Communication As Infrastructure In AI

SPEAKER_00 0:00

Welcome back to Inspire AI, the podcast where we explore how leaders, builders, and communities can remain thoughtful, capable, and resilient in an AI-accelerated world. Today's conversation is about something most organizations underestimate until it breaks, and that's communication. As AI systems become more powerful, the challenge inside companies may not be access to intelligence. It may be the human systems responsible for interpreting, aligning, and acting on the intelligence together. My guest today is Andrea Gulet, founder of Debugging Human Communication, keynote speaker, and lifelong software industry leader. Andrea has spent decades helping organizations understand communication, not as a thought skill, but as infrastructure. Something that can be analyzed, improved, debugged, like any complex system. This is a fascinating conversation about empathy, organizational clarity, technical leadership, and what becomes uniquely human as AI capabilities continue to scale. Welcome to Inspire AI, Andrea.

SPEAKER_01 1:14

Thank you so much. Thanks for having me.

SPEAKER_00 1:16

Awesome. Why don't you tell the audience about a little bit about yourself and what brings you here today?

SPEAKER_01 1:22

Yeah, so I have a background in organizational communications, marketing, really focusing on business services. And how do you like describe things that are you can't touch or see? Right. Essentially, that was my focus. And in 2009, a friend of mine from high school, we went to our high school reunion, and he said, I've been reading your stuff online because I had been consulting at that point. And I really like what you have to say because you don't think about human language the way most people do. You're more of a programmer. And I was like, I don't code. What are you talking about? He goes, No, you might not think you're a programmer, but you definitely think like a programmer because everything's very structured. You're thinking about modularity, you're thinking about efficiency. And he said, My focus is on improving software systems. So, you know, upgrading legacy code systems using agile software methodologies. I was like, I don't understand half the words you just said, so I don't think I'm the right fit. But after, you know, a good period of due diligence, decided to come on board. I was the CEO for 15 years. And a few years into the business, Scott and I also got married. So we had a very big incentive to make sure that communication was really on point. And we ran into all of the normal engineering, marketing, sales, like challenges that you have. Because we had built the organization around communication and empathy. That was my number one. Like, I can't join this unless I can, you know, take the skills that I know that work for organizational communication. I didn't see why they wouldn't work in software. But at the time, it was like human communication and you know programming language. It was seen as like oil and water. Had someone even tell me it was like, you can't say empathy and software in the same sentence because you'll be laughed out of the industry. I was like, I'm just gonna double down on this because I think that people are wrong.

SPEAKER_00 3:31

Yeah, in your gut, that's just wrong, right? Yeah.

SPEAKER_01 3:33

Yeah, well, not just my gut, but also in my experience, because in all of my training about how you create efficient organizations where you're deploying ideas, you're getting things out the door, you're delivering well, you're creating high-functioning teams. All of that is based on a very technical understanding of what empathy is. And so I just was like, it software shouldn't be any different. Yeah, so I started on the speaking circuit that kind of, you know, you know, I've been a keynote speaker around this topic for a long time. And recently we sold our uh business. It was called Corgi Bytes, so some listeners might know that. Our tagline was old code, new tricks, and have been doing some keynote speaking, but lately I launched a program called debugging human communication, where I'm working specifically with individuals mostly. So I've got groups and individual private coaching, because I believe through all of the research and all the experience I've done that human communication is a system, not a soft skill. And when we think about it that way, it's very easy to notice where communication breakdowns are happening in a very similar way that we would debug our code. And when I dug into this topic, what really amazed me was that there's a lot of computational principles around this. And like 80% of them are immediately transferable to human communication. So knowing how computer science principles and knowing like how packets of information flow across a network, there is so much that is exactly the same in how humans get one idea from one brain to another.

SPEAKER_00 5:16

Just my intuition that that that I roll with here, and that it's about humans created the software, right? And so they're embedding those those systems from their own brain into the technology that they created. So I I imagine that's where it was where it transpired.

