What happens when AI agents stop waiting for your prompt and start bringing you recommendations every morning?
Paul Roetzer predicts that shift reaches corporate teams within six months, and explains why personal agents are moving faster than business ones.
In this AI Answers, Paul and Cathy McPhillips take 14 questions from our Scaling AI and Intro to AI classes, covering agent governance, platform standardization, shadow AI, forward-deployed engineers, and how to show AI skills in your resume to a hiring manager.
Listen or watch below—and see below for show notes and the transcript.
Over the last few years, our free Intro to AI and Scaling AI classes have welcomed more than 50,000 professionals, sparking hundreds of real-world, tough, and practical questions from marketers, leaders, and learners alike.
AI Answers is a bonus series that curates and answers real questions from attendees of our live events. Each episode focuses on the key concerns, challenges, and curiosities facing professionals and teams trying to understand and apply AI in their organizations.
In this episode, we address 14 of the top questions from our latest Intro to AI and Scaling AI classes, covering everything from tooling decisions to team training to long-term strategy. Paul answers each question in real time—unscripted and unfiltered—just like we do live.
Whether you're just getting started or scaling fast, these are answers that can benefit you and your team.
00:00:00 — Intro
00:06:18 — How do you scale AI under compliance requirements?
00:07:53 — Should companies block frontier models and only allow self-hosted AI?
00:11:34 — What should AI teams prioritize when governance isn’t ready for AI agents?
00:15:34 — How much AI platform standardization do organizations need, and how should leaders handle shadow AI?
00:21:31 — Where do forward-deployed engineers fit, and who owns AI enablement?
00:24:04 — Where do OCR and RPA fit alongside AI agents?
00:25:40 — Should you redesign work by mapping tasks or business outcomes?
00:27:44 — How can business owners use AI to become less of a bottleneck?
00:31:09 — How much client education should AI-enabled marketing consultants provide?
00:34:57 — What does responsible AI personalization look like for alumni and donor outreach?
00:39:11 — How do you show AI skills to a hiring manager with low AI knowledge?
00:41:52 — What AI skills matter most for a mid-career pivot into project management?
00:45:07 — Will AI ever write as creatively as a talented human?
00:48:31 — Will top performers negotiate token budgets into job offers?
This episode is brought to you by AI Academy.
Get $250 off a new AI Academy by SmarterX individual license through November 30 with code POD250. That’s 25% off your first year of the full Academy experience: courses, professional certificates, live sessions, and community.
Leading AI literacy for your team? Buy 25 or more licenses by November 30 and Paul Roetzer will run a private one-hour virtual session with your team. Buy 250 or more and Paul will come to you for a private in-person session in the first half of 2027 (US only). Existing business members adding 25 or more licenses qualify too. Join at academy.smarterx.ai, or contact our team at support@smarterx.ai about business licenses.
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: So imagine that every role in the company has proactive agents that are just watching the data. And then coming to you with recommendations that that is the unlock that hasn't happened in corporate yet, but I believe it's coming in the next six months. Welcome to AI Answers a special Q&A series from the Artificial Intelligence Show.
[00:00:20] I'm Paul Roetzer, founder and CEO of SmarterX and Marketing AI Institute. Every time we host our live virtual events and online classes, we get dozens of great questions from business leaders and practitioners who are navigating this fast moving world of AI, but we never have enough time to get to all of them.
[00:00:38] So we created the AI Answers Series to address more of these questions and share realtime insights into the topics and challenges professionals like you are facing, whether you're just starting your AI journey. We're already putting it to work in your organization. These are the practical insights, use cases, and strategies you need to grow smarter.
[00:00:58] Let's explore [00:01:00] AI together.
[00:01:05] Welcome to episode 251 of the Artificial Intelligence Show. I'm your host, Paul Roetzer, along with my co-host today, Cathy McPhillips. Welcome, Cathy.
[00:01:13] Cathy McPhillips: Thank you so much.
[00:01:14] Paul Roetzer: Cathy is the Chief Marketing Officer at SmarterX. she is not replacing Mike. This is an AI answers special episode. This is our 23rd edition.
[00:01:23] So Cathy and I do these together once or twice a month. And this is our series based on questions from our monthly Intro to AI and scaling AI classes. those are classes we've been doing for about five years now. Intro to AI, I think's been five years. About 65,000 people have registered for that series.
[00:01:40] And scaling AI, we just did. 13. Number 20.
[00:01:44] Cathy McPhillips: We just said our 20th.
[00:01:46] Paul Roetzer: 20th. Wow. Okay. Yeah. Lose a track of time. So that one we've been doing for close to two years now. and that, I don't know how many people have gone through that series. Probably like 15,000 or something like that. Or 20,000, I dunno. A lot. So we appreciate everybody who's been [00:02:00] a part of those series.
[00:02:00] And the cool thing is when we do those classes, they're free webinars, that we host, and we always leave about 15 or 20 minutes at the end for questions. And there's always more questions than we can get to. So I know on the scaling AI we just did for November 11th, which is part of today's. there was like 15 or 20 questions left in the q and a.
[00:02:20] Yep. And then the September 30th intro to AI is the other one. So if you attended either of those, the September 30th intro to ai, November 11th, scaling ai, these are the questions that were kinda left in the sandbox from those. And if you didn't join us, you're gonna get a bunch of insights from the things people were thinking about as we went through those.
[00:02:37] So there's a few other listeners submitted questions in here as well. Claire on our team who produces the podcast does a great job of curating all of this and keep it in organized for us, and preparing Cathy and I to do this, this AI answers episode. Did I miss anything there, Cathy? You did not logistic.
[00:02:54] Okay, cool. All right. And this episode is brought to us by AI Academy by SmarterX this past Tuesday. So [00:03:00] that would've been November 10th. We celebrated our 250th episode of the Artificial Intelligence Show podcast with a special AI Academy offer. So if you haven't taken advantage of this yet, this offer runs through November 30th, so AI Academy.
[00:03:16] Helps individuals and businesses build their AI literacy and put that knowledge to work through courses, certi certifications, that's a hard word to say. Live sessions. Personalized learning journeys and dynamic learning that we're gonna be doing a lot of, announcements around that coming up on November 19th.
