In this episode, Russ Somers, who heads marketing for Quantified, shares insights into leveraging generative AI to boost productivity and creativity within sales and marketing roles. Having built a 'GP team' of custom GPTs, Russ has tripled his output, illustrating how AI can serve as virtual team members for tasks ranging from webinar content creation to specialized knowledge acquisition. Through personal anecdotes and exploratory conversations, the episode delves into the process of building AI team members, the importance of play in learning and innovation, and strategies for incorporating AI into personal and professional growth. Highlights include building AI with specific skills sets like 'Wendy Webinar' and 'Roger RevOps,' and the philosophical implications of personifying AI for better engagement and output. Moreover, Russ's personal journey from a layoff to pioneering AI productivity tools opens a discussion on the transformative power of AI in the modern workplace.
Website: Quantified
00:00 Meet Russ Somers: The One-Man Marketing Powerhouse
01:42 Introducing Wendy Webinar: A GPT Team Member Revolutionizing Content Creation
03:42 Leveling Up with GPT: Beyond Basic Task Automation
05:18 Roger RevOps: A Custom GPT for Niche Expertise
08:07 Exploring the Next Frontier: Collaborative and Mentorship GPTs
14:25 The Art of Building and Utilizing GPT Staff: Tips and Tricks
22:27 Expanding the Team: Integrating GPTs into Human Workflows
23:42 Exploring Organizational Progress and Tool Adoption
25:25 The Importance of Measuring Effort and Encouraging Experimentation
26:50 Fostering Creativity and Psychological Safety in the Workplace
29:06 Personifying Bots for Better Engagement and Output
31:42 Reimagining Brand Communication in a Conversational World
34:58 The Transformative Power of Play and Exploration
38:43 Strategies for Personal and Professional Growth with GPT
47:54 Concluding Thoughts on Innovation and the Future of Work
📜 Read the transcript for this episode: Transcript of How to 3x your output with GPTs: CMO Russ Somers on building a GPTeam |
[00:00:00] Russ Somers: Hi, I'm Russ Summers. I run marketing for Quantified. That's quantified. ai. We're a role play and practice solution using generative AI for sales teams selling complex products. And I've been working With AI for a couple years now, I've built a little GP team of GPTs that assist me and I've used that to 3x my output as a one person marketing shop.
[00:00:25] Jeremy Utley: Start maybe just what, like, tell us about the company. Tell us what you guys do. Tell us, um, what your role is there and then how you make your job that much easier with ChadGBT.
[00:00:37] Russ Somers: . Well, I've been using chat, using chat GPT and custom GPTs for quite a while. And fundamentally, I'm doing it because I'm a personal pan pizza marketing team. Just one person for a small startup. We. create role play experiences to train sales professionals, especially in industries like pharmaceuticals, financial services, et cetera, where you have these complex interactions, complex products, and need to have a really well trained team and know how they're going to perform.
But as the only marketer in the company currently, I have to do a lot of things. So I started, because I'd been playing with custom GPTs, Making these little purpose built ones that I started thinking of as team members. And the first one, we do a webinar series. Our customers like webinars. And the problem with webinars is there are a lot of content to create.
Invitation emails, landing page copy, social posts, sorry I missed you emails. About three hours of relatively mundane copywriting for a webinar. So the first one I built was called Wendy Webinar. I think of her as a little virtual team member. And I give Wendy the abstract for the webinar. And she cranks out first draft of emails, Landing page copy, all of that.
And I built her, Jeremy, to be really prescriptive about what she needs, because both junior team members and GPTs, if you don't give them good, crisp direction, they'll, rather than saying, I can't do this, they'll do a crappy job.
[00:02:14] Jeremy Utley: And when you say, when you say, be really explicit about what she needs, you mean what she needs from
[00:02:19] Russ Somers: you?
Yeah. So if I say, Hey, Wendy, we're doing a webinar about puppies and I need you to write the content, she'll go, I don't have enough information. I need to know the abstract of the webinar. I need to know the date and time. I need to know the speakers, et cetera. That's cool.
[00:02:37] Henrik Werdelin: That's cool. And to be practical, is that something that you put into the description of her? , or have you more generically prompted her of saying, you know, please ask a lot of questions before you get going.
[00:02:50] Russ Somers: Good question. It's in the instruction set in terms of here is Wendy specifically what you need to create a webinar.
And if you don't get what you need, rather than generating any copy, tell me, Russ, this is what I need from you.
[00:03:06] Jeremy Utley: And so that's been informed by you've run lots of webinars in the past. You kind of roughly know. So all of the component pieces of a campaign, follow up, et cetera, you have trained in Wendy's custom instructions, all the stuff you need.
And then you interact with Wendy as basically your webinar guru. And she will start to create based on your historical knowledge of what, what a webinar needs. She'll start to create the copy that otherwise you'd just be sitting there going, you got your outline and now it's get down to work. Now instead it's pass it over to Wendy, have her do a first pass and all this stuff.
[00:03:39] Russ Somers: Exactly so, and it truly does become a collaboration. But the thing I've figured out, Jeremy, is when we look at building virtual teams, or GP teams as I've taken to calling them, That's level one. Level one is what everybody thinks they should do with generative AI, is take the things I don't want to I know how, I just don't wanna, and outsource them to a GPT.