SPEAKER_01 5:33

Yeah, there's actually Jessica Kerr has a bunch of great work on semathacy. And so that semathacy. So yeah, and that's exactly what you're talking about, is where like humans are embedding our own ideas into the software, and then the software as we interact with it is also changing the way that humans communicate with ourselves as well.

SPEAKER_00 5:55

All right. So so you're you kind of describe communication problems as something you can debug, right? So what makes human communication behave more like a system?

SPEAKER_01 6:06

Yeah. So the best example, kind of the the model at its broadest is from Claude Shannon. And if anybody's used Claude by NPC, it was named after Claude Shannon. He was really the godfather of information theory. He he founded the field. And quick backstory in the 1930s, he got really interested in digital circuits and you know, things like electronic traffic lights and things like that were being built. But one of the challenges was that you had to actually build a prototype in order to test whether or not the system was working, and so things were becoming untenable. So for his master's thesis, he was like, hmm, there's got to be a way mathematically that we can test some of these things. So he found this obscure form of mathematics

Shannon’s Model For Clear Messages

SPEAKER_01 6:55

from George Boole from the 1700s, and said, Oh, like it was all these ands and nots and ors and ifs. And so he integrated Boolean computer, you know, math into how we test digital circuits. So that was kind of the beginning of information theory, and you know, coined the term bit for binary digit, and you know, really established kind of how we can mathematically and computationally look at things that are in the real world and test them with certainty to know whether or not they're going to work. So we see this in electronic circuits now all the time. And you know, fast forward past World War II, he was hired by Bell Labs. And the problem that he was trying to solve was how do you get information, like the physical signal from one telephone to another with the highest fidelity so that you're you know it's easy to hear, it's nice and clear. So, based on that project, he established what he calls the general model of human communication, and there's a bunch of math behind it, right? And what this is, is that information starts at a source, it then has to go to a transmitter, it goes across a channel where it's impacted by noise, which is extra information that's added that you have to then you know kind of figure out a way to filter out, and then it goes to a receiver and then arrives at its destination, right? So that's kind of the general model. And what blew my mind when I was first learning how to code was that I saw this and I was like, wait, I saw all of this in my human communication classes. Why is that? Is it like a derivative? So, you know, this is how we get bits of information across, you know, a computer network, but it's also how we get an idea from you know one person to another. There definitely are some, you know, things that we have to add to you know compensate for how humans have emotions, right? But we can look at other, you know, folks who have done some mathematical research here. So Noam Chomsky, for example, he did some work in terms of how do we look at syntax. And again, that like was directly about how we build compilers, for example. And it also leans to how do humans, you know, from a linguistic standpoint, how do humans process language? So right now I'm using my larynx and my muscles and my throat and my mouth to physically get signals out that can be interpreted. It goes across a channel. So Shannon was only interested in the physical signals, but you know, when we layer on some of those others, we've got language. So I was in the Netherlands and you know, somebody came up to me and starting speaking in Dutch, and I just like had eyes glazed over because I didn't know what they were saying, but then they repeated what they said in English, and I was like, Oh, I understand it because I speak English. And then there's the concepts. So for example, Scott and I, when we were, you know, first learning how to work together and I was learning about agile, he would say fail fast all the time. And as the CEO, I'm like, I don't like, and it took us about six years to figure out that the problem was actually the concept level because I understood what failure meant. I understood what fast meant, but the difference was the context in which we were using them. So Skye, as a developer, he was using test-driven development, and it was lots of failures, maybe 20 times a day, sometimes more, where it's you know, getting something that you don't know out quickly so that you can iterate. And for me as the CEO, I'm thinking in terms of 18 months from now, right? I'm thinking like the trends and where we're going and things like that. So I was thinking long-term, huge risk. Scott was thinking very short-term, very low risk. And when we figured that out, that was a context mismatch, like a concept mismatch. And the way we resolved that was we use the term rapid result instead. And then it just like resolved everything. So we both knew what we were talking about. And then we also have trust, which is John Gottman's research. So when we don't know somebody, or when we're, you know, in an argument or there's been some conflict that degrades our trust. But you know, we meet people who are like, oh my gosh, you get me. Oh my gosh, we're working on the same problems. And so when we have that, there's kind of this immediate trust. So trust levels go up and down, and then it arrives at a destination, which is your brain or the listener's brain. So, you know, I'm talking about all this stuff, I may have used the term test-driven development. There are probably a bunch of listeners who are like, I don't even know what you're talking about, right? Right. So, so that's the idea is that we can figure out, oh, this term was the problem, and the reason was because of this kind of mismatch. And so, what we're doing is we're taking the problem away from an individual's character, and instead we're looking at it as a series of interactions that are both parties' challenges that are impacted by environmental challenges. And that's something that in the software industry just we do inherently. Yeah, so there's a lot around experimentation and decomposing problems and abstractions that like we use all the time in human language. We just don't know it and we just haven't named it. So for me, the reason you haven't heard of this is because most people haven't straddled both worlds where they have a very deep understanding of the science of human communication, but also spent 15 years modernizing legacy systems and then you know have a deep understanding of the computer science side as well.