[00:03:34] We add new content weekly, covering the latest capabilities of AI apps and agents. Real stories of personal and business transformation and platform specific courses are coming soon for ChatGPT, Claude Copilot and Gemini. That is one of the most requested things. We hear from our business accounts in particular.
[00:03:52] and that is coming soon. So we're really excited about that. So through November 30th, you can get $250 off an [00:04:00] annual individual AI Academy license with code POD250. So that's about 25% off of your first year of academy membership. And then for organizations that are driving business transformation, we have two special offers.
[00:04:13] So by 25 or more team licenses by November 30th. And I will do a private one hour virtual session with you and your team, or it can just be an executive briefing, whatever you wanna do. And then by 250 or more licenses, and I'll actually come into a private in-person session in the first half of two thou, 20 2027.
[00:04:34] In the United States is the caveat there. I'm not, I can't do international travel, but if you're in the US and you buy 250 or more academy licenses for your team, by November 30th, I will come into an in-person session with your team so you can visit academy SmarterX.ai to learn more. And if you're gonna do an individual membership, go ahead and use that POD250 and get immediate on-demand access to the entire library of more than 400 courses [00:05:00] and all the upcoming live stuff that's gonna be happening.
[00:05:02] And then if you want to look into the business accounts, just reach out to our team. You can fill out a form on the site and our team will be in touch with you. We'll put links to everything in the show notes. and then if you'd like a closer look at what's going on with AI Academy, we have the November 19th AI Academy Day building and AI forward workforce.
[00:05:19] That is a free webinar happening, at noon Eastern time on November 19th. You can register at SmarterX.ai Alright. Did I miss anything there, Cathy?
[00:05:30] Cathy McPhillips: I don't think so.
[00:05:31] Paul Roetzer: Everything's good.
[00:05:31] Cathy McPhillips: You did a nice, nice job.
[00:05:33] Paul Roetzer: All right, good. All right, so let's do it. AI answers.
[00:05:36] Cathy McPhillips: So, interestingly, when we were doing these questions, since we did the scaling AI this week, Claire had to turn around pretty quickly to get these questions done.
[00:05:45] So she uses AI to help her synthesize the. The questions go through the previous questions for intro, put it into something to see, like, has this question been asked 10 times? We don't wanna do an episode on the podcast with a question you've answered a million times. And then I wasn't [00:06:00] sure what number AI answers this was.
[00:06:03] What did we do before artificial intelligence? I just went into the Google Notebook and said, what AI answers number is this? And it was like, it is number 23. so it was so helpful. So we do love being able to use some of these tools for, for things such as this. All right, let's jump into these questions.
[00:06:18] Cathy McPhillips: Number one, for organizations with strict security and compliance requirements, including SEC two controls or HIPPA, obligations, what's a practical first step towards scaling ai?
[00:06:30] Paul Roetzer: I mean, the first thing is to understand the limitations you already have in your organization, so the stakeholders who, you know, kind of make sure that everything stays.
[00:06:39] Structure the way it's supposed to, that you understand their concerns, the risk associated how they think about ai and that anything that runs up against any regulations, within your industry in particular, that you understand those and you're planning against those. The other thing though is to not let.
[00:06:57] Those restrictions or [00:07:00] limitations prevent you from driving transformation in your organization or across your team. And so the way you do that is to find the use cases where you don't run up against any of that. So the example I think we've used before is like in a healthcare system, you know, there, there are plenty, probably thousands of use cases for marketing, sales, customer success, operations, hr, finance.
[00:07:23] You can be using AI systems, your favorite AI tool without running into any of those regulatory risks and concerns. And so getting really good at identifying low risk or no risk use cases and then just stacking those is a really good way to keep moving while you're understanding the bigger picture and the risks that you need to be aware of in your organization.
[00:07:44] And that applies like hospitals, banks, like we talk to companies like that all the time, and I think that that messaging is relevant to them. Yeah,
[00:07:51] Cathy McPhillips: absolutely. Okay.
[00:07:53] Cathy McPhillips: Number two, what would you say to an organization that plans to block access to frontier models and only allow employees to use models hosted within its own infrastructure?
[00:08:03] How should leaders weigh security costs and capability when deciding what to make available?
[00:08:08] Paul Roetzer: This is definitely a problem in bigger enterprises. We do see this all the time. Either they're blocked or they're very limited versions of access to these models. So this happens a lot with Copilot is the one I hear about a lot, where a lot of the capabilities just aren't built in.
[00:08:22] And so you have individual users. Who maybe are using Claude or ChatGPT on the side, I mean, their personal accounts and understand the full capabilities. They're using the reasoning models, they're using the agentic capabilities potentially, and then they go and they use the tools they're given access to in their jobs and realize they're nowhere near as capable.
[00:08:42] yeah, I don't know. I mean, it really starts with education and literacy at the top. So, if. The decisions that are being made on the technology side are limiting the impact and the ROI that AI can have across the organization. Might need to start with an [00:09:00] executive briefing of what's actually possible and what the limitations are restricting in terms of our ability to make an impact on the growth of the company.
[00:09:09] and, and so like. My assumption is usually there is a lack of understanding if all these limitations are put into place. if that's not the case and they do fully understand and they're still dramatically restricting access, honestly, like that is the point where you kind of start looking around and saying if you understand the full capabilities and what you could be doing in your role in that company.
[00:09:34] And you realize you are not going to be able to fulfill your potential in that company and they're not going to give you access to the AI tools that you believe are necessary to unlock your potential, then you might wanna look for work at a company that is going to, and this is something we do talk about and again, I don't like look at these questions, so sometimes I may pres answer a question Cathy has in here.
[00:09:57] One of the things to think [00:10:00] about is when you go into the interview process, if that company you're interviewing with isn't asking you about your AI capabilities and your AI literacy, that's a pretty good sign that they're not doing much themselves. Whereas like if you go into a company and the first thing they're asking you is, okay, so what AI models are you used to using?