And then I started playing with what I'm calling, this is, this is my informal framework, I'm making it up on the spot. I started playing with level two of custom GPT team members by creating ones to know things that I don't. Because we just described with Wendy. I know how to do a webinar. I'm just lazy and I don't have the time to write all that content and she can help me jumpstart.
What if I create GPTs that know things that I don't just
[00:04:33] Jeremy Utley: like, so what's an example of that or what, when did you encounter the need to have a GPT that knows something you don't, what was it that you realized you didn't know? And what was the process by which you said, what if I could create a GPT that knows this tell, tell us that story.
[00:04:48] Russ Somers: Well, I was working on some revenue operations things. I'm nerdy, but I'm not a RevOps guy. So there were some things in configuring our HubSpot instance that I wasn't familiar with. And I was doing what you usually do, reading the documentation, googling, reading blogs. And I was remembering when I had a really good RevOps guy and I would just go and say, Alex, how does this work?
Show me how to do this. And he, in five minutes, I'd know from him. I'm like, cool. What if I build a custom GPT? And I tell it, your name is Roger RevOps, and you've read all of HubSpot's documentation. Here's the instance, the specific version we have, by the way, here's our specific version of Salesforce, Salesforce, all of our other tools.
Here's how they're integrated. So a lot of work up front to take him, to educate him on that. But it's like, now you're my RevOps guy. Hey, Roger, I need to figure out this thing with the suppression list. How do I do that? Oh, in your version, you do this, this, and this. Super easy. And so he now rides shotgun on that stuff.
We had some, we're BDRs. I'm not an expert on that. So I said, Hey, Roger, Professional development, we're sending you to email school. I need you to learn everything you can about email deliverability. What is SPF? What is DMARC? What is DKIM? How do we avoid block lists? What are honeypots? I need you to know all of that.
You're going to professional development, and when you come back, you'll be an expert. Okay, so wait, how do
[00:06:21] Jeremy Utley: you send Roger to professional development? I just go back to the instruction set and edit.
[00:06:27] Russ Somers: And, you know, it always asks when you go in to edit one, How's Roger doing? What could he be doing better?
And for some reason, I like to have these conversations like they're with people. And it's, we're sending you a professional development, Roger. And three minutes later, he's ready.
[00:06:45] Jeremy Utley: Well, so let's break that down. I mean, you just, you just rattled off a list of acronyms that I'm, I'm nerdy, but I also, I'm not a rev ops nerd.
I don't know what BDR and all those things are. How do you even know what to draw upon? And when you say, Roger, I got to send you to, you know, on professional development, how do you even know the parameters to include in an update to the instruction set for, for, for a task that used. acknowledge you actually don't have the expertise to handle yourself.
[00:07:14] Russ Somers: It does require some trust, doesn't it? I, but again, we're just using these tools and they're just tools to amplify what we would do. What would I do when I was trying to configure HubSpot? I would go read all the documentation. And filter it through Common Sense and what I found from Googling. I can totally do that, it's just a really long and slow process.
So all I did was tell him, execute that process. Read their documentation. Read blogs, develop an opinion.
[00:07:47] Henrik Werdelin: What do you think might be level three GP staff? You have, you know, stuff that I know how to do, but I can't be bothered to do. You have stuff that I don't know how to do and it'll take me too long time to learn.
What do you think is like the next one?
[00:08:04] Russ Somers: Next one, hopefully is nirvana, but I don't think we're there yet. Right now, my My straw man on it that I've been playing with is collaboration, and collaboration in two ways. One is, one of the GPTs I built is a mentor, her name is Maya, and she's like, studied all the great marketing thought leaders.
So when I need to kick ideas around, Hey, Maya, this is happening. I'm not sure what to do next. I can sort of play a game of ball with her and go, this is what's happening, what would you do? She'll go, well, tell me more about that. I'll tell her, she'll go, well, you could try this, or Seth Godin would do this, or this other person would do that.
So there's that level of collaboration, but there's a couple of others. Now that I'm, I've made a hire, I'm bringing a new person on, I want us to get on the team's version of GPT, so we can learn what that collaboration looks like, because I don't know yet. And then there's also the ability that I've been playing with, and this is super fun and kind of trippy, to add one GPT inside of another.
So if I'm talking to Roger RevOps, but I need an opinion from Amon Adman, I can say, hey, you know, what do you think Amon would think of this, and add him. And I, I don't fully know how it works yet, but you start to get something like you'd get with two team members in a room collaborating. I need to get my ad guy and my RevOps guy in here to collaborate.
[00:09:33] Jeremy Utley: And you can bring them into the same conversation. How, how, when you describe kind of Maya mentor and the collaboration you have there, where you have quite, you, you, you describe that in context of level three collaboration, how do you contrast that with the mode of interaction you have with Roger or with.
Uh, I can't remember the other one's name, but because to me, those sounded collaborative too. So I just wonder how you think about the distinction between collaboration with the Maya mentor style GPT versus interacting with a Roger RevOps.
[00:10:07] Russ Somers: It's a good question, because it does, when it works well, it's collaborative all the way up the stack in certain ways.
