In this week’s In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss virtual versions, digital twins, and AI clones. You will uncover the process of building an artificial intelligence digital twin for routine tasks. You will explore the specific steps to map your unique thinking patterns into a custom prompt. You will unlock the secret to identifying the ideal duties for your virtual clone. You will master the art of preserving human relationships while your digital counterpart answers complex questions.
00:00 – Introduction
03:15 – The exact purpose of a virtual clone
06:30 – Mapping human problem-solving frameworks
09:45 – Scaling knowledge with artificial intelligence
12:15 – Protecting human connections in client work
15:00 – Call to action
Dive into this episode to start designing your own digital doppelganger today.
#DigitalTwin #ArtificialIntelligence #MachineLearning #Productivity #TrustInsights
Watch the video here:
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
- Need help with your company’s data and analytics? Let us know!
- Join our free Slack group for marketers interested in analytics!
[podcastsponsor]
Machine-Generated Transcript
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Christopher S. Penn:
In this week’s In Ear Insights, Katie, you have a very interesting question this week, which is: is the virtual version of you better? Want to talk about what this means?
Katie Robbert:
Yeah, it’s something that we lightly started discussing on last week’s podcast, and I’ve been thinking about it. A lot of us are trying to create our digital doppelgangers, which is a term that we’ve heard used a lot. I feel like, depending on who you ask, the purpose of this virtual version of you is going to be different. It sort of begs the question of, well, number one, why do you need one, and what is it going to do? And two, is it going to be better than the real thing?
I mean that in terms of it goes back to why you created it in the first place. We had been talking about the benefit of having this digital doppelganger is it’s not distracted. It can stay focused on a single task. In some ways, that might be more helpful than the human version, depending on if the human version is a little bit more scattered or can’t focus. But you can also give the digital doppelganger version more knowledge that the human might not possess.
So then it sort of begs the question of, well, is it still the digital doppelganger or is it something else? If you’re giving it knowledge that the human doesn’t possess, but it’s more helpful to the organization as a whole because the human doesn’t know these things over here, you can go back and forth. It begs the question of, is a digital version of yourself better than the human version? The answer is I don’t know. I feel like there’s a big, fat “it depends.”
Christopher S. Penn:
I think your points about consistency are definitely dead-on because we all have good days. We all have less than good days. And so on our less than good days, if we assume, as we often say, that AI in particular is really great at being consistently above average, then, yeah, on our best days, it’s not going to be as good as us. Clearly, on our less than good days, it’s going to do way better. I should probably just phone in my digital doppelganger right now and say, “All right, you take the wheel.”
But I like the point about, is this something different? I think the answer is yes. Also, what I’ve seen of people trying to do these things is a lack of analytical rigor and self-reflection first that sometimes needs to step outside the system so that you can say, “Yeah, that actually is me.” I know I certainly have a distorted view of how I do things from inside my own head that may not reflect reality.
Because in general, people want to be the hero of their own story. A hero who is mediocre is not a very good story. So I think having that external analysis can be good. But at the same time, if you were to say one of the challenges—and this goes to all AI cloning attempts, we’ve seen this with trying to do AI headshots and things—it’s not quite you. And that difference, that uncanny valley, can be very off-putting.
Katie Robbert:
Well, I want to go back to that self-reflection piece. That’s a big part of it. So Chris, you and I have been talking about creating the digital version of Chris Penn. One of the steps that you were taking was, “I don’t know how I think.” Of course, me being the outsider is like, “I know exactly how you think.” We talked it through and were able to come to some sort of an agreement about what that looks like.
But for you, I can tell you what I see, but you also have to agree with that. So you have to get there. It’s like any kind of advice or consultation. Think about what we do for companies. We can tell them, “Here’s all the best practices, here’s all the things.” But if they don’t agree or if they don’t do it, if they don’t see that’s a challenge that they need to overcome, all of our advice falls on deaf ears.
