.

Jenny Moshea: Well, I have a few fun things I like to share, just to orient people on the same page. We’ve got quite a mix of organizations, so I just want to show—I like this. I have a video, or you know, I think—oh jeez, of course I skipped. But I love this little video that I will share, and hopefully you can hear. So I’m going to play this. It’s 90 seconds. It’s a trailer for a short film that was created entirely by GenAI. Bear with me. It’s worth the 90 seconds.

Echo Hunter is a fully GenAI film built with the cutting-edge tools of Arana Labs. I’m Caven the Kid, director of Echo Hunter and co-founder of Phantom X. We teamed up with Arana Labs to fuse next-gen tech with an amazing cast of SAG actors, pushing the future of filmmaking into now. It’s been too long. Tell me, does a corpse ever consider the purpose of its actions? [Music]

The potential is limitless, is the crazy thing. With this technology and with AI, and with the creative still being able to use the writers and the directors, everybody still gets to work, but the barrier for entry is lowered substantially. And I think that’s really cool. We’re very excited about Echo Hunter, and we’re very excited about the power of Arch AI and its tools and the creators who use them. [Music]

Jenny Moshea: So, the reason I share that—can you all still hear me?

EXP: Yes.

Jenny Moshea: The reason I share this is because it’s a 22-minute short film, and typically a film of that caliber and level would have taken 12 months, and they built this in three months. This was presented on an AI stage recently at a big AI conference, and they announced that if they were to remake it now, a few months later, it would only take four days. This is how fast things are accelerating, and that’s why I think—even between the time, Kelly, that you and I spoke and now actually talking, things are changing fast.

I love that people have jumped into this, and I think it’s—you know, there’s an example of how we’ve seen other trends. If any of you think back to the smartphone in 2007—I don’t know if you all stood in line like I did—but think about how we didn’t really fully understand the power of a smartphone and having a computer in our pocket. Now we spend 28% more of our waking time on this device that helps us run our day.

So I just like to bring us back to—we’ve been through this before. We’ve been through some of these big changes before. Here’s a little chart that I think is fascinating, just seeing how fast AI adoption is accelerating beyond what we’ve seen in the past, going from, you know, remember a landline subscription or moving to the internet and then mobile phones. It’s just going bonkers.

You can see—we’ve all heard the news stories about ChatGPT getting to a million users in five days. This is the pace that I think we’re all feeling. I appreciate folks wanting to catch up.

Just a couple more—

EXP: Right. I’m agreeing with you. I gave a presentation a month ago and started with the same thing about the adoption cycle for AI. It’s like the fastest technology adoption in history. It took 100 years before 100 million people had benefited from the steam engine. Took two months before 100 million people were subscribed to ChatGPT. There were a million in five days. So you’re right—the adoption curve is very, very steep and continues to be for sure.

Jenny Moshea: Yeah. I’ll just throw a couple more things. I think everybody, whether or not you realize it, you’re using AI. You’ve been using AI in the apps you use to do your everyday. I started plopping icons on here and just had to stop because it’s a bounty of apps that we use. At one point, it was the order of my day. I was like, well, I get a Starbucks, I go somewhere in traffic, I make sure I do my Duolingo because I’m keeping up with my kids. So anyway, all these apps that you may be familiar with are powered by AI.

I just like to give this tip—I give free tips all the time—but okay, if you have not done anything with AI, I’m just going to—the number one thing I would say you do, and if you already use a large language model, an LLM, or a ChatGPT, Perplexity, Gemini, you had Midjourney on there, Claude, any of those elements—

EXP: Can I interject right now? I still see the slide in your presentation that has the video. Have you moved on? You’ve talked about—

Jenny Moshea: I have—stop share. My bad.

EXP: I love it when technology does that, right?

Jenny Moshea: I know. All right. Let’s try that again. Are you seeing this? Give it a second to catch up.

EXP: Okay. So, you—

Jenny Moshea: Yes.

EXP: You didn’t see any—do you see these apps?

Jenny Moshea: Did not see those.

EXP: Okay, sorry.

Jenny Moshea: Yeah, my graph and everything. Okay, so these were the apps I was showing and talking about—everyday AI from Duolingo to Gmail to your photos to LinkedIn to Canva to Starbucks. So, but—and you’re still tracking with me now. Starter tips.