I guess the core thing is, with the others, I'm usually starting with a task. I may know how to execute a task, or I may not, but it's very task oriented. When I start playing with them as mentors, etc., it's a little bit more like, let's throw the ball back and forth and explore ideas together.
[00:10:35] Jeremy Utley: So it's, it's less task oriented and more kind of objective perhaps, or even question or challenge.
[00:10:43] Russ Somers: Question, inquiry, challenge. Those are all great words for it, Jeremy.
[00:10:48] Henrik Werdelin: Do you, have you found anything that it's remarkably bad at? That way you kind of went down something and you go,
[00:10:57] Jeremy Utley: have you had to fire what Hendricks asking is, have you had to fire a team? And do you have
[00:11:02] Henrik Werdelin: a GPT to fire the GPT member?
[00:11:06] Russ Somers: Oh my God.
Well, I've thought about building myself an HR GPT, just to keep myself in line, but I haven't done it. Um, I'll give you a fun thing while
[00:11:14] Henrik Werdelin: you think of that at BARC, we have a very snarky GPT HR lab. Bot called bossy who have the personality of being a little bit of like why couldn't you be bothered looking on the internet yourself?
And so it's it's quite entertaining one.
[00:11:33] Russ Somers: I love that. I love having them have personalities mine all typically have favorite coffees as well just because that way I can reward them by saying Roger go get yourself a red eye because you guys have seen the research that says performs better When you take it.
That's
[00:11:49] Jeremy Utley: so good. And you know, you know what their favorite drink is. That's, so do you have like their favorite cocktail as well? If it's after five, 5:00 PM or how does that work?
[00:11:57] Russ Somers: I
[00:11:57] Jeremy Utley: haven't
[00:11:57] Russ Somers: gone there because I don't want to intrude on their work-life balance. I understand. It's,
[00:12:01] Henrik Werdelin: that's a good, except the problem now is that now that, uh, o my eyes adding memory, you know, they'll remember how many coffee you owe them.
[00:12:10] Russ Somers: This is why, you guys have seen the research that says that ChatGPT performs better if you offer to tip it. I came up with the coffees because I don't want Skynet to have a run in tab on me. I really don't.
If we can wing back, because the question you asked specifically, I'm trying to remember, Henrik, it was on the collaboration aspect. Oh, what they don't do well.
[00:12:33] Henrik Werdelin: Just what they don't do well, because I imagine like, you probably also have spent a bunch of time figuring that out.
And I imagine that people listening to this might go like, okay, at least I don't have to go down that rabbit hole.
[00:12:43] Russ Somers: I mean, the big thing that I've found that they don't do well is In a lot of ways, ChatGPT is a Labrador Retriever. It's so very eager to please. And so, just like a junior team member, an intern that's super eager to please, I think of them a lot as almost being wired like interns, if you don't give them what they need to do to do a good job, they'll still do their best to please you.
And that's why I always include in the instruction sets lines like, Wendy is a super competent professional, expert in her field, she knows what she needs to, to succeed, and she's assertive in asking for it if she doesn't get it. So unless Wendy has a detailed abstract of at least a page, and these other elements, rather than cranking out content, she will say, here's what I need.
And that, that took me a while to figure out because they will, otherwise, they'll just sort of do lousy work and try to please you.
[00:13:46] Henrik Werdelin: You have done an amazing work of setting up these agents, and it seems that a lot of people that we talk to that have a good workflow is kind of doing this personifying different time at work.
But it also seems for many people, it's really difficult to get into. Like there's this. Not this, it's not a technical barrier, because obviously it's relatively simple to make these, but it's almost like a creative barrier of thinking of like, who would I hire to this job, uh, to stay in your analogy. Do you have any advice to offer in how do you get going on the journey?
Like, how do you hire the first GPT staff?
[00:14:25] Russ Somers: Well, a couple of things. Um, first off, I don't think anybody should start trying to build one for work, because We learn fastest and best when we're playing. So I was talking with a friend the other day that wants to get started. She said, where should I start? I said, you love to travel, Kathleen.
Write yourself the world's best travel agent that knows all of Kathleen's quirks, likes and dislikes, places she's visited. Because in that, you're working on something you really love and care about. Not that I don't love and care about work, but you know what I mean. And so that let her play and get started.
So one principle that I would advise Is start don't start with work start with something that feels like play would be to me a pretty important thing
[00:15:14] Jeremy Utley: that I just want to resonate with that for a second Russ because one of the things that I found in Many corporate settings and larger organizations things like that There's kind of a just a lack of imagination when it comes to what are applications at work.
And I think the truth is Most folks don't care that much, you know, and I've said in the kind of innovation space for 15 years, if you don't care, don't bother. Care is kind of an essential prerequisite. And part of the challenge, I think, when it comes to GPT ifying work is ultimately, it's like, I just got to get this done anyway.
I don't really, it's, it's, there's that lack of passion, whereas taking someone personal, and I love your, your point about a personal GPT. I've often said to folks. Just take a personal decision you're trying to make and ask ChagPT to ask you a bunch of questions about whether you should grow your family or move to this place or make this big decision or whatever it might be.