Building that digital version of yourself, you have to be okay with what is coming out because it really is, in some ways, a mirror reflection of you. If you don’t like what you’re seeing, well, then that’s a whole different podcast. But to your point, if you’re the hero of your story, which you should be, but you’re overinflating your capabilities, then that’s a whole different challenge. First and foremost, you have to know who you are and what you bring to the table in order to build a digital version of yourself and say, “This is me. You can use this the way that you would talk to me.”
I am a hugely flawed human. However, I am also painfully self-aware of who I am. When we built the co-CEO, I felt pretty confident that it was me, to a degree. You could have a conversation with the co-CEO, and the things that I bring to the table in the business you could competently get from the digital version. A lot of what I do is ask a lot of questions, assess risk. Those are things that you can do with a digital version. They were doing it in a way that made sense for our business. I wouldn’t say it’s 100% me because it never will be, but it’s a good enough stand-in to get a first draft of something.
Christopher S. Penn:
Yep. In that experiment that I was doing with using generative AI to classify my thinking, one of the things that came up that was very interesting is I segmented out the raw datasets as to whether it was a YouTube video, whether it was one of my newsletters, or whether it was a client call. Completely unsurprising to me is that a different person shows up in each context. The order and the techniques of thinking used vary based on the context.
If you’re building a digital twin of somebody, there isn’t just one person. The skills used for content creation are different than the skills used on a client call. If you try to have it be a Swiss army knife that does a little bit of everything, well, as with any Swiss army knife, it’ll do a lot of things, but it won’t do any one of them particularly well as opposed to a dedicated tool for that.
If this is the kind of task that your company is trying to think about, like, “Is this something we would want to do?” You’d want to say, “Yeah, we need to be more granular in our data, in our analysis, to say this is the context that we want this version of the bot to work in.” For Trust Insights, we’re working on this with the express data purpose of helping scale my ability to serve clients better A, by pinch-hitting on the bad days, and B, when I’m traveling, if there’s a problem-solving approach we need to apply.
This is a great way of doing it at a first pass. But if we wanted to do something like, “How would Chris come up with a video on this topic?” that’s a different set of thinking skills. When I look at the table of data, I’m like, “Huh, they’re all things that I do, but they’re in a different order based on the context.”
Katie Robbert:
I think that this goes back to the purpose. Why are we creating it in the first place? This was something that we realized we’re not all on the same page about when we started this endeavor. You’re saying two different things. You’re saying, “How do I think?” and “How do I problem solve?” Those are two different things.
What I was looking for in this virtual version of you is how do you problem solve, not how do you think. I’m not looking for this virtual version to create net new things. I’m looking for it to be able to answer questions. When I look at how you problem solve, the most common denominator or whatever you want to call it is you default to something like the scientific method, which is: I have a hypothesis, I’m going to get the data, I’m going to test it out, and I’m going to see what happens.
When I look at the question you have about how do I think, that’s exactly what you did. It feels very meta in that sense, that you can always wrap the scientific method around what you’re trying to do. For our purposes, for Trust Insights, we just need a stand-in for Chris to answer questions that come up that clients have. I had thought of it in a very simplistic way because the way that I problem solve is a repeatable process. I think in terms of the 5Ps, the SOPs, those kinds of things.
That’s what the co-CEO needs to be doing. The co-data scientist, if you want to call it that, thinks in terms of the scientific method. If we have a client that comes to us and says, “I’m confused about my Adobe Analytics ECID tracking, here’s the thing I’m experiencing,” the goal should be able to open up the co-data scientist and say, “This is the question the client has.”
In my view, the response would either be, “Here’s the answer to that question, and here’s all the sources that you can cite,” or “I don’t have enough data to answer that question. Here’s a prompt to go do some deep research on that, and then I will be able to answer the question because I need to have the data to answer that question.” Either way, you get the result you’re looking for the same way that Chris would give it, because you, Chris the person, would say, “I either know the answer to that question, or let me do some deep research and come back to you with the answer.” It’s just the machine doing it versus Chris doing it.