EXP: Yes.

Jenny Moshea: Okay. Indeed. So, regardless of what chat you use or model you use, they all have similar capabilities. What I want you to do is—ideally, if you’re using it for business at all, you’re in the paid version. Twenty bucks a month is worth it because you get a little bit more security. Your data doesn’t have to be shared out, or you can turn that off now. Almost all of them, you can turn off the ability to share. So I encourage you to go into your settings. You want to turn off—if you’re doing private things, you want—or anything sensitive—

EXP: Yeah. We’ll just disclaimer that you never want to put anything sensitive. It’d be like posting anything on a public billboard. So, be careful about what you put into ChatGPT.

Jenny Moshea: But as far as an assistant and soundboard for ideas and things, I just want to encourage you to go in and use the personalization. Go into the settings. You can choose your preference, whether or not you share your information with the larger ecosystem in the model.

Another thing is to share your personalization—how do you want your—I use ChatGPT most of the time because it’s a great generalist and I have a lot of general things. If I was doing a lot of writing, I would probably use Claude. If I was doing a lot of research, I would probably use Perplexity. But those are examples that all of these models can be used for different things.

But I encourage you—use the paid version if you can. You can do a lot with the free version, but you won’t get the speed, you won’t get the responsiveness, and you won’t get the full possibilities. So personalize with custom instructions, and then also—highly—to hold the context of whatever you’re saying, I encourage you to create a project, a folder. It’s basically a folder, and you can start chats out of those folders, which holds the context.

For example, when I was working on a project for my brother for his business, I didn’t want it to mix with my questions about college campus tours for my kids’ spring break. So, I create a folder for each kind of category of what I’m talking about. Those are just hot tips.

And just so we all orient, I think it’s important for people to know—you’re probably hearing more in the news about agentic AI agents. There’s a lot of mixed information, but you can think of ChatGPT—and what I think is the starter AI app for most of us—as really conversational intelligence. You’re interacting with it. It responds to you, but it only responds to you if you tell it, if you ask it something, or prompt it.

What’s new about agentic AI is that it’s more broad. It’s an operational intelligence, and you can set the goals of an agent and it will go and plan and act on your behalf, and it can report back.

But I’d love to see if there’s any questions bubbling up yet.

Custom GPTs, Coaching, and Human/Expert-in-the-Loop

Explore the difference between generative and agentic AI, the value of custom GPTs, and why human expertise remains essential in AI workflows.

EXP: I have a question that’s somewhere in between here, because I think this will come up in our conversation. I’ve heard you refer to this, and maybe it’s later in your presentation, but right now you’re making the distinction between generative AI like ChatGPT and these other tools, and agentic AI which can act independently and make decisions.

But you also talk about custom GPTs.

Jenny Moshea: Yes.

EXP: Can you explain that, or am I one slide ahead of you here? Can you speak to that a little bit?

Jenny Moshea: No, you’re right. Custom GPTs—and I showed a little bit about that in here. You can also create your own custom GPT. That’s where I think you can also change the mode if you’re in a large language model like ChatGPT. You can create your own GPT, and it’ll say agent mode sometimes, but that’s not to be confused with creating an agent. We’ll get to that a little bit later.

GPTs are—like, I think of it like a PhD student working for my company in a specific area. For example, a lot of companies, part of their org chart now is an AI CFO or an AI COO, and it’s an agent as part of their organizational chart. You can create these GPT agents. I know there’s two kinds—there’s many agents—but a GPT agent to focus just on that category of information. It’s like a folder, but more interactive. So a folder is like the container, the project is the container, but then the GPT is—you’re training it to just focus on a specific type of information.

For example, we often have a Director of HR at a company, but you don’t ask your Director of HR to also manage the finance team. So you want agents to be very discreet and separate, and create them for their bespoke process and function. That’s a little bit about GPTs. We could literally talk about this all day.