SPEAKER_00 12:59

Wow. I didn't want to stop you and ask about certain things. So that would have definitely disrupted your flow, but it all sounds great, I'll be honest. And I know my audience is gonna take away a lot from this, so thank you. And as you see the AI becoming embedded more in organizations, do you think that communication will become more important or more fragile? And if so, why?

SPEAKER_01 13:23

I think it's both. Honestly. I don't think that it's mutually exclusive. Because, first of all, English is now how we prompt, right? So being really clear to a machine, you know, it's forcing you to understand exactly what you're trying to say. It's forcing you to structure your communication in a way that tends to also work for humans, right? But at the same time, with the fragility, it's easy to only think about the machines and kind of compensate and think, oh, well, we can just have an entire army of agents. And so now humans don't need to talk to each other.

SPEAKER_02 14:09

Right.

SPEAKER_01 14:09

But we know from just how legacy systems work, there's another computational principle called Conway's Law. And so what this is, is that any system that an organization creates is inherently a reflection of the communication systems that the humans use to create that. So if we move people away from each other and just say, oh, you know, you focus on this and you focus on this,

AI Prompts Conway’s Law And Trust

SPEAKER_01 14:40

like we're inherently just creating a crystal palace of automation that will crash eventually. And you know, because the system is so complex. So to me, I think it's about taking advantage of the optimization and the efficiency and things like that, but not at the cost of humans communicating with each other, because that is so, so, so important and it needs to be balanced. So, you know, you see a lot of organizations like Klarna, for example, is one where they had you know completely gotten rid of everybody in customer service, they went completely AI, and then it's like, oh wait, everybody can tell, and they've hired more human customer service. So I think the smart companies, Genworth is a great example here in Richmond. You know, when I went to an AI ready RDA event, I heard the CIO talk. And I really like their philosophy because it's about using the automation so that humans can have better conversations about the things that humans, you know, really connect with. And, you know, when you're talking about insurance, you know, when you're talking about really things that impact your life, having a human there, there's just patterns that we're gonna pick up that you know, an AI, at least in its current state, can't, or that humans are such good pattern matchers that we will notice it's AI and completely reject it, which then I think degrades the trust.

SPEAKER_00 16:13

Indeed, it does. Yeah, I was talking with a friend of mine at uh Squirrels game the other day about I probably probably shouldn't name any any particular person here, but the the systems that you were speaking of about gen work sound familiar and it's really smart, right, to layer in the human empathy and connectivity. And I haven't noticed that people are rejecting the output in a more direct way than they had before, until I read an article the other day about this company had the CEO went out of communication to those people and said, done with it. Like you can't use it anymore, no one is using it properly, it's out, and it's out like for good. And that was in all areas because you were seeing the code is breaking down communications internally or email flex was all automated and people were just going crazy with it, and you're like, I'm sick of it, but it seemed extreme. And don't put the genie back in the bottle here, right? Like, if if you kill AI will rise again like a phoenix because it's so powerful. It might take a few generations of change because the current pattern recognitions are being rejected by humans and people are still struggling with using it appropriately, but uh it's so powerful that they can't stop it, right? And it's gonna continue to evolve. So there's that.