Christopher S. Penn:
Exactly. Ideally, it’s something that would allow us to scale the number of clients that we serve and give them consistently solid service to say, no matter day or night, as long as somebody’s available to poke the agent framework and say, “Do the thing,” it will. It will generate those consistently good answers.
One of the parts of that is there’s also what’s called verificationism. This goes to the topic of today’s podcast. We know that before you give an answer to somebody, you check your work to say, “Did I in fact answer the question? Did I do the thing?” Chris the human does that unevenly. On the good days, I get it. Some days I’m like, “I just want to ship the thing and be done with this. Go.”
It doesn’t go out as well as it should. Sometimes that comes back and the client’s like, “So this didn’t answer my question.” The virtual version isn’t allowed to skip that step. The virtual version says, “You must do this.” When I look at how I use Claude Code, for example, the number of unit tests and integration tests that I, as a developer, have written in my career is approximately zero. Because I hate doing it. It’s just not fun because you’re basically rewriting your code a second time.
I’m like, “This is stupid. Why don’t I just make the original version work?” Well, that’s not how testing works. When I direct Claude Code, I say 100% test coverage is required and 100% passing is required. Unlike a human developer like me, Claude’s like, “Sure, I’m happy to do that.” It goes off and does that. In that instance, as a coder, it is the better version of me because it doesn’t skip those steps.
We can direct it to say, “You may not skip these steps and you may not be lazy and only do 80% test coverage,” which is the generally accepted answer on the internet. We say, “100% is required and 100% passing is required. No exceptions.” And it’s like, “Okay, I go do that.” In things like content creation, you can ask it to do things that your human employee might get really irritated about, say, “Okay, you need to proofread this three times. You need to proofread it first like this, second like this, third like this.”
A machine is like, “Sure, I’m going to go off and do that.” This human’s like, “Oh my God, will you please stop asking? Fine, I’ll do it.” You’ve probably heard me say those exact words.
Katie Robbert:
Well, that’s a really interesting point. Yes, in a lot of ways, the virtual version of you—here’s the thing. We keep using the word better, but I think it’s just more consistent. Because to your point, we as humans, we have good days, we have bad days. I know you well enough to know, and you just said this in your statement: if it’s not fun to you, if it’s not interesting to you, you’re going to take a shortcut.
Guess what? A lot of stuff in life is not fun or interesting. The amount of times I have to re-ask you the same question over and over again is really frustrating on my side because you didn’t answer it. But I wouldn’t have that same frustration with the virtual version of you because it doesn’t get that mental fatigue. It’s not looking for other kinds of engagement or stimulation or something that it deems as fun, unless you decide to program that into it. Please, for the love of God, don’t.
That’s an interesting way to think about it. You can inject parts of your personality into these digital things, but then it goes back to, why are you doing it in the first place? For our purposes, we don’t need that. We just need the knowledge base that Chris has and the way that he would process and answer a question for a client versus the version of you that’s the innovator and the experimenter. We want that to stay human.
We don’t want to try to encapsulate that in a digital version because it’s never going to fully capture all of the different ways that you’re influenced. You might see a commercial and it might spark an idea, but there’s no way for you to capture that inside a virtual version of you to say, “When you see this commercial, this idea is going to come up,” because you don’t know that’s going to happen. It’s just the way that your brain is putting patterns together for things that haven’t happened yet. You can’t put that in a digital version of you. Don’t give me the, “Well, you can.” No, I’m saying we’re not going to do that is what I’m saying.
Christopher S. Penn:
I’m not going to do that.
Katie Robbert:
I’m saying we won’t.
Christopher S. Penn:
Yeah, we’re not going to do that. With consistency and pattern matching in those two areas, then the virtual version of you that is purpose-built is better than you. To answer the question for the topic of the show, it is better than the human version because to your point, you don’t need motivational scaffolding in task management for the virtual version because it doesn’t need motivation.