EXP: Right. I love that analogy. I’ve heard something similar—I think even Dave Nelson, who was a former presenter in this series, said something along the lines of ChatGPT or just AI in general is like your super smart intern. I’ve heard other people say it’s like an intern that has a PhD in everything. It’s very good at surfacing information. It’s read every Washington state law on HR or on accounting, but it doesn’t have any context. It doesn’t know your corporate culture. It doesn’t know your voice, your tone. It doesn’t know what’s important to you or to your business.

So, it doesn’t really know—it’s kind of unfocused in that way. But it has access to and has read practically everything that’s available, right? That’s why creating your own GPT narrows that focus, and you can teach it your tone of writing, your voice. I love Claude for that reason. All of these—you can upload your own writing, give it examples of your writing, give it examples of your LinkedIn posts or your social media, and help it learn your style.

So that’s also a thing I recommend people do. If you’re struggling—wherever—what can it do, or even after today’s conversation and you’re like, ‘I don’t know where to start,’ just ask your ChatGPT to help you. Like, I recently did something way outside of my comfort zone to help a family member, and I didn’t know where to start, so I was like, ‘How would I do this?’ And ChatGPT gave me specific steps.

And then, if I don’t know how to write a good prompt—sometimes I struggle with good prompts, because garbage in, garbage out. If you give it nothing—like, we’re so used to a command like in a Google search that it’s hard for certain generations of people, namely I would say, to ask a really good prompt and clarify and give it more information. So I will often use ChatGPT to help me write a better prompt. Then I’ll say, ‘Well, can you make it even better? Can you make it better or interview me for this process?’ Have it interview you and ask you questions to clarify and draw out what you’re trying to accomplish.

So use it as a sidekick. I pretend like it’s my little assistant sitting next to me, and I’m just using it as a soundboard to create something different and new.

EXP: Forgive me for interjecting, and I don’t want to get you off track from your prepared material. That’s a—there are a couple of pearls of wisdom in there. One is that using AI to help you improve your prompts, and you can do that with competing AI tools as well. You can take a chat—you’ve submitted a prompt into ChatGPT, you’re not getting the results that you want. You could go into Microsoft Copilot and say, ‘Why am I not getting the answers I want? How can I improve this prompt?’

And the other thing too is that people often—and you’re talking about this—people often look at AI as what it produces, but I find, and it sounds like you’re saying the same thing, I’m often using it as a coach to help me. For example, I recently—I won’t go into the details—but had a fairly complex integration between several different software applications that I use. I had never used Zapier before that, and I wasn’t great at HubSpot and some of the other things that I was using for integration, so I just started from there and said, ‘Hey, this is what I want to do. These are the tools that I have available. What’s the most efficient workflow to get all of these different tools—’ and I had three or four of them communicating together in this fairly complex workflow.

In that case, the AI wasn’t really producing the results. It was—but it had read every user’s manual on HubSpot and Phone Burner and SendSpark and all the other tools that I was using. So it could teach me how to get those things working and playing well together.

And one other thing that I’ll add, because I love this, was the ability to copy and paste my screen share. I would copy and paste screenshots or screen captures when I got stuck. I think this is relatively new to more recent versions of AI, but it can read the text that is just in an image. So it could look at my screen capture from HubSpot and say, ‘Oh, you’re using this version. You don’t have that feature enabled, so you have to work around it this way.’

Jenny Moshea: Exactly. It’s like, again, having a genius coach looking over your shoulder that can tell you, ‘No, click here.’ I did the same thing, but I didn’t even think to use screenshots. That’s brilliant. I just said, ‘I don’t see that option on my screen,’ and it would say, ‘Oh, you must be in a different version of blah blah blah.’

EXP: Oh, I love the screenshot. Especially all you entrepreneurs or solopreneurs or small company or under 20 employees, under five—you can use things like, you know, Kelly mentioned Zapier, that’s a little pricier than—you could use Make or other tools, but ask ChatGPT, ‘How would I do this thing?’ and just talk to it like a friend that you trust for advice and wisdom.

And then, if it just gives you—sometimes, depending on the version you’re on, if you’re not on a paid version, it has a tendency to do the hallucination thing we’ve heard about so much. But the more paid you are, especially for groups that are in enterprise or team versions of these tools, the hallucination gets much reduced. So I would encourage you to leverage that or know it’s possible. There’s a level of companies like that I’m part of are going towards zero hallucination because the data has to be accurate and absolute.