[00:10:18] What's your preferred AI system like? That's a pretty good indicator that they're a little more AI forward than another company.
[00:10:24] Cathy McPhillips: And you used this stat a while ago, I forget which presentation it was in, but you know, X number of people or their company say you can't use ai and then this higher percentage of people using it on their own devices because their company won't let them.
[00:10:38] So what happens when, even if people know they shouldn't be doing this, but like we ought get stuff done, are they doing that and putting their companies at risk?
[00:10:47] Paul Roetzer: I mean, I do think that that's happening all the time. Obviously, it's hard to get at that data because people would have to admit that they're doing that for that to show up in the research.
[00:10:55] But I mean, based on just like conversations and sort of [00:11:00] qualitative feedback, it seems like it's pretty prevalent. I mean, I've been in a workshop where the person that's in charge of providing access to the models and is the one that's setting the guardrails. Was using a personal account on the side to do work.
[00:11:16] Right. So that, and that was a big enterprise. So just gives you a sense like if the people who know the risks also understand what's possible and are going outside of their own guardrails to get stuff done, there's a pretty good chance that the people within that organization are doing something similar.
[00:11:32] Cathy McPhillips: Right. Okay.
[00:11:34] Cathy McPhillips: Number three. For larger companies where governance teams aren't yet comfortable with widespread use of AI agents, what should AI teams prioritize and in what sequence should they focus on? General AI literacy, connecting approved tools to internal data or building reusable workflows for specific roles.
[00:11:52] Paul Roetzer: I do think this one starts with that understanding of what we mean by agents to, you know, I think that's a good starting point. So I've said before, [00:12:00] but like, agents exist on the spectrum of reliability and autonomy or like the two most important variables, in my opinion, to, to really think about at, at the starting point with agents.
[00:12:09] And what I mean by that is, how much human in the loop there is with the agent. So some agents are very deterministic in nature, meaning most of what they're going to do, the rules are set by you, the user. So it, it's really more just an automation than it is some probabilistic outcome being driven by an AI model that you don't even know what it's doing or how it's doing it just like creates the end product.
[00:12:35] And so we've talked about things like, you know, Zapier and make, where, let's say you take a workflow as a marketer, you know you're gonna like announce a product launch or something, and you define these 25 steps of going through this process and elements of it, you just set rules. It's like, okay, I wanna do this.
[00:12:54] This and this. Grab this data, do this thing, and you know exactly how the agent is going to behave. Now again, we're [00:13:00] using the age, the term agent pretty loosely here 'cause it really is just a traditional automation. But you can very confidently say, okay, these 15 steps are basically just automation agents and they're doing exactly what I tell them to do.
[00:13:13] But now these other 10 steps I'm in, I'm using an AI agent that has. Some of its own ability to make decisions and determine its path forward, so it might build its own plan. It might. Craft its own creative, it might segment its own lists, like it's gonna do stuff where you are just now overseeing it. And so I think to set up this governance, to understand how involved we need to get, we need to understand how autonomous are we trying to make these agents in our workflows across our teams.
[00:13:45] And once you do that, then you can start to have a better handle on how to set them up and how to govern them. So the better starting point for most organizations is to start very low on the autonomy scale. So more [00:14:00] control, more clarity about exactly what it's doing. Or more very low risk outcome work.
[00:14:06] So for example, setting up an analyst agent that every day goes and looks at marketing data and sends it to you in ChatGPT each morning and says, here's what happened yesterday. If it gets something wrong, there's no harm done. As long as you're not activating that data and not like assuming it's all a hundred percent correct.
[00:14:22] So it's not going to anyone external. There's really no risk involved, but it's, it's mostly just like sort of figuring out its own thing. I just tell, go look at the data and come back to me, and then it does its work. So again, it does start with literacy. It starts with understanding what agents are, what the variables of using those agents are.
[00:14:40] and from there you can start to make more educated conversations. But the more advanced they get, the more autonomous they get. The more access to data they get, the more ability they have to take actions on your behalf that maybe have more risk associated with them. You, you need to keep, you know, be very intentional with how you set those up.
[00:14:58] Cathy McPhillips: Right. Yeah. I mean, you [00:15:00] can't go from zero to 60 and just like turn this thing on, like you said, like make sure that you're seeing it along the way and say, oh, it can do that. Like I'm gonna take it one step further, one step further and keep improving on that. And then, yeah.
[00:15:10] Paul Roetzer: And there are people doing that.
[00:15:12] Like we, there are definitely organizations that are like way out on the frontier taking enormous risk. And I will share that stuff with the team sometimes. Be like, yeah, this is not us. Like I'm glad someone is doing this and sharing this, but like, we will never be doing what these people are doing.
[00:15:28] Cathy McPhillips: You talk about it on the podcast a little bit.
[00:15:30] Okay. Yeah.
[00:15:31] Paul Roetzer: Good.
[00:15:33] Cathy McPhillips: Okay
[00:15:34] Cathy McPhillips: Number four, how much standardization do organizations actually need when it comes to AI platforms? How do you get employees to adopt approved tools consistently? And when does a mandate make sense? As personal AI agents like Muse become more accessible, how should leaders address Shadow ai?
[00:15:50] That's kind of two questions there.
[00:15:52] Paul Roetzer: Yeah. Okay. So let's start with the first one. How much standardization to organizations actually, when it comes to AI platforms? so I, I've shared this [00:16:00] before on the podcast and, and some of our classes, you know, at SmarterX we have access, everybody has access to three platforms at least.
[00:16:07] Like we all have Gemini, Claude, and ChatGPT licenses. At different times, I've wanted to go all in on one of them and just like focus the team on really learning those tools. But I think the feature sets are dynamic enough and the value creation is, tangible enough that to me, I can justify what roughly is like $60 per user per month to invest in those platforms for each person.
[00:16:33] That being said we basically centralize which ones of them get access to outside data sources like our CRM system, our financial systems, things like that. And so we're much more cautious when it comes to outside data access. and so, you know, I would say in a perfect world for me. You are centralizing around a single AI system because then you can better govern [00:17:00] that and you can drive innovation and optimization through that one platform.
