Pod Insights The Brand Called You AI at Work: Insights on Adoption, Safety & Transformation with Kate Marshall, Founder of TheGr.ai About this episode Welcome to another episode of TBCY, where host Ashutosh Garg is joined by Kate Marshall, founder of TheGr.ai and author of AI at Work. In this insightful conversation, Kate Marshall shares her expertise on artificial intelligence adoption, the skills needed to thrive in the AI-driven workplace, and practical advice for professionals at every stage of their AI journey. Discover why AI literacy is no longer optional, how to navigate organizational challenges with AI tools, and how to protect your data in an evolving digital landscape. Whether you’re an AI beginner or looking to optimize your workflow, this episode delivers actionable strategies to harness the power of AI effectively.
Transcript 0:08 Speaker 00 Welcome to another episode of the brand called you a podcast and podcast show that brings you leadership lessons, knowledge, experience and wisdom from thousands of successful individuals from around the world. I'm your host, Ashutosh Gargan. Today I'm delighted to welcome a very senior professional with a deep understanding of AI from Denver, USA. Kate Marshall. Kate, welcome to the show.
0:31 Speaker 01 Thank you so much for having me.
0:33 Speaker 00 Thank you. Kate is the founder of The Gray, a consultancy focused on AI adoption, training and governance. She's a fractional CAIO and she's an author of a book titled AI at Work. And we've got to be talking about her book today. So Kate, before we come to your book, a few questions for you.
0:55 Speaker 01 Yeah.
0:55 Speaker 00 You've spent over 20 years in cybersecurity, training and project management. What were the defining moments that shaped your approach to teaching technology adoption?
1:07 Speaker 01 Yeah, thank you. So with this emerging technology, one of the things that we have seen is just how sticky and how challenging the adoption process is. And that's the same thing in cybersecurity and cybersecurity training. So in my past world, I focused on taking really technical topics and making them understandable for the the normal consumer that had to pay for the training and then they would bring their their technical groups in. And it really translated well to what's happening in AI in this era right now of it just when you look at terminals and command lines and trying to get a sense of how these
1:51 Speaker 01 LLMs work, it's, it's not as hard as it looks. Once you get in there, it's just getting people, you know, to jump in and try it and play.
2:01 Speaker 00 Amazing. And what patterns have you noticed which in professionals were struggling with new technology? And how did these insights influence your work in AI adoption?
2:17 Speaker 01 Yeah, awesome question on that one. I
2:21 Speaker 01 I think the biggest thing that I am still seeing as a struggle as an issue is time. Uh, people having the time and space to learn these tools and because they're evolving so quickly. Uh, Claude alone has come up with, I think, twenty changes to their platform just in the past six weeks. And so as soon as you learn something, as soon as you build something, it's changed, it's bigger, better, faster, which is a good thing. But it's really overwhelming. And so that's what I see as the most challenging thing. And the people and the organizations that actually create time and space for people to learn these new technologies. They're the ones that are really seeing the transformation in their people.
3:05 Speaker 00 Hmm. Very interesting. And what prompted you to start the Gray? And what gap were you trying to solve in the way organizations approach AI?
3:15 Speaker 01 Yeah. We are giving people keys to a car and we're not teaching them how to drive. And that's really the biggest challenge that I continue to see, and that's why I started the Gray. We were giving access to different tools, different large language models. And there were a lot of training classes on this is what it is, this is how the car works. Um, you know, this is the, the backend of the engine. And that's not really helpful to especially your non-technical professionals that are just trying to learn how to use them in their daily work. And so what I started doing, um, back when I was, you know, at the cybersecurity training, um, organization
4:00 Speaker 01 was really helping people learn how to drive, um, not how the car works, but how to drive um, and, and understand how to implement these things in their real work. Oh, and there's a massive need at this point. I, I don't know anybody, myself included, who, you know, doesn't need help learning how to optimize and how to integrate these tools into work.
4:22 Speaker 00 Absolutely. And there is a huge need indeed all over the world.
4:27 Speaker 01 Yeah.
4:27 Speaker 00 Here, let's come to your book, AI at Work.