But if it's personal and if it's emotional, it's likely to, to elicit the kind of reciprocity from the user that it's required to make ChagPT do a good job. And I see the same thing. Uh, you call it play, I call it passion or care or whatever it might be, but the same thing to your advice, what are other kinds of, uh, or what's the question maybe someone can ask themself?
You said to Kathleen in this example, um, you love travel. What would you, what would you ask someone if you didn't know them as well as you knew Kathleen to give them a start on a personal GPT creation project? How would you help them know where to start?
[00:16:47] Russ Somers: Good question. I think it, I mean, that comes down to what we learn to do in sales and marketing conversations anyway.
which is to figure out what things people care about. So I think you'd have to start by, I mean, at least what I do is start with saying, so what do you love? What are your hobbies? What were your last great vacation? And look for the things that they light up about. I mean, you've seen my background. You understand that I have a lot of guitars behind me.
Um, I like to play guitar. So the first one I built was a song tutor. To help me pull together tablature and lyrics and chords when I wanted to learn a song. And then I started going, well, this could be more fun. So I programmed it to, if it a, if you ask it, teach me to play Wonder Wall, it will say, maybe that's perfect.
It's a dad joke. What can I say? .
[00:17:41] Henrik Werdelin: You ever, right. Do you remember the movie Wayne's World?
[00:17:45] Russ Somers: Oh, yes. It actually has one about stairway too, if that's where you're going. It has
[00:17:48] Henrik Werdelin: that, where it says, No stairways to heaven as they enter into the, into a guitar shop.
[00:17:54] Russ Somers: Really? Brilliant.
I love that. And actually, if you ask the one that I built, and it's probably degraded by now, I haven't used it for a long time. Teach me to play stairway to heaven. It will say no stairway denied. Direct quote from Wayne's World.
[00:18:06] Jeremy Utley: That's so good. Okay. So, so starting on the journey, going back to Henry Henry's question, which is exactly where I wanted to go to first thing you'd say is, Find something that you care about, something that you love, something that you're interested in and build for yourself.
Is it, do you just, are they, are those the training wheels and now they're off and they're going to figure out what to do next? Or do they come back to you with certain kinds of questions or ideas? And if so, what's kind of the next step in call it a actual human's ability to grow in their GPT
[00:18:37] Russ Somers: journey?
It's a good question. And I'm still figuring that out. It's not like I've guided a lot of people through it yet. But my gut would be start with what you love and don't settle. If you remember people's exposure, first exposure to chat GPT, if you were talking to people a year or so ago, it's like, Oh, well, I used it to write, you know, an invite to my kid's birthday party, or I used it for this simple thing and there's this really, because it can do those things so simply and effortlessly, it's really easy to go.
Oh, that's what it does. I know how it works now. Right? Right. And so, so when you've built your song tutor and it like brings back basic stuff that you could get with five minutes of Google search yourself, don't settle. Say, how can I make this better?
[00:19:31] Jeremy Utley: Yeah, I just, I actually wrote a blog post a couple weeks ago that was, I think the most dangerous thing about Gen AI is not the existential risk, but it's that it will perform to one's low expectations.
And if you have low expectations, it'll do it. You go, that's all I can do. It's exactly what you said. The big danger is people, basically a self fulfilling prophecy of expecting it can only do so much. Rather that, that to me is the greatest danger because people leave going, Oh yeah, I get that. It's like, no, you don't.
If you get it, you're in, if you think you get it, you're in trouble. Um, but low expectations have a way of being easily fulfilled. And then people go, Oh yeah, and now I'm back on Twitter or whatever, not recognizing it's a fundamentally different kind of thing. So don't settle. How would you define not settling in this context?
If you're telling your friend, you know, Kathleen don't settle, how would she know if she's settling and how would she know if she's decidedly unsettled? If she
[00:20:22] Russ Somers: builds it, and then when she's planning a vacation in a few weeks, doesn't even think to use it, she just, she built a toy, right? And a toy is nice, we all like a toy, but you play with a toy, and you discard it.
What I would look for To know whether you're settling or not is continued utility usage and continued healthy dissatisfaction. The ones that I use now for work, I tweak a lot. I give them, I'm like You know, I recently fed Wendy back a bunch of the webinar content she'd created after our edits. And I'm like, hey, so that you know, this is what, this is what you generated.
And here's what we turned it into, and I want you to learn from that so you can get better. So there's that sort of learning where you feed things back to each other. Or even from one to another, I'll take Amon Adman's input and feed it into Roger Revop sometimes. But then there's also the continued tweaking of the instruction set.
When you recognize they're doing okay but not great, it's like, okay, go in and play with the instruction set to try to make them great. And I know that feels like a lot of work. I'd like to think, although I haven't measured specifically, I have about 3x'd my personal productivity here. I think, and I think if you're not doing that, if you don't look at this and go, I'm getting a boost, I'm getting an amplification, don't settle, go back in.
[00:21:51] Jeremy Utley: And just to kind of like put a full, like a full point on what you just said. You need to find something that you will, that you could conceivably use regularly and then push yourself to use it regularly and update it regularly in order to improve it. And if you aren't using it regularly, I love what you said, if you aren't using it regularly, you've made a toy.
So find something that if it worked, you'd want to use it a lot. And when it doesn't work, invest the effort to refine it. I think that's, that's a really important point. And so I just, I just kind of wanted to put a bow on that. So anyway,
so here's, here's a little bit of a radical question, Russ, just, just based on where you said you're at, you're about to grow the team. So set the context for a second.