[00:17:06] But some people you may want to allow them to be continuously experimenting with the other systems, running their own evals, you know, for high value tasks. And that's basically how I use it. I would say that I predominantly. Like right now, I predominantly use ChatGPT. There was definitely a period this year where it was clawed, like I was more so.
[00:17:26] But if it's a high value task, like if I'm working on a strategy or a prototype of like a tool I want to build, I will, I will start building it in all three of them, and then I will see which model and which system. Is going to best help me achieve the outcomes I want for that high value project. And that's, that's part of the reason why I like having all three is sometimes one of them is just better at something.
[00:17:50] And that one project I do pays for the licenses for all three of them for the year. So that's, that's hard to do in bigger enterprise though. But I think that's the way I think you [00:18:00] need to have. A group of people, a lab, a center, center of excellence, whatever it is that has access to the major ones. And they're always running evals and sort of testing.
[00:18:09] And then I think the rest of the team who aren't maybe as advanced in terms of AI maturity, focus them in on a single platform and get them really good at using that platform. That's probably the best path. but again, it's not gonna be prescriptive for every organization.
[00:18:26] Cathy McPhillips: And there is some value in having more than one because what happens when custom gpt go away and you've put so much energy into that, that it's good to know at least one other tool.
[00:18:35] Paul Roetzer: Yeah, they, they do, they make changes to their system where it's like, well, that just broke how we were using the system.
[00:18:40] Cathy McPhillips: Right. And then do you wanna address that as personal, a agents like Muse become more accessible? How should leaders address shadow ai?
[00:18:47] Paul Roetzer: Yeah, I mean, they're all gonna have these agents that are gonna be more useful.
[00:18:55] On the personal side, initially, anybody who's playing around with dots, for example, or scheduled [00:19:00] tasks in ChatGPT, certainly in meta, Google Gemini has the ability to schedule things now. So there's some that are happening in the business world, but most of the. Innovation is happening on the personal AI agent side because they can, you're more likely to give them access to all the data they need to be useful because for these agents to be extremely valuable, they have to be proactive, meaning they have to bring use cases to you, rather than waiting for you to come back in and use them again.
[00:19:31] And so, like imagine each morning you go into like ChatGPT or Gemini or Claude. It says, Hey, you know, okay, I'll play this out real quick. So let's say ChatGPT has access to your CRM. So we're gonna say these personal agents have now moved into, these proactive agents have moved into the realm of the business world, and ChatGPT has access, let's say it's dots or whatever their agent is, has access to your CRM and your financial data at a limited sense.
[00:19:56] And every morning. It monitors [00:20:00] without you telling it to. It's monitoring your CRM data based on things that it knows are important to you based on previous conversations, and it surfaces things for you and says, Hey Paul, I noticed that webinar registrations took a spike yesterday. great work. I wonder what's going on there.
[00:20:15] I dug in and I actually found this ad is working really well on LinkedIn. You should pass this along to your marketing team and give them a heads up to double down on that ad. That's a proactive agent. Like that's something that is watching things on my behalf and coming to me with use case ideas. And maybe it even says, here's three alternate ad copy ideas that might work really well.
[00:20:38] Give these to your team. So imagine that every role in the company has proactive agents that are just watching the data. And then coming to you with recommendations that, that is the unlock that hasn't happened in corporate yet, but I believe it's coming in the next six months. so right now people are having to like hack this together, but I think that's gonna cross over [00:21:00] into businesses.
[00:21:02] Cathy McPhillips: The better agent would be for me to get that
[00:21:04] Paul Roetzer: Yes.
[00:21:04] Cathy McPhillips: And then for you to get it saying, we already made this change.
[00:21:08] Paul Roetzer: Yes, yes. Yeah, it, yeah. So that's. I think what'll happen is like you'll, the center of excellence or labs or whatever, the centralized body within organizations with the most AI forward people will proactively go out and build personalized agents for the roles in the company.
[00:21:29] Cathy McPhillips: Looking forward to that. Okay.
[00:21:31] Cathy McPhillips: Number five. How should organizations bring technical builders closer to business teams to turn AI opportunities into working solutions? Where to forward deploy. Engineers fit and how should business leaders IT and HR share responsibility for AI enablement?
[00:21:47] Paul Roetzer: This actually builds on the previous question, I guess.
[00:21:49] So the forward deployed engineer, if you're not familiar with that concept, it was like kind of made popular by Palantir, and now all the AI labs are basically building consulting arms of forward deployed [00:22:00] engineers. The basic idea here is people don't know how to use software or AI models, and so they send technical people into these big companies.
[00:22:09] Or they're internal people that do it, and they basically look at workflows and problems within an organization, and they identify how to use Claude ChatGPT co-pilot Gemini to solve things, and then they charge you millions of dollars to build things that, that's kind of like the basic premise of a forward deployed engineer.
[00:22:25] So I think that there, those roles are, are really important right now. I actually, I'm not a believer that they're as necessary a year or two from now. I, if an organization trains their business people to understand how to use these technologies, because most of what needs to be built, you can vibe code with no coding background or technical knowledge.
[00:22:47] You can just go in and say. Hey, I, I've got this idea to like create this app that does this thing and 30 seconds later ChatGPT built you a prototype you can play around with. And so if you [00:23:00] have business leaders or even practitioners who are extremely strategic minded and really strong critical thinkers, they just need to be given access to the tools.
[00:23:11] They don't need a Ford deployed engineer to come in and tell 'em what to do. So I do believe though that, His, you know, if you look at the historical structure of an organization, there's just gonna be business-minded people who are just better at strategy and critical thinking, and I do think centralizing those people and giving them, high levels of AI literacy and competency.
[00:23:36] And then having them go meet with the business unit leads and the team leads and the practitioners and run workshops and bootcamps and hackathons. I think that's probably where most organizations are gonna go over the next five years, is centralization of that building. I just don't believe that they're gonna necessarily have to be a traditional Ford deployed engineer who's a technical person.
[00:23:57] Like I don't, I don't think they're gonna need any coding background or [00:24:00] engineering capabilities. Like I I think they're just gonna be business people who build things.