4:30 Speaker 01 Yeah.
4:31 Speaker 00 You say the gap between I should be using AI and I am using AI keeps growing. Why do so many professionals remain stuck in that gap?
4:45 Speaker 01 There's a couple of reasons and one of the bigger ones is people don't really know what's safe versus what's allowed versus what's not allowed. So organizations are struggling to communicate what they want to see from their people. And I'm, I'm looking at this from a corporate professional organizational lens right now. And so there's not a great communication path on these are the tools that you can use, these are how you can use them, you know, here's your AI policy, we have an AI manifesto. It's really muddy right
5:27 Speaker 01 now because organizations, they're figuring it out for themselves right now, too. And so I think that is the biggest gap that I see. And then I also see a lot of organizations think, okay, we just turned on Copilot, Microsoft 365, check that box, everybody's got access to AI right now. But I don't think that's a great representation of the scope of what you can do with these AI tools. And so there's a lot of frustration of people saying, I really want to try Claude and I want to, you know, have ChatGPT on my system and their organization is saying, sorry, we've got Copilot, go have fun, you know,
6:08 Speaker 01 and, and Copilot has its uses. It is a, it is a good tool because it's deeply embedded in the Microsoft ecosystem, but it is, it is not a great representation of all the AI tools that you can use at this point. And so I feel like there's a lot of overwhelm and there's a lot of gap in what tools can I use? I need to stay safe? How do I, you know, protect data? How, and, and really, again, how do I figure out how to use these things?
6:37 Speaker 00 Great response, you know, and I was going to ask you a follow up question on this one because I've seen this in many organizations. Leaders or leadership of the organization say, we have signed up with Claude or with ChatGPT or something else, and someone else wants to use something else. Now that becomes a conflict and therefore the gap in my personal view seems to grow even wider. Because they want to use AI, but not necessarily the tool that is being, you know, forced upon them, if I can use that term.
7:11 Speaker 01 Sure, and and to make it even more challenging, these tools are evolving so quickly. And so I have clients that are constantly chasing the latest greatest, you know, greatest version of something else. And so, you know, they will deeply embed their memory and they'll turn on GPTs and ChatGPT. But then the new shiny thing comes out in Claude and they want to switch over to that. Asking an organization to be able to chase these models, it's too much. And so really, it's you choose an enterprise version, you choose the best possible one for the current times and where your organization is going. And then unfortunately, you stick with it.
7:55 Speaker 01 And I know that there's there's tough frustrations with that, with the things you want to do. But there are ways that you can still play with the latest and greatest tools, just not with your corporate data. So one of the things that I especially note in the book is, look, if you are forced to be used Microsoft Copilot at work, put Claude on your system at home or put it on your phone and learn some of these other tools and just don't co-mingle data.
8:24 Speaker 00 Amazing. The idea that AI literacy is no longer optional is very powerful. But tell me, Kate, what does true AI literacy actually look like in practice?
8:38 Speaker 01 Yeah, it's evolving, it is constantly evolving. But the one thing that I will. Compare it to really is it's like an exoskeleton at this point. So we have evolved now from using our L L Ms as like fancy Googling and just treating it like search and now A I fluency really looks like A I plus human augmenting work, doing it together, getting the best out of both model and people and we're and really turning into full orchestration at this point. So there's no great definition of what true A like A I literacy is because it's
9:23 Speaker 01 going to evolve week by week depending on the new features that come out, but it's evolved beyond just prompting, just, you know, content creation and it's really about augmenting that work and orchestration together.
9:38 Speaker 00 Amazing. One of your key principles is you own the outcome.
9:45 Speaker 00 How should professionals rethink accountability when they're working with A I?
9:50 Speaker 01 Yeah, yeah. I am seeing people, whether it's in meetings, whether it's in their emails or their presentation, they will treat the output that comes from A I as, you know, the gospel and they will even have a hard time defending it or say, oh, that's just what the A I gave me. That's not, that's not a good answer. And so, you know, A I is the tool, A I is your thinking partner, but you are the professional and the output is always under your name. You're always accountable for it. That should be full stop. And so we're seeing people, you know, that are, they're producing incredibly long emails or, you know,
10:35 Speaker 01 just automating their blogs and they're not even really reviewing or absorbing the content that's created. And so I work with people to step back and rework it. So you're not just putting something in, getting something massive out, doing a quick skim and then shipping it. It's really.