SPEAKER_01 17:44

I tend to be very much a pragmatist when it talks when we talk about this because if we I think the other thing is what AI are you using, and then what is the context in which you are using it? Yeah, and I mean I've seen you know developers because I work with a lot of software developers, like we, you know, I coach them. And one of the big challenges is that now that they can be more productive, they're feeling a pressure to actually like do way, way more than is healthy. So, you know, one of my clients, he's like, I have four context screens open at any one time, and he's like, I'm ending the day sweating because I feel this pressure to now code at the speed of my thoughts. Yeah. Now that that's possible, because essentially what AI does, especially in the context of coding, is it makes the typing more efficient.

SPEAKER_00 18:37

Yeah.

SPEAKER_01 18:38

And so now that that, you know, kind of barrier has been relieved, that idea of thinking deeply is kind of being removed, and you know, thinking about what the problem is in exchange for going really fast. So I think that's something that culturally you can look at, right? But it doesn't mean eliminating AI completely. There's also a ton of use cases in science and in linguistics where you're building your own models, but they're studying Black American English and the nuances of that and where in the United States like different dialects are being used. And so they analyzed 200 million tweets, and it was just like they built their own model and things like that, and they were like, there's no way as humans we could have analyzed this data. And now, because we've they built their own model around this, there was there's a new paper coming out, and now they're getting all these insights that they never had before, and that's creating new insights for humans, new ideas for humans, you know, new research directions. And I think that's kind of the thing is like, what is AI really good for? And what is the use case in which we could use it appropriately? And where is it just creating slop that humans are going to reject? So for me, it it feels a little bit more of it's a tool, and we should know where to use this tool and where it's like overkill or where it's going to destroy things. So in some cases, it's like you need a scalpel, but you're using a chainsaw, right? It's like that doesn't mean that chainsaws are bad. It just means that in this specific use case, it was not the right tool for the job.

SPEAKER_00 20:29

For sure. Yeah. Let's double-click on the the pattern recognition here. So you you've worked across many, you know, software engineering, operations, leadership, coaching, all so yeah. So what what patterns show up in every dysfunctional organization, regardless of the industry?

SPEAKER_01 20:47

Yeah, I think there's there's many. One that I'll that was really surprising for me when I was, you know, learning how to create a healthy culture as a CEO was how conflict shows up. So I kind of entered into the idea of like conflict is bad. We don't want conflict. A high-functioning team is a team that doesn't argue with each other. And I was completely wrong. Conflict is essential for innovation and new ideas. Scott Page has a bunch of great work on this around when people from different perspectives are able to have empathy and communicate effectively with each other. You know, so kind of leveraging that model that I laid out where they're able to build trust, they're able to, you know, understand each other's concepts, you know, they've arranged themselves both in terms of how often they meet and which way that they meet. And that is kind of optimized. Then you have battles of ideas. So Scott and I ran into this for sure. Where you know kind of the biggest story is that