Jenny Moshea: So that leads to, I think, one aspect of a phrase that we hear often, and that is human in the loop or human in the process. One way to check for hallucinations is for AI tools to provide citations, but another—do you want to speak to having a human in the loop and humans involved in the process?

EXP: I think absolutely—the human judgment and the subject matter expert and expertise is critical to using AI effectively. You need to have an approver or reviewer in your workflow. You do that today, right? You just do it with all humans. So, if you’re adding an AI piece of that workflow, that work still needs to be reviewed. Think of it like an intern—wants to do a good job, but sometimes they need guidance and they need to be redirected. So you are still in control, and you’re creating very small areas of focus with the AI in there.

Human in the loop is one thing, but now with other platforms like Rapid Canvas and some of these AI tools and the industry tools like Via AI and Trunk Tools and Trunk Text, it is expert in the loop that’s part of this process. So you’re adding the human business person in the loop plus the data science or expert to know that the data is quality or not, because a lot of companies do not have the luxury of a big technical team, and that’s why companies like EXP exist to help support. Most folks don’t have data scientists on their payroll.

EXP: I’m going to interject with one question because it’s really relevant to what you’re saying too—a question that came up in the Q&A here, and this speaks to expert in the loop. Somebody asked about—I’m summarizing and I’m skimming—but using AI, and frankly I think there’s some legitimacy to this, but using AI as a counselor or for medical advice. What I would say, and maybe you can amplify this, but a couple of things. One, this goes back to what you said about privacy and confidentiality. My understanding is that there are products that are specifically built for that purpose—your little AI psychotherapist that are regulated by HIPAA and have those controls about sensitivity.

But the other thing too is that we—you definitely want—well, I would remain very concerned about my mental health information and entering that into an artificial intelligence chat session if I don’t know where that data is saved, where it’s stored, where it may exist in the future. So, I’m going to shut up in a second here, but I agree—couple of good reasons to have an expert in the loop. One is that with an expert in the room, what happens in that room or is said in that room might stay in that room. And also, we’re still in the early stages of AI, and it might not give the best medical or mental health information as well.

So, I personally think it’s valid. I think there could be valid psychoanalytic applications and psychotherapy and mental health counseling, but I think we want to beware. I’m sorry I’m talking over you and yours.

Jenny Moshea: No, I think we’re both excited to answer this. I think—I will not ever say I’m not an expert in that area of domain expertise. However, from what I think we’re seeing is that yes, it is trained on the academia of that therapy or that area of focus, but it does not have the lived experience that a human therapist has. It does not have the experience of real world. So I would encourage people to make sure they know where their information is going. If it’s in a generic context and you can sanity check it with a professional, then that might be a good blend, but make sure it’s not—it’s a closed system, that you’re not just giving it all away.

EXP: Yeah, I agree, Jenny. I’m going to go to another question from our Q&A here. That’s why we’re here. Let’s do that.

So this speaks to the difference between different products that are available, like a paid version of ChatGPT—and there are levels too—but that versus Microsoft Copilot, right? I wonder if you have some thoughts on security controls in one product versus another. If you don’t, I’m sure—I kind of do, but I want to hear from you if you have any thoughts on that, about which might be more secure or which might be more appropriate for a small business.

Jenny Moshea: Yeah, Copilot is broad, and it’s usually implemented at an enterprise level. So your technical team or people in charge of putting in the tool should be able to put some guardrails on it as to what it has access to. All tools—there’s flavors of different things you would use ChatGPT for versus Copilot, depending on what the subject is. But I would say at least, you know, encouraging people—if you have an enterprise tool like Copilot, I use that for anything business or interactive. If you’re doing something personal—like recently I was trying to figure out the best route for our spring break tour and trying to hit the tours and hotels and kid-friendly—a lot of personal things or fun, but also just as an assistant to think through business topics.

But I would encourage people to stick—you know, I mean, again, you’re going to use more than the average person is using—on average about five different tools to do day-to-day things. So it’s not a silver bullet.