10:56 Speaker 01 Communicating with your LLM, using it as a thinking partner. And so that way, the output is not surprising to you. And you, you know, you own the outcome of what the output is.
11:08 Speaker 00 Amazing. Yep. You also recommend choosing one LLM and sticking with it for three days, which I find it difficult to do myself.
11:18 Speaker 01 Yeah.
11:19 Speaker 00 What changes do you see when someone actually commits to that discipline? And how does one handle hallucinations of one LLM? Because I tend to ask something from one LLM and then copy paste in another and say, do a quick fact check. And more often than not, there are errors.
11:41 Speaker 01 Sure, sure, sure. Well, right now, I've kind of evolved my thinking on this one a bit. I firmly believe that you should choose an LLM, stick with it for at least a month. You need to train it on things like your tone of voice, your role, the way that you like to communicate. You know, there are, there are things that you can put into memory. Say, please don't do the M dashes. You know, you still see that or the if this, then that kind of context memory. But what I'm recommending now at this point is you choose your LLM for work, or oftentimes that one's chosen for you. And then you choose a different one for personal life and they need to get to know you.
12:26 Speaker 01 So if you are talking to somebody and they just have that generic, out of the box version of ChatGPT and they're not loving what, you know, the output is, or they're, they're saying it's not working for me. It's because it hasn't been trained on you, your style, the way that you like to work, the things that you care about. And so you really have to get in, interact with it, have it understand you. And then that will really help it become your superpower.
12:57 Speaker 01 And the free versions are not the full representation of what an LLM can do. And so I also recommend like. Yes, pay the twenty dollars a month for a plan because you get the latest and greatest models, you get the best features, and you have that ability to have it really understand your memory and get to train it. And now that we're seeing skills pop up, those skills that you can upload in those markdown files, it's getting that much better at really becoming your your partner in work.
13:32 Speaker 00 But you know, someone has asked me an interesting question on this twenty dollars. They said, oh, I pay twenty dollars, but four of us use the same chat G P T account.
13:42 Speaker 01 Oh, no.
13:43 Speaker 00 How on earth are you going to train that L L M for four people?
13:48 Speaker 01 Yeah.
13:49 Speaker 00 What you're suggesting is everyone should have their own account.
13:52 Speaker 01 Hundred percent. And there are times where I will actually turn on incognito mode, and it's just a quick little button, because I'll be talking to my kids about something completely random, and I don't want the chat G P T memory to think that I care about it. So Godzilla, you know, has nothing to do with my work, but my kids, you know, like to learn about all of the different things with Godzilla. But no, it's it's a great point on you should absolutely not share these things if you want it to work the best way that it can for you. And then addressing hallucinations, because I didn't I didn't
14:33 Speaker 01 address that in that earlier question, these models are getting better and better at not hallucinating. But if it is a project that you you need to know that it is factual, I always just ask for its sources. And perplexity has been the star of that one for a long time. But now you can prompt even, you know, Claude or chat or, you know, Grok does a great job at this, too, where it will actually give you the sources. So you know, and you can look further for yourself. But I, still at this point, I would not just trust outputs, you know, especially if they're important.
15:10 Speaker 00 Yeah, you know, I tried this for my latest book. I was looking for multiple quotations from different scriptures. And one of these L L M's gave me half a dozen of them. I copied all of them, put them on a perplexity and perplexity, but none of these, none of these exist.
15:30 Speaker 00 I went back and I said, these are all wrong. And they said, yes, you're right. They're all wrong. You know, so.
15:35 Speaker 01 And thank goodness you did that.
15:37 Speaker 00 Yeah. Otherwise, you know, I would have had to explain to my publisher where I got these ones from.
15:44 Speaker 01 Yeah.