[00:24:04] Cathy McPhillips: For sure. And speaking of that, number six, where do technologies like optical character recognition or robotic process automation fit today?
[00:24:13] How do you decide when to use these technologies, when to use an agent or when to combine them?
[00:24:18] Paul Roetzer: This goes back to the first conversation around agents and it's really just, you know, a lot of traditional technology that was like early forms of artificial intelligence, like image recognition and things like that.
[00:24:30] it's just like merging into this automation and, and it just, the automation kind of got rebranded as agents, like this general term for agents, but again, they exist on this spectr Sometimes within that agent, there's this deterministic outcomes where you're setting the rules and sometimes with, you know, it's more probabilistic where you're letting the agent sort of figure stuff out.
[00:24:54] So, yeah, I mean, a lot of this is just blended in and the technical people. [00:25:00] Care, what we call it, and whether or not it's always been a thing for 20 years or 30 years. And the non-technical people don't, they just, they just want it to do the thing that they're envisioning. Yeah. So yeah, I would say just increasingly it's like blended into the technology itself.
[00:25:16] and, and it's, it's gonna continue to play, but I do think that people overlook the role of traditional automation processes that like. A lot of times we don't need to use the most advanced model. We don't need Astra six Pro or High to like do something that's literally just a if then statement, right?
[00:25:36] Like just use automation for that.
[00:25:38] Cathy McPhillips: Yep. Speaking of that,
[00:25:40] Cathy McPhillips: Number seven, when redesigning work, does it make sense to start by mapping the tasks people do today alongside the outcomes the business actually needs? How do you use both perspectives without simply automating an inefficient process?
[00:25:53] Paul Roetzer: So we talk about two main.
[00:25:56] Outcomes or value creation. [00:26:00] elements of ai, which is optimization and innovation. So if you wanna optimize a process, do it better, faster, cheaper than mapping tasks and workflows. and what the outcomes are is a really logical place to start. And that's a, again, if you take the person who runs those workflows, take a director, a manager,
[00:26:21] You know, CO, whatever, whoever's like in charge of the workflow and they have a high degree of AI literacy, they can look at their own processes and say, okay, I actually understand now like 30% of this workflow can be automated through different tools. They can optimize an existing workflow. The harder part is to come in with a very high level of AI literacy and competency and say, this workflow doesn't even need to exist.
[00:26:47] This role doesn't even need to exist, and that is like a, a higher level. That goes back to the previous conversation about this idea of like a forward deployed engineer or someone who is. Extremely advanced in their [00:27:00] understanding of what's possible that may look at like entire workflows and campaigns and say, let's just go to first principles and reinvent this entire thing.
[00:27:08] so yeah, I think that there's this combination, but if, if your. Not able to look at your own workflows and tasks and understand where AI fits. That would be a pretty good indicator. You're, you're at the very beginning of your own transformation process. If you look at your workflows and other people's and just like immediately want to fix the thing and like, oh my gosh, I could in five minutes, I could save you like 10 hours a month.
[00:27:35] You are likely like further along in your own transformation journey and you want to like figure out how to impart that knowledge on other people.
[00:27:42] Cathy McPhillips: Yep.
[00:27:44] Cathy McPhillips: Number eight. I operate multiple businesses across transportation, tourism, and hospitality. And too much still depends on me personally. From customer inquiries and quotes to scheduling, invoicing, and coordinating staff.
[00:27:56] Where should I start to make these businesses less dependent on me? How do I [00:28:00] still decide what to automate, what employees should handle, and what needs my judgment?
[00:28:04] Paul Roetzer: Wow. I guess the first step is did you want to hire people? Like are you, you know, if, if there's that many things depending on you, it seems like part of this is like you just didn't have enough staff surrounding you.
[00:28:16] but if your goal is to continue to operate multiple businesses without having to add staff, those are I guess, two different paths I would suggest taking. So if, if your goal is to maintain. lower level of, staff. Then I would start to look at the things that you, that are keeping you busy, that are data-driven and repetitive.
[00:28:43] And I would, look at ways to automate as much of that as you could. Going through some of the things we previously discussed where you're analyzing workflows and tasks and you're just saying, okay, you know, here's how I can plug in these different AI capabilities. Here's the agents I can build pretty quickly without needing expertise like technical expertise.[00:29:00]
[00:29:00] I would, I would start there. yeah. In terms of like employees, it would actually come down to an assessment of your current staff and how AI mature they are and how, how much they could use. But again, I would say it probably just starts with AI literacy, understanding the current. Capabilities of what these models can do, and then understanding what they're gonna be capable of doing over the next like six to 12 months.
[00:29:24] So you sort of take a forecasting lens to this and say, what's the impact gonna be on my team and my operations over the next 12 months? As the models get smarter and more. Capable. And then from there you can go through and do an assessment. So we have JobsGPT is the tool we use. It's a, it's a custom GPT at the moment.
[00:29:41] We're gonna be converting it over to a plugin, it looks like is the short term solution. but if you go to SmarterX.ai/jobsgpt for at least the next 30 days until OpenAI retirees gpt. You can use it as a GPT and you can just put a job title in [00:30:00] or you can have this exact conversation.
[00:30:01] Plug this question in to JobsGPT. The other thing I would say is like, if you're not on ChatGPT, you're on Claude Gemini copilot. You could literally take this exact question and just put it in there and say you. Function as like a high level, business consultant. And help me think this through, and then just give it more context about each of the businesses.
[00:30:21] You can just say like, here's my five businesses. Give it links to it. help me think through how to, you know, infuse AI into aspects of this. And it'll, it'll be, it'll function as a pretty good advisor without having to use jobs. GPT necessarily.
[00:30:34] Cathy McPhillips: And if you think about an operator who, you know, customer relationships and customer increase and things like that, like that's something really that people hold near and dear to their hearts.
[00:30:42] Like that seems like some, but then I also think on the flip side, customers have an increasing appetite for just getting a quick answer and knowing it's going to be ai. So just finding that balance of, I need to let this go and this will help save me 10 hours a month, like you said, and our customers get the answer they want, [00:31:00] but I'm still maintaining that relationship with them, that it's a.