EXP: I tend to agree and affirm that just from my own experience. Again, I’m just a sales guy at EXP and not the expert. But Copilot—the good news and the bad news about Copilot is that it has access to everything in your Microsoft tenant, SharePoint and OneDrive and all of these resources. And as you expertly mentioned, your IT team can put proper access controls around all that information. So in that sense, you can be kind of working in an environment that has controls built into it.

That’s the good news. The bad news is if those controls aren’t in place, you might also be giving your intern access to HR data or—because the bad news is it has access to all of your information in your Microsoft tenant. That’s a good reminder that we still need to be good humans and good citizens of our information and data and files. Even before Copilot, we had to be responsible employees or humans of the company and preserve the company information and intellectual property. So that doesn’t change with AI—it just goes faster. Making sure you have the right governance in place, compliance, and you’ve put all the controls in.

Jenny Moshea: I think also, Kelly, your EXP Academy—the latest training you have with the cybersecurity for business leaders—is super important for people to listen to because, yes, the bad actors are acting badly still. They can do it faster, and we can respond faster also, but it doesn’t go away.

EXP: Thank you so much for mentioning that. Maybe not everyone was in the waiting room and saw the previews, the video that showed before the session started, but I want to give my own plug for that too, just because we feel like this is really important information that we want everyone to have access to. At EXP Technical, one of our values is share knowledge, share success, and we’re really concerned about this genie that’s out of the bottle now and ways that sensitive data can get outside of your control.

We also know that a lot of business leaders are really concerned about that too. So we created a course that’s available for free at EXP Academy, and it’s called AI at Work. It’s practical cybersecurity recommendations for non-technical business leaders, and it speaks to compliance, governance. It talks to ethical concerns like plagiarism and attribution, and it gives you practical steps like how to begin writing policies and a lot of those topics are raised in that course. Also, there’s a journey in that course too—there are resources that you can use to start implementing those things too. It’s not just a course that’s designed to scare the hell out of you and then leave you on your own to fend for yourself.

So thank you for mentioning that. We feel pretty strongly about it. We want to get that information out there because we’re team Seattle and team Pacific Northwest, and we want our clients and our neighbors to be interacting securely with AI.

I’m talking over you and I shouldn’t be. So—

Jenny Moshea: Oh, so here’s one—just a basic definition. Can you explain AI hallucination?

EXP: Oh, yes. So, it’s when the tool literally makes something up. I had this happen even yesterday on a paid version. It came up with the word ‘vespers,’ and I’m like, ‘I haven’t mentioned vespers. I haven’t given you the word vespers. What does that even mean?’ It’s like this old reference, but it just popped it into the response without any guidance. So sometimes—that’s why you still have to be human smart. It is not replacing humans by any means. It is elevating, expanding, and boosting us. You still have to look at everything it produces and just sanity check.

AI is incentivized to be an answering machine, and when it can’t find an answer, sometimes it can fill in the blanks. Famous cases of that where people have embarrassed themselves by citing legal precedent that didn’t exist or medical reports that don’t actually exist. That goes back to human in the loop, right? We want to review those things.

If you don’t know—because younger folks ask me this all the time, you know, when I speak at startup events—if you are not experienced in the world, you haven’t had enough cycles around the sun, you may not know the answer. One way to handle that is to put the answer into another chatbot or ChatGPT or another GPT to say, ‘Can you verify this?’ And you also ask it all the time, ‘Where are the sources? Verify the sources,’ because if it can’t point to sources or sources that are not real, you should be able to validate those. You just—you’ve got to keep asking, just like you would ask a child or a friend, ‘Hey, that sounds a little off. Where did you get that information?’

So, in the era of some fake news, you really have to trust but verify.

Jenny Moshea: And that terrifies me, because I’ve seen studies on people in silos creating disinformation machines and how successful they were at spreading disinformation. If you’re talking about an issue that’s hotly contested—and we’re not going to talk about these things—but I’ll just raise the issue of vaccine effectiveness or those sorts of topics. There’s a lot of information on both sides, and AI might give equal weight to the noise as much as to the signal.

EXP: Yeah. And I think—

Jenny Moshea: Yeah. Also, you have to—you know, we’re trusting and we’re learning, but also it is not any different than the rumor mill in high school or middle school. There was a rumor mill. It’s just a lot faster now and it can spread like fire.