15:44 Speaker 00 But get back to you. You strongly emphasize data safety. And what is the biggest mistake professionals make? When AI tools at work are not being used well?
15:59 Speaker 01 Yeah, um, so if employers don't have an enterprise system turned on for their people, or if they don't have AI policies that are clearly outlined, the easiest thing that you can do is put in PII, just that personal information from clients into an LLM where it's not set up correctly. It doesn't have the guardrails and the boundaries in place. And then that information can, you know, go out to the universe because we're we're training these models on on other things. And so one of the things that I really stress is you have to think about the data that you are feeding it and what you are giving it.
16:41 Speaker 01 And what I am seeing now is people are doing a much better job about thinking through what they're putting into their LLM, but they're not thinking about these other tools that are more ancillary. So a good example of that is it voice dictation tools. So I use one called Whisper Flow. I use it on everything. I love it. It's it's done a great job increasing just my my speed and my processing time. But all of that information goes up into a cloud. And so unless you have something that's either enterprise version or private, that data, that customer information that I could just be speaking and not thinking about where it's going, it could be at risk.
17:23 Speaker 01 And so it's really just being thoughtful about, look, if this gets out into the world, what's the risk associated with it? And this is where organizations need to be really thoughtful about how they're allowing the tools to be used, what the policies are, and getting on these enterprise versions.
17:44 Speaker 00 Very interesting.
17:47 Speaker 00 Your trust check reject model is very practical.
17:52 Speaker 01 Yeah.
17:52 Speaker 00 How can someone apply this quickly in a real world situation?
17:58 Speaker 01 Oh, I love that question. And when I talk to people about this, they, there's this sinking feeling of you're telling me that everything that I do has to go through this model. And our brains are great at doing this in milliseconds. And so this isn't a, you know, a mental habit of every time you put something into an L L M, you have to sit back and think about, you know, what the data is and, you know, check if it matters. This is very much a, look, is this an email that is very low risk? Or is this a, you know, project that is going to go out to a ton of coworkers? Is this something that could us get, could get us in trouble legally?
18:41 Speaker 01 So really, it is a super short, one second, maybe two seconds tops, thought of, okay, how high risk is this? Can I easily just trust the model? And how well do I know the information? Do I need to check things like as, as your example that you did with the book? Or, you know, completely reject it and, you know, think through your, the, the data that you have already. But I'd say 70% of the time, we're already doing this subconsciously anyway.
19:16 Speaker 00 Correct. Well said.
19:17 Speaker 01 Yeah.
19:18 Speaker 00 You describe AI as something that should feel like a steady advantage.
19:23 Speaker 01 Yeah.
19:24 Speaker 00 Rather than a disruption. What does that future workplace look like?
19:32 Speaker 01 I don't know. I don't know. Um, and so what, uh, what I do know is there will be massive disruption. And the future is about to get really weird as we're bringing in AI agents into the organization. And there is this whole mess of who is going to manage the agents, who is going to train the agents, who is going to partner with them on the output? Um, what's going to happen when you have an agent that manages you, you know, and as a human, what is that going to look like? And so when I think about AI feeling like a steady advantage, it's really
20:15 Speaker 01 Steps. It is steps to becoming better at your role using AI tools. And having that advantage over the organizations that aren't thinking through it, aren't doing those steps. But but there will be massive disruption. And I don't know anybody that can clearly clearly say and clearly define what that's going to look like. It's just it's just gonna get weird. So but that that goes back to the point of the the AI plus human exoskeleton. It's how quickly can you get to that point where AI is your thinking partner? It is your buddy in all things to be able to get the best out of both you and the models
21:00 Speaker 01 to to get great work done at speed and scale.
21:03 Speaker 00 Well said. I have time for two more questions for you, Kate.
21:08 Speaker 00 My next question is, do you believe AI will widen or reduce inequality in the workplace? And moving further, you and I are sitting in countries where we have access to a lot of AI with every possible tool. Do you think this is going to widen the gap between countries?