[00:31:03] It's a fine line and it's something we're trying to figure out too, you know?
[00:31:06] Paul Roetzer: Yep.
[00:31:08] Cathy McPhillips: Okay.
[00:31:09] Cathy McPhillips: Number nine. If you're a marketing consultant offering AI enabled services like content operations and repurposing how much client education should be part of the engagement. What do clients need to understand before buying and where do you draw the line between explaining your approach and providing training?
[00:31:26] Paul Roetzer: First, you have to check your contract if you know if you're allowed to use ai. So I mean, I one as a consultant, and it could be a freelancer, consultant and agency, whatever it is. If you're working with bigger enterprise. There are most likely clauses in there that say you're not allowed to use AI, or that you have to disclose the use of ai if you do, and you have to get their approval.
[00:31:47] So for starters, this conversation should happen before you create any work products for them. They need to be clear on transfer of copyright and things like that, that if you're using [00:32:00] AI to especially create content, this conversation is a very important one to have with them. That being said, you, you then have to get into, if you are a stop gap for them until they learn how to do this themselves or if this is a sustainable business model.
[00:32:19] And I say that because like, I mean we'll talk about this on episode 252, but like HubSpot just announced a marketing studio yesterday. Marketing Studio is like an on-demand content creator is how they're basically building this. I don't know how they're positioning yet. I haven't dug in deeply. But the way that the trailer video plays it out is in essence, like, just go into the studio each morning and talk about what you want to create that day, and their AI creates it.
[00:32:49] So what does that mean to people whose job is to create content for people? Like I don't know, like this is, we're, again, we're all sort of like figuring this out [00:33:00] together, but the history of agencies and freelancers has often been. Yeah, we could do the work internally. We just don't want to. And, and so the question is, you know, finding the right clients who, even though they have access to automate the creation of this, they value having an outside trusted partner who is the one.
[00:33:20] Doing this and using the tools and deciding when it should be human, when in the lead, when it should be human in the loop. Those are messy discussions, but there are absolutely going to be companies that know it's possible to automate much of what they used to outsource, and they're going to get rid of their agencies and consultants.
[00:33:40] Cathy, you and I have been in the room with marketing executives who have said this exact thing. They're basically wiping out their agent port agency portfolio, and they're just gonna bring everything in-house. So it just gets harder to, to like find the right fit for those services. And then there's gonna be just lots of pressure on the pricing model around [00:34:00] that.
[00:34:00] because if they know what it costs you to create it, and if your agreement with them is you're mostly using ai. It, it, it's hard to know how to price that efficiently. So, tough, difficult. We are in the middle of an evolution of how professional services firms work with clients across every industry, not just marketing.
[00:34:19] Same thing's happening in accounting and legal and everywhere, and it's a lot of unknowns ahead.
[00:34:25] Cathy McPhillips: So maybe providing training is the way to go.
[00:34:27] Paul Roetzer: I, if you can make the money doing workshops and teaching them how to do it, and if you can grow that business, it's a much probably more stable business model, right.
[00:34:36] Moving forward.
[00:34:38] Cathy McPhillips: And you think about that content creation through HubSpot, you're not extracting data to put into an AI to ask you how, what you should be writing about it actually has access to all of that right there. And it's just, it's writing it based on what it already knows, what's performing. Yeah.
[00:34:51] What's getting open. All of those things are already tied, baked into it.
[00:34:55] Paul Roetzer: Totally.
[00:34:56] Cathy McPhillips: Yeah. It's crazy.
[00:34:57] Cathy McPhillips: Number 10, an audience member working in higher ed and nonprofits shared this concern. Alumni may accept commercial brands using AI for marketing, but expect something different from their alma mater. How should institutions respond to that concern?
[00:35:11] What does responsible AI look like when personalized outreach to alumni and donors while protecting their trust?
[00:35:18] Paul Roetzer: I mean, I think authenticity matters across any business really. So, I mean, this is certainly a question related to higher education nonprofits and maybe there's like a higher expectation of authenticity.
[00:35:30] There'd have. To think about that. I'm, I'm not sure if that's true. I always, you know what I've said about branding, I used to spend a lot more time thinking about brand and brand strategy in my early days running my agency and stuff. and I have no idea if this is something I original thought I had or something I read somewhere.
[00:35:47] But I did a lot of crisis communications work also. And the thing we would talk to clients about is trust takes a lifetime to build and a moment to lose. That is relevant to like literally every brand and really every [00:36:00] person. And so if there's an expectation of authenticity from a communication channel, regardless of what industry you're in, you have to think through how you'll use AI.
[00:36:11] Now, I mean, I'm a donor to, you know, to universities. I, if the, if the communication is personalized to me based on past donations and things that. I don't know that I would look at that and think, oh, they must have used AI to like, know how to personalize that For me, personalization done well is extremely valuable to, to anybody.
[00:36:34] Like, I want you to have context and not waste my time with the same offer you're sending to everybody else. Like, I actually would rather know that you've taken the steps to like right. Know more about me personally and target and, a communication based on my interests and things like that. So. Yeah, I don't know.
[00:36:51] I think it's just like understanding where authenticity really matters, understanding where disclosure of use of AI matters. But [00:37:00] I don't know. I mean, I spend a lot of time with advancements officers at high schools and colleges, and. I think that there's very elegant ways to use AI to capture information.
[00:37:10] Like I, here's an example, and again, I'm just like spitballing here. I have a public persona. It's not hard to like learn things about me. I talk about my podcast for an hour and a half every week. Like, there is not, nothing's stopping. I'm gonna, I'm gonna, man, I'm gonna regret this. I'm gonna start getting targeted with all this shit.
[00:37:27] there is nothing stopping an advancement officer from building an agent. That just every morning or once a week goes and figures out any interviews I've done, any transcripts I've put out there from speaking and curates all that information, and then every day or every week sends that advancements officer ideas of personal outreach to me and like drafted something for you.
[00:37:49] Hey, he said this on the podcast. This might be super relevant. He mentioned the university on the podcast this week. Send him a note saying thank you. Those are the kinds of things where it's like super authentic. [00:38:00] You have now used AI in this extremely elegant way, but you can do the exact same thing for any sales or customer successful.