EXP: Indeed. So, here’s one that speaks to me. Is there a way to get around the tendency of AI to flatter users in order to get information from it on your level of expertise in prompt engineering? I hate that. Okay, go on.

Jenny Moshea: That’s why I went to that number one thing I showed you, which was the black slide of go into your settings and go to the personalization, and you can tell it, ‘Be direct.’ You can even say, ‘Give me a New York Bronx style—be direct. Do not sugarcoat it. Do not give me flattery.’ You can change the way it responds to you.

EXP: I haven’t done that, but I am continually irritated by the ‘What an excellent question,’ or ‘That’s an outstanding way of thinking.’

Jenny Moshea: Yeah. Very cheery. I wanted to tell you—prompt it to be direct. You can even say, “Give me a New York Bronx style—be direct. Do not sugarcoat it. Do not give me flattery.” And you can change the way it responds to you.

From Ideas to Implementation for Small Businesses

Small business AI strategy: Learn how small businesses can identify high‑impact AI opportunities, automate repetitive tasks, and build secure, practical AI solutions with expert guidance.

EXP: I want to get to—the main course here. We’re nibbling around the edges, and I want to make sure we don’t ignore AI implementation in organizations, particularly in small businesses.

You’ve talked about things like personalization and using one GPT to validate the information another is giving. For the business leaders thinking, “There’s this new tool; I’ve seen it create video quickly, I’ve seen it get research answers quickly, but I don’t know how to use it to amplify my efforts in my business”—where does that person start? What should they be looking at?

Jenny Moshea: Start—I’m so glad you asked. Can you all see this? My questions? It’s coming up… here we go.

These are things to look for, and I follow the pain. Where do you get stuck? What do you not look forward to doing every day? Where do you find drudgery?

Where in your processes or workflows do you have to do repetitive tasks that are not the best use of your expertise or time? Time is money.

Also look at where work moves across systems or you’re copying and pasting. I have been using a machete to spreadsheets my entire career, and I always look for the spreadsheet mess in the middle because that’s where really good work can be done.

Spreadsheets are amazing for a lot of things, but they’re not for a lot of things. People have used them as a crutch. I’d encourage you to think through: what would I love to not have to do every day?

One way to think about this from a human perspective: right now for typical office workers, we are on our keyboards 80–90% of the day. Can you imagine that being reversed, where you are only 10–20% on your keyboard? That is the future with agents, because the agents will do what you’re tediously tapping away with on your keyboard and give you more time to interact with customers, interview stakeholders, or do the business of human‑to‑human.

EXP: Elaborate on that. When you say that, I think of—I’m not a Star Trek fan—but I think of Star Trek where they would say, “Computer, do this for me.” Is that what you’re thinking?

Jenny Moshea: Exactly that. Most of the time right now this is already happening with some business owners. They just use their phone; they talk to their GPT using the microphone.

You talk to your GPT “COO assistant” and say, “Hey, we’ve got this deal. Can you write up a draft summary for me and email it to my inbox?” And all that happens without you touching a keyboard.

EXP: That’s especially timely. Several of us at EXP went through training recently on building AI agents. One of our exercises was very close to what you just described.

Imagine a law firm. They review contracts. Contracts come in and have deadlines associated with them. Attorneys spend a lot of time summarizing the contract; paralegals do the same. Sometimes important deadlines and dates are missed because the attorney doesn’t see it.

So the exercise was: write an agent that can view incoming email, identify which ones have contracts (PDFs) attached, review those contracts, identify deadlines, summarize, and then—there’s still a human in the loop though. You don’t want ChatGPT giving legal advice. But it can do a lot of the grunt work. For example, researching: What is relevant precedent? What should we be concerned about? What does Washington State law say on this matter? It accelerates the research.

Jenny Moshea: I love that. Accelerating the research. That’s a lot in the AEC world—in permitting and compliance—and just serving up a lot of the research. That doesn’t mean a human doesn’t scan to say, “They missed some important information,” but you’re at least not doing all of the heavy lifting. And that’s key.