21:29 Speaker 01 I don't, actually, because it is now becoming more and more accessible to all countries, all people. When I think about knowledge and how, you know, a hundred years ago, knowledge was it was power, it was for the elite, it was, you know, the the thing that got you ahead. Well, now everybody has this knowledge and the the value is going to zero, which on one hand, I'm really fearful for colleges and universities of, you know, what those need to evolve to look like. When, you know, it's not about the knowledge, it is about how you what you do with it.
22:14 Speaker 01 What are what are your strategies and what what do you use to to do with the knowledge that you have? And I think that's the place that we're in. But when we're talking about inequality, my fear is for women specifically and and these non-technical professionals.
22:34 Speaker 01 We are just wired a little bit differently and and I'm I'm categorizing here, but I I think the majority, when I look at the people that are forward leaning, that they are not afraid to fail, they're experimenting, that is more on on the male's side. And I'm part of academies and different groups, and I'll look at the leaderboard, and it is five percent female that are really leaning into this. And then when you also look at just the the jobs and the roles that are being taken over quickly by AI, it is the admin and support, it is
23:14 Speaker 01 the data entry, it is a lot of those roles that that females historically are great at. And I worry that if we're not leaning forward and training and pushing ourselves and trying and testing and failing and doing all of those things, that is where I will see that inequality widen. Because, you know, it it is that old quote of your job's not going to be replaced by AI, it's going to be replaced by people who know how to use it. Well, right now, what I'm seeing are the people that are really leaning forward to deeply understand how to build these agents, how to build these automations. It it's more male than female, which is really unfortunate.
23:57 Speaker 00 Very well said. Great response. And my last question to you, Kate. If you had one piece of advice to everyone picking up your book, AI at Work, what would you say on how they should navigate AI today?
24:17 Speaker 01 I think you just need to start. I think everybody knows that they should learn it at this point. I I run into very few people that are just, you know, arms crossed, not not interested in incorporating it at all. But it comes back to What are you trying to do with it? That is where a lot of people get stuck. And so the reason that I put the book together the way that I did is because I saw certain patterns of, okay, if you are just starting out baseline, there is a series of checkboxes that will help you get, you know, to the next level pretty quickly. And it is just starting and going through that. And so one of the things that I always recommend is if you have no idea where to start, or
25:02 Speaker 01 if you don't know what you could be doing, turn on the voice mode on any of your L L Ms, all very easy to do, grab a favorite beverage and just talk and just, you know, say, this is my role, or I'm starting from complete beginner, or I'm thinking about this. And it's a, it's a beautiful brainstorming partner. Many times I'll go off on and go on walks and have just really lovely conversations with my A I tool while, while I'm on a walk. And I, I think that's where you need to just start understanding how you can use it and know that we are all beginners in this.
25:44 Speaker 01 So the, the fear of I'm going to look stupid, or, you know, I don't even know what to do. We are all feeling that at all levels.
25:52 Speaker 00 Absolutely.
25:53 Speaker 01 It's really just start.
25:55 Speaker 00 Absolutely. And on that note, Kate, I just want to say, what a fascinating conversation this one has been for me, my viewers and my listeners. You know, you gave us some very interesting thoughts about A I. And yet, the message that came through so strongly and powerfully was, A I is here to stay. You can't duck the, the fact that it's going to be there as a part of our lives. And if you haven't started, get started now. Thank you for speaking to me and good luck to you.
26:26 Speaker 01 Oh, thank you so much for having me.
26:30 Speaker 00 Thank you for listening to The Brand Called You, videocast and podcast, a platform that brings you knowledge, experience and wisdom of hundreds of successful individuals from around the world.
26:43 Speaker 00 Do visit our website, W W W dot T B C Y dot in to watch and listen to the stories of many more individuals. You can also follow us on YouTube, Facebook, Instagram and Twitter. Just search for The Brand Called You.
Previous Mastering Global Leadership: Patrick Brown, Founder & CEO of Unity Communications Next Building the Future of Tech Services: Chant Vartanian (Founder & CEO) and James H. Herring (CIO) of M-Theory Group Page updated November 14, 2026 · transcript and summary are machine-generated