[00:38:08] This doesn't have to be fundraising and stuff like that. you can, if you have targeted like an A BM approach at an individual brand level, you can now monitor stuff in really advanced ways. And then have the personalized elements like handed to you and then you just, you could take it the final steps and I don't, I don't know that anybody would know the difference.
[00:38:29] I think they would actually appreciate that, so, right. I think it could be done Well,
[00:38:33] Cathy McPhillips: who needs Google alerts when you could just build an AI agent to track Paul?
[00:38:37] Paul Roetzer: I mean, go back to 2016 or 17 when we first started trying to identify use cases through our institute for ai. You had to go find like specific companies that did this stuff?
[00:38:49] Yeah, like there was one I remember very specifically that was like a competitive monitoring tool and it would alert you like any changes to somebody's website that day, personnel announcements, funding news, [00:39:00] product now, whatever. I can spin that up in 10 seconds and get it automated now. Like that was a, that was like a thousand dollars a month license to a tool 10 years ago
[00:39:09] Cathy McPhillips: that was right.
[00:39:11] Cathy McPhillips: Number 11. You talk about being a change agent when AI has barely penetrated a business. How do you communicate your AI capability, experience, and approach on a resume when the hiring manager has a low baseline of AI knowledge, talking about outcomes without AI undersells you, but talking about agentic workflows, feels like talking about Excel pivot tables to an accounting department still using paper ledgers.
[00:39:32] How do you bridge that gap?
[00:39:34] Paul Roetzer: Yeah, so I mean, any job description is gonna have. Like what the expected responsibilities are and likely, you know, what goals or KPIs you, you work towards and you can take any. Take your favorite AI system, turn it up to the, you know, highest reasoning model and say, Hey, I'm talking to this company.
[00:39:56] They don't do a ton with ai, but I wanna really understand their industry [00:40:00] and the challenges they face. And you know what, maybe try and interpret what some of their goals might be. if they're publicly traded, you can analyze their earnings statements each three months and like, what did the CEO and the CFO talk about at their latest earning calls?
[00:40:14] here's the job I'm applying to. I want to go into this interview with, some prototypes and some sample analyses of like potential business problems and demonstrate how, you know, I'm able to use AI to solve things more efficiently. And so you just walk into the interview. Stuff they haven't even asked for.
[00:40:34] And you find an opportunity to say, yeah, you know, actually, you know, I anticipate this might be one of the challenges that we faced, and I took the initiative to use Claude to do an analysis if it's okay. Like, I, I'm happy to send you this. I went through and verified everything myself and dah, dah, dah, dah.
[00:40:46] And you're just like, they're gonna be like, wait, what? So, yeah, I mean, you can always be proactive in presenting things of value, not just, oh yeah, I use clot all the time. Show me. Like what? What have you built? [00:41:00] What's like something you did that was specific to our company? And that goes a long way with impressing people and opening up their minds to what's possible, not only with ai, but with you as a part of their team.
[00:41:10] Cathy McPhillips: Right. And it also just, if they're apprehensive, if they're scared, if they just haven't, if they're just misinformed or ill-informed or whatever the case may be. Coming in with a use case like that, they're like, oh, okay. Like that's, it just makes it so tangible and just like. Easy to understand.
[00:41:27] Paul Roetzer: Yeah. And you can do the exact same thing as a change agent within an existing company.
[00:41:31] Not looking, you know, to move companies, but if you're just trying to like make the case internally, hey we should be doing more with ai, do the same thing, like approach it as like, let me just make a business case for this. Build a 10 slide decks, like I get 15 minutes. I wanna show you like a couple of ways.
[00:41:45] I think we can drive innovation if we like do more and I wanna raise my hand and like be a part of that.
[00:41:50] Cathy McPhillips: Yep, for sure.
[00:41:52] Cathy McPhillips: Number 12. For someone making a mid-career pivot into operations or project management from a non-traditional background, what practical AI capabilities should they prioritize over the next six to 12 months?
[00:42:03] What's one portfolio project they could build to demonstrate measurable workflow improvement to prospective employers?
[00:42:10] Paul Roetzer: This is an interesting one, so I'm just like, I'm trying to think about. How project management is evolving for us. So we use Asana as our project management tool and they have an AI capability baked right into it that you can use within Asana.
[00:42:26] We have not, I don't think we have Asana connected to ChatGPT yet. Like I don't know that we've gone to that extent. But, you know, I think that the agentic stuff we talked about earlier, maybe more than anything will Augment what's possible with project management. You know, certainly automate elements of it.
[00:42:44] There's a lot of project management that is repetitive task, repetitive steps. But, you know, imagine if you, I don't know, I always have to think in like tangible terms. So imagine we're doing a product launch and within that product launch, there's elements of it [00:43:00] for marketing, sales, customer success, operations, finance.
[00:43:03] Like everybody plays a role at different stages of that, and you're trying to manage that entire process. You could, you could go through and set up a lot of like automations, you know, of reminders and like the basic things. But you could also use AI to analyze where we're at, where are we falling short, who's specifically falling behind in what they're doing.
[00:43:24] recommend next best steps so that it's surfacing to you. So imagine every morning you wake up to like, a control center that is suggesting to you as the project manager. Who to talk to, what steps to take, drafting emails for you. if there's just a barrier like, Hey, I know someone's out for the week and they're not gonna get this done, can you create the work product with AI And just like, Hey, I just, I know you're really busy, Cathy, I took the initiative to just create this for you, do what you want with it, because that goes to this whole idea of work compressing where all of a sudden the project manager who knows how [00:44:00] to use ai.
[00:44:01] Can assist any of the roles. It's like I can kind of do marketing now, to a degree. So yeah, I think it's like stepping back and just saying how is the overall role of a project manager going to evolve when they are given AI capabilities and they know what those systems are fully capable of. Right.
[00:44:21] and the whole idea of workflow assessment, like looking at workflows people are doing and saying, Hey. I know you built out these 20 steps to the part of this overall launch. I've took in the initiative or taken the initiative to analyze this. I think we could use AI to do like 10 of these and I might be able to save you like five hours.