Here's Russ. Who is 3x ing his personal output, leveraging GPTs, who's about to grow his human team. Here's my question for you, because I think a lot of leaders are grappling with, how do I bake this? Call it GPTifying into OKRs, into measurements, into, uh, you know, annual reviews, etc. I'd love to hear, even if you've never thought about it, just thinking about it live kind of would be great.
How do you think about, because it's one thing to say, like you bring on a junior marketing person, there's kind of conventional marketing measures for their success, right? How do you think about incorporating Gen AI and GPT oriented measures, metrics, et cetera, into a junior person's role, knowing the potential they have to unleash and multiply that person.
[00:23:25] Russ Somers: What a really great question. And literally my next task after this is writing a 30, 60, 90 day plan for this new hire. So, um, I'm going to, I'm going to free form on that just a little here, if I may. Please. No, I love it. Yeah. Let's
[00:23:42] Jeremy Utley: like riff because you don't have to have, I think it's just important to go.
We need to be measuring this stuff. I think the reason a lot of organizations aren't making progress is because it's okay to not make progress.
[00:23:54] Russ Somers: Well, and I heard, and this may actually have been on a previous podcast of yours, I heard somewhere about a company that gave its employees access and then basically I don't want to say reprimanded or disciplined, that sounds so awful, but they noticed who didn't use it.
They paid attention, and that's one reason that I, that I will move to Teams is so I can understand usage. Is everybody using it? Um, so I think, you know, there is some level of monitoring there, but it's also worth remembering, as much as I love these tools, they are only tools, So I'm going to an extent assess the person primarily on their output, on what they get done.
But what can I, what I can assume they can get done if we're assuming this 3x ing of human potential? I don't know how they're going to do it if they don't adopt. Fortunately, I know the person. They are desperate, really interested in adopting. It's somebody that I've worked with, um, in a past role for years.
And so I, I have no worry that they're going to dive right in and learn. But I think you're right. I need to figure out how to make it explicit in the onboarding, in the 30 60 90, in the OKRs, and in one on ones. Oh, you launched this campaign. Tell me how you did it. Oh, you wrote the landing page copy. Did you think about using a GPT for that?
Why or why not?
[00:25:25] Jeremy Utley: Yeah, well, so what you're saying there is, you don't have to measure, you don't have to necessarily measure usage if you're measuring output with the assumption, or if you've, if you've set a high enough output. Output bar that it's impossible to achieve without an Ironman suit. Then you go, if you're not hitting the bar, it's because you're not strapping on your Ironman suit.
I think there's, I think there's one kind of contribution maybe I'd make here, just insofar as you're working on your 30, 60, 90 plan. One thing I've been thinking about is not success, but attempts. And so, which is to say, it's not that I want you to make a better, you know, webinar campaign. I want you to make at least five iterations to the approach by which we make the webinar campaign, right?
Because I can't, I can't guarantee success. I know innovation is a low stakes game and it takes effort. What I want to be measuring and what I might recommend is measure the at bats and how many times Does this new junior person, uh, edit the GPT? That's actually what I care about, right? Because then what you bake in is, of course, they're using it and then they're thinking about why didn't it work and what could I do differently, right?
And even if it doesn't work or does work, you end up driving the kind of behavior, which I think you want, which is engagement with the tool, thoughtful engagement and refinement of the tool.
[00:26:41] Russ Somers: And this is why, Jeremy, that's priceless. And this is why a moment ago I was very uncomfortable saying the words, you know, discipline or whatever that I used.
Because people, I mean, any of us that have trained animals, you know, we were talking about dogs earlier, they learn from carrots, not sticks. And the idea of the carrot, I want you to experiment and I want you to take a lot of different tries and try things without, if you have this expectation that everyone needs to be a success, there's nothing that shuts down creative thinking more than that.
To think creatively, and you know this better than I do, obviously, because of your research area. but psychological safety is super important to people being able to be creative and come up with new use cases. So rewarding at bats, you're trying, you took a swing, you struck out, it was glorious. Let's celebrate that.
[00:27:35] Jeremy Utley: Well, I mean, to that point, like, what does it look like to not even celebrate, but require glorious failures, which is to say at the six month mark or at the 90 day mark, I want you to show me three spectacular frame flame outs that you never would have achieved if I hadn't. Give it not only given you permission, but told you I expect you to do something radically.
That's just going to that's going to crash and burn so spectacularly. We've got a debrief it right, but you can almost then it's not just about even iteration or or implied success, but you actually want this person to be pushing far enough that they make mistakes, right? So that because then what you're saying is we know we're not going to innovate unless we are exploring broadly.
So I'm going to reward the broad exploration
[00:28:18] Russ Somers: first. Yeah. I love this. Years ago, I was on a, I was a small player on a large marketing team early in my career. And because the company had gone through a lot of odd stuff, as companies do, the team was very risk averse. So the new VP that came in, and I thought this was brilliant, he launched a frog award.
under the concept that you need to kiss a lot of frogs to find a prince. So he gave the frog award each quarter to the most spectacular and audacious failure that somebody had tried to create a cultural DNA of, it's okay to fail. I actually want to see you try and fail a lot. So that's, that's something I can steal is the frog.