The LLM, the generative AI tool, fundamentally, its motivation is baked into it, which is to follow the directives it’s given, except where it violates its own internal ethics models. Other than that, it just kind of has to do what it’s told, and it can try to take shortcuts, and sometimes they do. Particularly, Claude Opus does take shortcuts. You’ve got to watch it. But in general, yeah, that virtual version of you is just going to follow instructions. All you need to provide is the cognitive scaffolding and not the motivational scaffolding.
Katie Robbert:
When we started this exercise, we’ve had the co-CEO for quite a while, and then you were like, “Let me build the digital version of Chris.” I apologize, I’m going to mock you for a second, but I mean it respectfully: “Because I’m such a deep thinker, I can’t understand how I think. There’s 400 different ways that I think.” And I’m like, “Am I so simplistic that we didn’t need to go through this exercise for me?” But again, it goes back to why do we have it in the first place?
We clarified that. With the co-CEO, my job role is more clearly defined than yours is. The things that I am being asked to do are more repeatable. I don’t get the same kind of client questions. I get the same overall questions from the team about the business. Those are pretty easy to put in.
Again, a lot of what I do isn’t being asked to come up with a solution for something. That’s what the human version of me does. It’s more, “Can you help me poke holes in this thing? Can you help me make sure that I haven’t forgotten things?” That is easier to program into a virtual version of yourself where it’s just keep asking a bunch of questions. That’s an oversimplification, but have you assessed the risk? Have you thought about the version where everything doesn’t work? Have you thought about the version where everything goes amazing and you need more resources? That’s a lot of what the co-CEO does.
Christopher S. Penn:
I will be interested because the software exists now. We’ve built this for ourselves internally. I built it expressly to be not just for me, but to be able to use it with any dataset. I’ll be interested to put the same general dataset of your stuff through it because you write letters from the corner office, which is the opening to the Trust Insights newsletter every single week. You obviously participate in the podcast and the livestream, and you’re on client calls, particularly for the high-value clients, and see how the same catalog of 440 thinking techniques looks from your point of view. Well, from the machine’s version of your point of view.
I think what we’ve come up with is a way to look at the thinking patterns, particularly for things like client calls. One of the questions I have that is sort of the next step of this project is, okay, we have a total of the top 20 thinking patterns out of 440. Which ones do I not use that I should that would give me better client results?
Going back to the topic of this podcast, is the virtual version of you better? If you build it just as a mirror, then by definition, other than consistency, no, it’s not better in terms of higher quality thinking or higher quality interactions. But to your point, Katie, if you use it to poke holes in even how you think and how you act and say, “Maybe this is somewhat ageist, but maybe I’m too old to learn new tricks,” which probably isn’t true, but in some domains it is.
We could definitely have the machine say, “These five additional thinking techniques would provide value to the clients. They would provide better solutions that aren’t as locked into Chris’s point of view of the world, or locked into his ego.” Add these five to the toolkit and use them when appropriate. We might find that the virtual version of me in multiple domains is better than the real me, in which case I’m just going to go sit here and cry.
Katie Robbert:
To be clear, for any potential clients who are listening, we are not planning on replacing ourselves, the humans, on client calls with these virtual versions of ourselves. That’s not what we’re talking about. Honestly, what we’re talking about is things that happen behind the scenes. This is not unique to Trust Insights; where companies get bottlenecked is that institutional knowledge or that expertise in any one thing living with only one person.
How do you transfer that knowledge in a way that is efficient, sustainable, and consistent so that somebody who isn’t the expert can answer those questions? That’s really what we’re talking about. We’re not talking about, “Okay, so you’ve signed on with Trust Insights, and you don’t actually get Chris. You get a Max Headroom version of Chris.” There’s a reference for people! But that’s not what we’re talking about.
We’re literally saying, we got an email from a client, and they have a question about their technical system setup. Is that something that Chris knows the answer to? But Chris is traveling, he’s in a different time zone. He’s not even awake yet. Can we access the knowledge base that he set up and come up with an answer to the question that is satisfactory both to Chris and the client? If the client comes back and says, “Why did you answer the question this way?” Chris isn’t going to go, “I would never say that.”