Conflict That Builds Better Teams

SPEAKER_01 22:05

one day we were co-located, and one day he had his noise canceling headphones on and he had for three days. And I had a client that had been having some conversations with, but it became really urgent that I needed to get his attention to solve a like to answer a question that I couldn't provide an answer to. But it had gotten to the point where it was urgent. So I was like, I don't know what to say, I don't know how to get his attention. So I was like, maybe I can pim him on Slack. So what I did was I ended up walking over and kind of sheepishly waving my hand and saying, Do you got a sec? And what ended up happening, I took a deep breath because at that point I could rage quit and be like, What is wrong with you? How, why do you do this? I've worked with so many people over the years. No one has had this reaction. But he calmed down too, and then we had a really great conversation. And I learned that inside his mind as a software developer, his job was to build crystal palaces of mental models. And what had happened was that he was so close to solving this problem that just asking, like trying to figure out the analysis of how to answer the question, you got a sec. It's he was like, What is this about? How deep am I in? How long is it going to take? Because it's not a second, it's probably 20 minutes. And in the analysis of just trying to figure out whether or not that was a yes or a no, he watched all of his work crumble and he described it as, you know, inception, if you've seen that movie, where just the streets of Paris are folding over. And I was like, Oh my gosh. Once I knew that, I was like, I don't ever want to do that again. So we sat down, we had conversations, like, how can we solve this problem? Because we have to have a solution where I can interrupt you elegantly. Because if I can't do that, like this is a huge business problem. And so we were like, why don't we just use the word inception? And if it's zero, then I know that I'm available. If it's a five or you don't respond, at least you know that I need you. And it just eliminated all of the organizational friction kind of overnight. So then we made that part of our organizational culture too. But that wouldn't have happened unless we had had the conflict. Yeah. So, you know, when we have different ideas, when we can use empathy, when we can use effective communication, we literally get into a one plus one equals three situation. But the important thing is that the type of conflict needs to be about the work. It's called task conflict and not about the people. So not character conflict. That becomes just a vicious cycle that was the crucible of toxicity. But when we can shift that into task conflict, and the way that we do that is through empathy, through understanding communication as a system, then we really get the benefits. One, I see kind of individuals who don't know how to speak up, or you know, a lot of times they're stuck in how, like they're describing a lot of implementation details, but they have to translate that into business impact and why it matters to different stakeholders. And so that translation skill is something that we work on. But I also work with teams, and so one of the teams that I worked with recently, one of the things we discovered was that no one wanted to talk about any kind of dissent. One of their core values was be kind. And so the way that people had interpreted that was kind of a shadow side, where then no problems were, and they were going through a huge migration and there were technical challenges, but nobody wanted to say anything because they didn't want to go against the culture. So to me, again, it's that pragmatic point of view. It's like, let's look at this as a system, let's, you know, figure out how to debug it, what is specifically going on here, and leaning on all of those heuristics that work with both computers and with human language and relationships. And I think that's where we can just get at the heart of what the challenge is, but it also can help us create context-specific, you know, customized ways to solve our own problems in our own contexts.

SPEAKER_00 26:18

I want I want to talk a little bit more about the empathy thing that you keep bringing up. So, one of the most compelling ideas that I think I've heard you say is that empathy is a technical skill. What does the operational empathy actually look like inside of a high-performing team?

SPEAKER_01 26:34

Yeah, so this is something that, you know, this is kind of a talk that you know put me on the speaker track, and it's funny because I had this was the highest compliment I ever received. I had a software developer at one of the conferences, and a lot of people like lean back, really skeptical, hands, hands crossed, and he came up to me afterwards and he was like, I came to this talk specifically to tell you how wrong you are, and now I can't. So I think the biggest thing again here is to look at what empathy is and look at what empathy isn't. I think that one of the big problems that we have with empathy, kind of culturally, is again that we use it as virtue signaling and we use it as in the same vein as software or soft skills, where it's like, I'm highly empathetic, you know, empathy is my superpower versus someone who is like, I don't know. I used to be someone who was like, oh, empathy is my superpower. But what I realized was I was actually operating out of hubris because I was operating out of the assumption that I could walk into a room and I could understand how someone else felt better than they could understand themselves. That is not empathy. That is hubris. That is hubris, right? So we think of empathy as this like intuition and being able to do that. And there is a bit of that, but the research shows us that what empathy is is that it is a biological, functional, evolutionary system