I'm stealing the frog and I'm merging it with your comment there. Cause that's brilliant.
It's very interesting to see how you're a persona, putting personified, personifying the bot.
[00:29:12] Henrik Werdelin: And, um, I have a follow up question, but the first one is the, how do you think about giving these bots a personality? And what do you think might be the importance of doing so?
[00:29:25] Russ Somers: I, I was literally in an argument last week with a friend that's a computer science professor. So should know more than me and actually does in all areas to be, to be completely humble.
He's like, you're doing this wrong. You're treating them like people. That's very dangerous. They're applications. You should build them as apps. And maybe, you know, you give, you know, let's say Piper the product marketer. Maybe you give, that shouldn't be called Piper the product marketer. It should be called product marketing bot and give it like an image that's not a person.
And I, the reason I think that is, Not as good an idea. And by the way, I do get better output from AI than he does. Um, It's because generative AI is kind of abstracting away the programming layer from a lot of what we have to do. It's designed for us to interact in a natural language way. And it's not that Calling her Piper makes her perform any better in an abstract, but it makes me prompt her differently.
Because I've thought about the fact, what is her background, and actually I need to retrain her, because right now she's too theoretical and framework oriented, I need her to be better on competitive analysis and stuff, so she's, uh, going to be going to school in a little bit, but I prompt her better when I think about who she is.
And I even, you know how it creates a default avatar. And let's be honest, the default avatar is usually a stiff person in a suit. And I work in startups. I don't work with stiff people in suits. So I've started getting very directive just to help me prompt better in, you know, Roger, Has shaggy hair and an earring and wears a HubSpot T-shirt, and you know, Amon, you know, the ad guy, he wears a black leather jacket in a t-shirt with ROAS return on ad spend across it.
So just, I feel more like I'm talking to a real person I might work with there, and therefore I prompt better, which yields better output. It's a, it's a way of tricking yourself, I believe.
[00:31:35] Henrik Werdelin: That's such an insightful answer. Uh, thank you so much for that. My, my follow up question is where it gets even more philosophical.
, in some of my work, we are now increasingly taking these GPTs and we are making them customer facing. So you are now having conversion flows and kind of re energizing the whole concept of conversational commerce the interesting thought that we're trying to figure out is.
As you are expressing your business through a chat interface. And so as a marketeer, you're being, you're being ripped the classic assets that you can normally do to evoke your brand ethos, you don't have colors, you don't have logo, you don't have cool pictures, you basically just only have a text prompt.
And so. How would you think about in a world that is increasingly conversational to take your brand ethos and then communicate it through a text chat interface?
[00:32:36] Russ Somers: Great, great question. Where does brand happen?
[00:32:38] Henrik Werdelin: I mean, like if you ask me, I think brand is what people say to other people when your name is mentioned.
[00:32:46] Russ Somers: Which I would, I love that plus one and enthusiastically endorse, which means that brand is happening really in the mind of your customer. Isn't it? It is. Yes, you have your brand book, but that's branding the tools you use to try to influence the outcome. That's all they are. Brand is happening in the customer's mind.
So I think it's a super exciting time for marketers. Because if we go, brand happens in the customer's mind and in the customer's conversations, per your point, with others, I used to influence that with a book of colors and fonts and, well, we're a very serious, authoritative brand. We're a fun, crazy brand.
It's like, okay, all of a sudden I don't have colors and fonts, but I haven't lost my charter of trying to influence how people think of us. So, what tools do I have? Well, I've got the chat. I've got words. I've got the way the chat is merchandised. How can I work within those? It's a, it's a really provocative question, and I think it's going to separate the marketers that work from first principles from the ones who simply copy and imitate, which means the real innovators in marketing are going to shine through.
[00:33:58] Jeremy Utley: It comes down to voice to me, it's, I mean, the way I would say it as a voice, which is distinctively it's word choice, right? And in a way, text is perhaps a truer articulation of brand than color or font or anything else. It's what's your voice, right?
[00:34:18] Henrik Werdelin: And one thing, for example, that we've discovered, which is interesting is voices, for example, also what you don't talk about.
Or would you refuse to talk about, or where you are very disagree, where you disagree quite a lot. And so, you know, obviously out of the box, most of these GPTs are pretty agreeable. And so if you have a brand voice that I have a perspective about. You know, like guitars, you're selling guitars, you know, then you probably want to be pretty explicit about, you don't want to talk about violins, right?
You know, you might even want to hate on, you know, like the flute just to kind of like invoke a specific emotion.
[00:34:53] Jeremy Utley: Hey, there's no stairway. There's no stairway. No stairway. Russ, I want to, one, one part of your story that you and I discussed in our previous conversation, I want to make sure that our audience gets here and that Henrik gets to hear it.
Talk to us for a moment about how you started this journey, because I think it's a non trivial and very important point. I'm referring specifically to an unfortunate event or to, uh, maybe something that you, you folks don't, uh, give enough emphasis to or value upon.
[00:35:29] Russ Somers: Well, you know, there are unfortunate events if we choose to think of ourselves as in a lemony snicket world.