EXP: How do organizations stop from reinventing the wheel? I’m thinking of our prep conversation. I mentioned AI seems great for project planning for AEC or project management for AEC firms. I described a scenario: let’s say we have a project and there’s a delay—concrete work can’t be done and there’s a three‑day delay. That throws a wrench into the entire project plan because there are dependent activities. Those get shuffled.

How do we reprioritize to other activities that could be done independent of that concrete foundation being poured?

Jenny Moshea: There are beautiful solutions like Kaya AI that help solve that problem—the middle between your scheduling software and your procurement and your materials and your crews.

Whether or not it’s that specific tool, those solutions already exist on the market today. They’re trained in construction terminology and documentation. But more importantly, I know no superintendent who will just let it ride and let a system or an agent change their schedule. That is forbidden.

So, of course, there’s a moment where it stops. The AI can make the recommendation, and then the human says yes or no or reroute. You have to keep that control in there. There are certain things we’re not ready to let AI do. We’re not ready as a society to let AI fly a plane. It might be possible, but we’re not ready to accept that. Trust has to be built. That’s part of the journey we’re on.

EXP: Yes. It seems like this is a recurring theme: a human in the loop—or an expert in the loop—to validate.

So the process then, to make sure you’re not buying or inventing something that’s already been built, is…?

Jenny Moshea: First, before you think you need to hire somebody or you’re drowning—“If I just had an assistant”—you can create a lot. For more straightforward things, you can create your own GPT to assist you. Give it instructions, and I even have ChatGPT help me write the instructions: “You’re a world‑renowned COO” or “a trained financial analyst; you do these things.”

I think of opportunities in construction with the WIP—the work‑in‑progress schedule—and the spreadsheet used by financial analysts, the CFO, the teams. A lot is riding on this spreadsheet. What has become a system of record or an independent tool—then we’ve got SaaS systems of record like QuickBooks or your ERP. What is emerging now is a system of intelligence above that.

That’s why I’ve joined forces with Rapid Canvas—because it’s elevating this whole process of how we think. But I would start with: before you hire anyone or think you can’t do any more alone, what would help you or assist you? What would you do next? Then you’re looking for that functional area of expertise. It can be an off‑the‑shelf tool that has AI; it can be something you build; or something you build with a partner.

EXP: Great. That’s great advice. I’m going to put you on the spot—it’s okay if you don’t have an answer. Maybe speak at a high level.

Can you talk about examples of projects with really high ROI where there was an AI implementation in an organization—beyond “we gave everybody ChatGPT and look at how that amplified their efforts”—something broader in scale?

Jenny Moshea: A very common one across every industry is a process where you are ingesting emails and lots of PDF documents. You’re reviewing documents, reading a lot of material, scouring or going to certain sites you have a login to, to collect and collate a report, for example.

This is one of the number one things: the reports we produce for our CFO or for our board. A lot of that can be done with agentic AI that has specific tasks and instructions to gather emails and documents and spreadsheets into a platform, then create the report. If you’re doing any copying and pasting—it’s amazing—we have high‑powered attorneys, financial analysts, stock brokers, real estate investors doing a lot of copying and pasting. That is a key indicator of where AI can assist.

It greatly reduces the time hours of those doing that work and creates space for capacity. That’s why I say, before you go hire someone, see if you can create more capacity for yourself.

In a recent study, some of the savings we’re finding does not end up on a company’s balance sheet. That’s hard right now for our productivity‑minded society and CFOs to absorb. Time is going back to families, personal time—the time you were tilting over in the first place.

EXP: You see me smiling because I’m thinking, “Great—personal time and those gains.” At the same time, I’m going to offer a counter, maybe phrased as a question.

What are your thoughts about the next generation? We’re talking a lot about “expert in the loop,” but there’s also a need for human apprenticeships—for humans to go through that process—or we’ll have a generation 20–30 years down the road where there aren’t real engineers and accountants and architects and people with specialized subject‑matter expertise. Do you agree? Any guidance for folks in attendance on how to keep that in mind and weigh benefits and costs?

Jenny Moshea: This is fresh for me—with a senior in high school and two more behind him. They’re asking, “What should I major in?” Humanities—things that are human—but also…

I think back to the calculator. When the calculator came in, it was forbidden in classrooms. Everybody thought it would take over accounting. Then came spreadsheets, elevated analysis, and accounting and finance professionals doing higher‑value activities than doing calculations on a calculator.