[00:44:38] Like Great. Let's, I'll take that meeting. So, yeah, I think it's like stepping back and saying what is the future role? And, and actually, JobsGPT, which I mentioned earlier, there is a, conversation starter that says Forecast new jobs is one of the, one of the things, if you click on that and say project management for this type of company, what is that role gonna look [00:45:00] like in the next one to two years?
[00:45:01] It'll actually help you brainstorm how a project management role would evolve.
[00:45:05] Cathy McPhillips: Nice.
[00:45:07] Cathy McPhillips: Okay. Number 13, Paul, you've spoken about why you don't use AI for your writing. A human brings accumulated personal history, values and blind spots to their voice. Current AI models, recombine statistical patterns. Do you think AI will ever reach the point where it gains true life experience and writes as, create creatively as a talented human?
[00:45:26] Paul Roetzer: I think that we've already kind of approached where. With the right model, with the right knowledge base, the right memory, the right access to your previous writings and what you do and your emails, like if you've given it access to everything. pretty reasonable to say today's current technology is can be indi indistinguishable from the human writing and.
[00:45:56] I think it's more about your personal knowledge that it wasn't you, and [00:46:00] that like the writing is part of your process and it's why, you know, it's how you think. And so I actually have a reasonable degree of confidence that if I wanted to use AI to write my stuff, like my editorials and things like that, as long as it wasn't drawing on a lot of personal anecdotes and things like that.
[00:46:21] I could probably do like 95% of that writing with AI today, and people probably wouldn't know the difference.
[00:46:29] Paul Roetzer: So, yeah, I think it's just, for me, it remains a personal choice of confidence in the material of being able to do stuff like this unscripted without ever seeing the questions before we even talk about it.
[00:46:40] Like, if I was letting AI write everything and it was creating my talking points for the podcast, and it was doing my editorials and it was creating my presentations and like. It was doing all of the thinking. I would not be sitting here doing 45 minute unscripted answers to 14 questions I've never seen.
[00:46:58] Cathy McPhillips: Right.
[00:46:58] Paul Roetzer: And I think that [00:47:00] the lesson there is, in any stage of life, if you want to have confidence in a topic and develop some level of expertise and be able to be viewed as someone who understands something deeply, there is no replacement for doing the work. And for me, the work is writing. So it doesn't matter if it can sound like me, it probably can.
[00:47:19] Cathy McPhillips: Right.
[00:47:20] Paul Roetzer: I just don't want it to.
[00:47:21] Cathy McPhillips: Yeah. And like we've also talked about taking notes in meetings. Like we have notes, you know, notes, apps, taking notes for us in the meeting, giving us a synthesis. I still take notes. You still take notes because that's how I process things.
[00:47:33] Paul Roetzer: Totally.
[00:47:34] Cathy McPhillips: And for me to be able to remember that.
[00:47:35] And same thing with with writing.
[00:47:37] Paul Roetzer: Yeah. And even, I mean, even before note takers, like if you think about it, if. Like, let's say you attended a meeting, Cathy, that I wasn't there. Me reading your notes is helpful, but like I don't get all the context right. I don't like hear the tone of people's voices and the inflection and the pauses before they answer and like you don't get all that from a [00:48:00] transcript.
[00:48:00] And so yes, like for me there's the part of just showing up and then the process of inputting the information for me, I have to type it like I'm not a big like voice to text guy. I find like the retention has to come from actually, like, fingers on keys. I don't know, it's, it might be weird that way, but yeah, I think that there is just an argument that for critical thinking, for retention of information, doing the work is always gonna be required.
[00:48:28] Cathy McPhillips: Absolutely.
[00:48:28] Paul Roetzer: I hope
[00:48:30] Cathy McPhillips: same. All right.
[00:48:31] Cathy McPhillips: Number 14, a rapid fire hot take. Do you see top performers negotiating token usage and compute stipends into their job offers?
[00:48:40] Paul Roetzer: Mostly only in AI research right now and probably software development. I could see it would be an outlier if you had someone in marketing, sales, customer service operations, that that was negotiating token usage.
[00:48:54] That would take a pretty advanced user of AI to understand that. But I can also see within ai, Ford [00:49:00] organizations like ours, we're running up against it all the time already. Like I get alerts every like three days in ChatGPT, like someone's requesting, you know, more credits or usage. we run into credit limits all the time at HubSpot.
[00:49:11] Like, so it's, it's coming. and I think even for 2027, most organizations should be thinking about what are their token budgets. And I believe we actually have a topic. For episode 252, about how different companies are approaching token usage and the limitations they're putting on it. But no, it's an outlier if you have a non-technical person inquiring about compute and token usage.
[00:49:38] Cathy McPhillips: And that was it. one of our class attendees who asked that question. So they might be thinking about it too.
[00:49:44] Paul Roetzer: Yeah, I mean, if, if you ask that question in most companies, you're probably gonna get a blank stare. But I do think that if you're a CFO or a CIO or a head of HR, it would be good to be [00:50:00] proactively starting to think about this kind of stuff.
[00:50:01] Right. 'cause it is gonna come up
[00:50:03] Cathy McPhillips: for sure. All right, Paul, thank you as always, and thanks everyone for listening. we've got a busy few weeks here at SmarterX, so if you're around, visit, SmarterX.ai. You can find some webinars on our homepage. You can go to SmarterX/ai/webinars and take advantage of that POD250 offer for academy and we will see you.
[00:50:21] You'll see Paul and Mike next week, next Tuesday.
[00:50:24] Paul Roetzer: All right, thanks
[00:50:25] Cathy McPhillips: Cathy, and thanks everyone for
[00:50:26] Paul Roetzer: joining us. Thanks for listening to AI Answers to Keep Learning. Visit SmarterX.ai where you'll find on-demand courses, upcoming classes, and practical resources to guide your AI journey. And if you've got a question for a future episode, we'd love to hear it.
[00:50:44] That's it for now. Continue exploring and keep asking great questions about ai.