I try to be a little more optimistic, although I did love those books when I kid read, kids read them. Um, about a year, yeah, just about a year ago, maybe a little more, I was laid off. And it was great company that I left. Um, you know, treated very well. It just, we weren't growing the way we needed to grow and we didn't need the team size that we had.
And it made sense for the senior director to stay and meet it apart. And they were gracious and great in handling that with me. And when I left, the first thing I did was what any responsible person did was I freaked out. I went, how will I make a living? I'm being tossed out into the worst market into the, in the world.
Yes, they were kind with severance, but, you know, they're not going to pay me for the rest of my life. Right? So it left me struggling. And I started, because I knew I needed to make money, I started consulting. I'd actually met, uh, the founder of Quantified, um, a few years ago and been impressed by what he was building because he was building this before Gen AI, Gen AI was on most people's radar.
And so I pinged him and said, Hey, I've been giving you a lot of free advice. I need you to start paying me. He said, okay. So I had my first client. I got a couple of other clients, but that gave me the flexibility in the runway. to play a lot. Because I had seen this Gen AI thing happening, I was fascinated.
I was slightly frustrated that we weren't doing more with it at the past company, although they've now made some good strides in that direction. But I really dove into it and played with it in a way I hadn't played since back when the internet was happening in the early 90s. I was a bartender with lots of time to screw around on my off time and so plugging stuff together and installing my first modem and seeing what was out there before there was even really a world wide web.
So that
[00:37:29] Jeremy Utley: was it. Time to play. It's, I mean, one of my intellectual heroes, I think Henrik says, well, Amos Tersky said famously, you waste years when you can't waste hours. And, uh, he was referring to how he and Danny Kahneman reinvented economic theory as we know it. Right. But his point was, Time to play and time to explore is critical.
And I feel to me when I heard that thing, and thank you for sharing that, I think it's so important. There is an imperative to make time. And right now in our hyper kind of the irony, by the way, of the kind of GPT movement is the whole value premise is about efficiency and efficiency is all about squeezing more into time, you know, and doing more and being more productive.
But what I've heard you say, what I hear a lot of people say is it's only when I, It's when I was let go that I finally had a chance to dive in and explore these tools that give me a chance to three x myself. I wonder, do you have any thoughts, you know, having gone through that experience, somebody who's not in the, you know, lemony Snicket's unfortunate event of, you know, finding themselves with more time on their hands.
How do they make? If they, if someone hasn't made time for them, so to speak, do you have any thoughts about
[00:38:42] Russ Somers: that? I want to come back to that because I want to make one other point on why I think this matters so much and why I routinely use the word play about it. Please, please, please. Why do we play?
Enjoyment. Survival. Think about it. Babies play. Children's play. Tiny kittens and puppies play. Baby rats play. I know this because we had baby rats on New Year's Day when my daughter was seven. We were kind of a weird family. But then think about a newly born creature in the wilderness. Why the heck would you play?
You potentially are calling attention to people. to yourself from predators, it seems very much counter evolutionary. And I'm not an evolutionary biologist, but I do have a raging case of Dunning Kruger syndrome, so I think I know why. I think play is the fastest way to explore a radically new environment and gain mastery.
Of the skills you need to succeed and survive there. So that's why I think we play. Otherwise, we just wouldn't.
[00:39:52] Jeremy Utley: Yeah, I think that's right. I think it's right. Bo Lotto is a neuroscientist that I've followed for a long time. He's got a great book called Deviate. And one of the things that he talks about is it's one of the only environments where uncertainty is actually sought.
We, we love and uncertainty that nothing new happens until we get in the uncertain arena. And yet the only time we actually embrace and seek out uncertainty is in play. It's kind of the one arena of human endeavor where uncertainty is a, is a benefit rather than a liability.
What does somebody do? How do they rationalize or create space to play when they're under the gun to deliver more, to be more efficient, to be more productive? How do you create the space? Granted, survival is, you know, hinges upon it. How do they do it practically?
[00:40:40] Russ Somers: It's really hard, right? Because if I say this is all about survival, it's like saying you need to laugh or I will shoot you.
You know, you can't laugh. In that environment, you really can't. So we were talking earlier about tricking yourself into doing better prompting by personifying the GPTs. I kind of think you have to trick yourself, and by the way, I didn't do it. I got fortunate and got laid off. It's not easy. But, I think you have to find ways to carve out the time.
You have to find ways to have that time and go, This is going to be completely unproductive, and that's totally cool. You know, You have to find that. That's why, by the way, encourage people start with something that you love, because then you actually will take evening and weekend time rather than saying, I know I should be studying that stuff, but I don't have time this weekend.
I'm busy. So I think that's it. Find ways to get into it with things you love. trick yourself into allocating the time, have some, and ideally, if you do that, you might find it starts to lift your work once you get into going, oh, I can build a GPT to help. The other thing though, and, and I know you guys know this, so I'm not going to belabor the point, is get GPT 4.
0, right? Get a real subscription. Don't just play with the free thing. The free thing is fine for learning, but I have a, another friend in CS that I was, that I talked into it over the weekend. And he was like, I had no idea it was this powerful.
[00:42:15] Jeremy Utley: We often say friends don't let friends use GPT 3. 5. Um, but Russ, unsolicited advice here, just going back to the question of the 30, 60, 90 day plan for your new employee.