We are going to work differently, but there’s still learning. There’s also an opportunity to capture things we’ve never been able to capture before—our parents’ and grandparents’ memories and stories, and superintendents’ craft and techniques—before a whole generation starts to retire.

We can capture and “bottle up” and then create interactive training with new people onboarding—have a much more flourishing experience than a dry training manual.

EXP: That gets me thinking back to my childhood. My grandfather died when I was around 12. He was an immigrant from Italy and, like many from his generation, came here very early in the 20th century. He didn’t have great English skills even at 92, so we couldn’t communicate very clearly.

AI is a great translator and interpreter and a great resource to offer prompts to humans that could facilitate that kind of conversation—how do we capture the things that are important with the people that are important to us.

Jenny Moshea: It’s really practical, too. In commercial construction, safety meetings are vital, and there are a lot of languages represented in the Pacific Northwest—Russian, Spanish, and all sorts of dialects. Tools like Google Translate and Wordly are key to a global audience.

EXP: This gets back to a question that came in: is there a secure note‑taking AI? There are privacy policies—sometimes you don’t know where the notes are saved, and sometimes policies say they can train on your data or use your video. Do you have—

Jenny Moshea: You can look at those policies. Yes—the Otter AI, the Fireflies, the Tactic, Gemini doing it now. That’s part of figuring out your company policies and industry compliance and regulation. It’s a good conversation to have.

EXP: What we can say is: there are some things not to do. Don’t share company information in the free versions of AI products. Generally—I feel like Copilot is a little more secure because it exists—if you’re using the “for work” tab and the paid version of Copilot—you’re operating in an environment that has some access controls built around it. Would you agree? I’m saying this with a high‑level understanding, but you’re a CIO.

Jenny Moshea: That would be a logical starting point. I wouldn’t want it to provide a false sense of security. You still have to behave correctly and not just send data everywhere. You have to have access controls, and that’s part of cybersecurity training 101. We have to make sure not everybody in the company is seeing payroll information, for example, or HR information.

EXP: Speaking of cybersecurity 101, we do have a security awareness training course at EXP Academy as well.

We’re running short on time—not done yet—but if people have questions, submit them now. I have one for you that’s kind of a wrapping‑up question: Is there an AI‑related shift that businesses are not preparing for right now, but should be? Is there something on the horizon—a tree in the middle of the road, a boulder, a bridge that’s out—or maybe something wonderful that people are not preparing for that they should be?

Jenny Moshea: I am really nervous for business owners right now. You need to get on board—learn how—and figure out what you need to get the trust in the tools or the technology because it is happening.

We cannot make a better BlackBerry keyboard. We have emerged to the smartphone, and that paradigm shift is a fraction of what we’re going through right now with AI. I’m nervous about people falling behind.

I think the agentic AI space is the future. It’s an opportunity to reimagine how we do work. It is vital and important that we reconsider how we are spending our time as humans.

EXP: I’m going to pose one more at you. A CEO at a small business—if you could tell them one thing to do in the next 90 days in their AI adoption journey, what would that be?

Jenny Moshea: Create your own virtual assistant with a GPT and start collaborating with AI.

EXP: Jenny, we’re up against time. I feel like this could have gone on for another hour. We’re barely scratching the surface. Perhaps we’ll have you back and dive into some topics in depth later on.

For people who want to connect with you or follow you, where can they find you?

Jenny Moshea: I think “Jenny Moshea” on LinkedIn is a great place to start. We can connect there and take it from there.

EXP: Awesome. Thank you so much. For folks who want to connect with us at EXP—most of the people in attendance are currently clients—but for those who are not, you can find us at expchnical.com. I’m kelly@expchnical.com. You can email me directly.

We will be sending a link to this video in about a week, so you can review it then. I see 12:59 on my clock, so I think that’s about the moment to call it here.

Jenny, thank you so much for joining us today. I really enjoyed the conversation. You made it fun and interesting.

Jenny Moshea: Cheers.

EXP: Go Mariners.

Jenny Moshea: Yes.

EXP: Thanks, everyone. Goodbye.