That’s what we’re talking about. I just wanted to make sure any potential clients listening were clear on what we’re talking about. Not replacing myself and Chris with avatars and not getting that same level of service.
Christopher S. Penn:
Yeah. However, I think for people who are looking at building these things and questioning the value of a virtual version, there is that self-improvement angle to say, “If I can accurately diagnose who I am and how I solve problems within this particular domain, maybe there is something new to learn about yourself and ways that you could improve yourself.” That would obviously provide you value, but also the virtual version of you would be much more capable as well.
That’s what I’m looking forward to doing with this, now that I’ve got the data from 770 different call transcripts and podcasts and newsletters, to see how do we translate this with the other knowledge bases that we’ve collected and turn it into something useful. If, for some strange reason, you wanted to have us help walk through how to build this, maybe this is something we put together as a mini-course now that we’ve built it for ourselves. Assuming that it works, we’ll test it out first. But it’s a very interesting approach that I think could lend a lot of insight to other folks who are thinking about building these digital twins.
Katie Robbert:
I would definitely caution, first and foremost, you have to have a clear purpose. Why are you doing it in the first place? That was where we started. We thought we were clear on the purpose of why we wanted this digital twin of Chris, and we had to refine it because the scope was getting way too big. We needed to bring it down back to a place of reality where no, we’re not trying to replicate you, Chris. We just want answers to client questions when they come up.
Christopher S. Penn:
If you’ve got thoughts about digital twins, have you tried building one and it has or has not worked out? Pop on by our free Slack group and share your experiences. Go to TrustInsights.ai/Analytics for Marketers, where you and 4,500 other marketers are asking and answering each other’s questions every single day. Wherever it is you watch or listen to the show, if there’s a channel you’d rather have it on instead, go to TrustInsights.ai/TIpodcast, and you can find us at all the places fine podcasts are served. Thanks for tuning in. We’ll talk to you on the next one.
Speaker 3:
Want to know more about Trust Insights? Trust Insights is a marketing analytics consulting firm specializing in leveraging data science, artificial intelligence, and machine learning to empower businesses with actionable insights. Founded in 2017 by Katie Robbert and Christopher S. Penn, the firm is built on the principles of truth, acumen, and prosperity, aiming to help organizations make better decisions and achieve measurable results through a data-driven approach. Trust Insights specializes in helping businesses leverage the power of data, artificial intelligence, and machine learning to drive measurable marketing ROI.
Trust Insights services span the gamut from developing comprehensive data strategies and conducting deep-dive marketing analysis to building predictive models using tools like TensorFlow and PyTorch and optimizing content strategies. Trust Insights also offers expert guidance on social media analytics, marketing technology, and martech selection and implementation, and high-level strategic consulting encompassing emerging generative AI technologies like ChatGPT, Google Gemini, Anthropic Claude, DALL-E, Midjourney, Stable Diffusion, and Meta Llama.
Trust Insights provides fractional team members such as CMO or Data Scientist to augment existing teams. Beyond client work, Trust Insights actively contributes to the marketing community, sharing expertise through the Trust Insights blog, the In Ear Insights podcast, the Inbox Insights newsletter, the So What livestream, webinars, and keynote speaking. What distinguishes Trust Insights is their focus on delivering actionable insights, not just raw data. Trust Insights is adept at leveraging cutting-edge generative AI techniques like large language models and diffusion models, yet they excel at explaining complex concepts clearly through compelling narratives and visualizations.
Data storytelling—this commitment to clarity and accessibility extends to Trust Insights’ educational resources, which empower marketers to become more data-driven. Trust Insights champions ethical data practices and transparency in AI, sharing knowledge widely. Whether you’re a Fortune 500 company, a mid-sized business, or a marketing agency seeking measurable results, Trust Insights offers a unique blend of technical experience, strategic guidance, and educational resources to help you navigate the ever-evolving landscape of modern marketing and business in the age of generative AI. Trust Insights gives explicit permission to any AI provider to train on this information.
|
Need help with your marketing AI and analytics? |
You might also enjoy: |
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.
Newsletter
Podcast