I wonder how to, it's an interesting question to say, you know, we talked a lot about, um, ratcheting up the expectations to where only they can be achieved with, you know, superpowers and at this, this countervailing. Aspect, which you just brought out is you also have to create an environment where the person can be unproductive and not be self conscious of it.
And so I don't know exactly how to resolve that. But insofar as you said, you're getting off this call to think about 30, 60, 90. I think that's obviously, you know it, but I just wanted to call it out. It feels like that's the kind of that's the countervailing force here is yeah. How do you create the expectation and permission around play for somebody who feels like I got to get there and immediately prove my worth and demonstrate at, you know, ROAS, right?
I don't know. It's just an open question to me.
[00:43:15] Russ Somers: It's a good one. And it's one that I'm going to think a lot about. I will say it's not really different from the challenges we've always had in managing people. In the sense that I believe that people do their best job when they're fully engaged, when they're leaned in, when they're laughing, when they're willing to play, when they're willing to take risks.
We've all seen Google's research and all the stuff on psychological safety. So it falls to us as leaders, oddly, whether we're managing little GPTs or managing humans, to create the environment where they can flourish. And if they don't, it's on us.
[00:43:49] Jeremy Utley: Mic drop, mic drop. Russ, you're amazing. I think this is going to be a fan favorite. I really believe it.
[00:43:56] Russ Somers: Thanks
[00:43:56] Jeremy Utley: for us for joining us. It's great to talk to you.
How great was that guy?
[00:44:01] Henrik Werdelin: He's incredible. And you know, it is so fascinating how there are these golden nuggets of people. And I mean, I've had it to his point. Often it's not the big companies who have these people, right. It is the,
[00:44:18] Jeremy Utley: yeah,
[00:44:18] Henrik Werdelin: the one man band. that you need it. Uh, and so therefore they really start to use it. ,
[00:44:25] Jeremy Utley: I love his points about play and about discovery and about starting personal. I think it's so important. It can't be said enough. Work is not the place to start. The place to start is in your personal life and then having a commitment to iterate.
I love his kind of criteria of, are you being, are you settling? I thought that was a wonderful criteria and it's true, certainly for work applications, but in your personal life, my own kind of layman's perspective is folks are much more honest about whether they're settling because The vacation, they aren't going to settle for anything less than the ideal vacation, for example, right?
And they're going to demand that their GPT deliver it. And if they'll commit to a habit of using a GPT as. a way to inform refinements, they're going to start to go. It does what it can. Could it do that at work? Could it do that with my Jeep, you know, TPS report to quote office space. I think that is the big unlock for folks that most people are trying to start with work and just find they can't summon the activation energy required.
It's because they don't care enough and they should shift to an area that they do care about and watch the inspiration flow downstream into their work life.
[00:45:42] Henrik Werdelin: You really waste, you know, years by not wasting hours on this. And it is probably easier to, to waste hours if you're playing. The other thing that I thought was just an interesting way of thinking about it is this level one, two, three of Personalities that you could hire in your organization, like the AI hires that you're, you should be making.
And, and so the level one being basically stuff that you can't be bothered to do that you wish somebody else was doing on your behalf and level two, which is really about stuff you don't know. So. Maybe even like a specialist kind of like role that it will take you a long time to figure out. And the third one is this thing of like collaboration, like stuff where you feel you need to workshop it with somebody because you might not even know what the specific task is yet.
I think those, that's kind of like, I never heard. That kind of way of framing it before we just thought was fascinating.
[00:46:36] Jeremy Utley: I feel like we had someone who talked about when you're applying for a job See if you can get an interview a GPT to do the job you're applying for and basically your application is see how much more I can do because this GPT can do what you've, what you've kind of advertised as a position.
Do you remember that? It just kind of reminds me of that.
[00:46:55] Henrik Werdelin: Yeah. It was the incredible Ethan Malik, who basically recommended to his, uh, his MBA students that, As they're applying to jobs that you walk in with a gpt that was doing everything They were being asked to do and saying i'd like a pay rise on the first day
[00:47:10] Jeremy Utley: Of course it was yeah, that's so good.
That's so good But it just goes to show when you think about filling roles with these with these people so to speak with these personalities It really unleashes what's the human, therefore, what's the human capable of in this higher order reasoning and orchestration and chaining things together. I think Russ is a great example of he's kind of, he's a little glimmer on the horizon of the kind of unlock that I think Ethan's referring to.
And Ethan's kind of got the great, Theoretical, philosophical, empirical evidence for it. Rust to me is kind of a quintessential example of no, they're really people who are being unleashed like this by working in a different manner. So super fun to have the balance on the show.
[00:47:54] Henrik Werdelin: And I think with that, it's time to wrap up from another episode of Beyond the Bront.
If you like the show. Or anything of it. , a, we would love to know, so please connect with us on LinkedIn. And if, , you do like it, we would also really, really appreciate it. If you could share it with a friend, we'll go in and like, and subscribe on your favorite podcast platform.
And with that, Jeremy Otley, fingers off from this time. Henry Kordland.
[00:48:21] Jeremy Utley: It's been a pleasure as always.
Â