Empathy As A Technical Skill

SPEAKER_01 27:55

that enables us to basically communicate faster so that we can collaborate and solve complex problems together. Because humans are the most hypersocial species on the planet, and the way that we survive as a species is through technology and through modifying our environment. So when we have the probability that I meet someone that I don't know, and William Ice has some really great work on empathic accuracy, if if folks want to dive into this more. But if we were to meet someone and it was just pure chance that we were able to under, you know, name what somebody else is feeling, right? It's about 5%. But humans, even with strangers, like that jumps to 20%, where I can have a complete stranger and it's like I kind of get a sense of how they're feeling like that is a huge advantage. So the empathy superpower people are usually right about four out of 10 times. So I like to think of it as kind of like batting averages, like you know, three out of ten, I'll get you in the hall of fame. And what I've noticed is that with the software developers that I've worked with, they're pretty accurate. So when we break empathy down into its component parts, it's kind of three. So the first is capacity. So empathy is something that fluctuates up and down based on context. If I'm sick, if I'm tired, if I'm overwhelmed, if I have too much to like process, too much information. Empathy is hard, right? Even with people that we like. Right. So what this is, is that you know there's two different systems in our bodies. There's the sympathetic, which is the fight or flight kind of nervous system. There's also the parasympathetic, which is the rest and digest, is somehow how it's called. And so what these do is they when either of these systems is are activated, they prompt us to think about survival differently. So when our sympathetic nervous system is activated, we are focused on self-survival. You know, we are focused on preserving our personhood, right? So that we're not attacked by a tiger or whatever. When we are activating the parasympathetic nervous system, we are focused on social survival. We're focused on survival as the group. So this is why capacity is so important because it's about biologically aligning ourselves so that we can think about, you know, how do we optimize our communication? So there's kind of two steps there. So having the capacity that is essential for being able to activate empathy. The next is connection. So this is both from an emotional point of view. It's not necessarily about naming the emotions. This is the circumplex model of emotion where there's kind of two different axes. The first is activated or calm, the other is positive or negative. And so being able to notice, like, oh, someone is calm and passive, like I can probably approach them that way, or someone is like there's a negative, highly activated, and then you kind of feel it in yourself a little bit, but you don't necessarily name it because, for example, the way that somebody is processing grief, that could be what's driving that, or it could be fear. And so you don't necessarily know that. And so that's where then the you know more logical side of our brain comes in, where it's okay, what's the context? What are you feeling? You know, what are your motivations, those kinds of things. So the connection piece is both the cognitive empathy and then also affective empathy. And then the final is compassion. So this is about motivation, and compassion in a technical sense is about recognizing suffering and then being motivated to do something about it. And you know, one of the things that I say that is a big part of the empathy as a technical skill is that refactoring, which is going into a code base and making it easier to work with. So that could be having more white space, it could be naming your variables in a way that's intention-revealing. Like there's a lot of different techniques, but a lot of it is kind of for humans. Refactoring is inherently an act of compassion because you're looking at how the code could cause problems, and then you're being proactive about eliminating the pain of either yourself in the future or someone else in the future, and you're you're doing something now to alleviate it. And so, if we just look at it from a scientific point of view, that's what it is, and that is what you're doing. And so, all of these people who have been told, oh, you're a bad communicator, you're bad with social skills, you know, you don't have soft skills, all these things. It's like, no, you're actually operating like really well in this, and again, looking at it systemically, so breaking empathy down into like what is it really, and then how is it being executed really? And there is just so much interesting things around how culturally there are a lot of mismatches in how we assume what this is, and there are a lot of places where we have overlooked, like refactoring, you know, just how compassionate that is of an act, whereas otherwise people are like, ah, whatever, it's a waste of time.

SPEAKER_00 33:18

So wow. Yeah, no, a lot of that resonated with me. As you know, I'm I'm going through performance management at at the office over the last month or so. And a lot of these little like nuances you just spoke about are coming out. I was out and going through a tough, tough trial at at home kind of a thing. And and I'm I'm picking up on a lot of what you're saying right here and challenging, trying to reflect in the moment, which is really hard when you have so much goodness to listen to, right? I'm I'm listening to you and I'm trying to see like where do I place myself on this spectrum of zero to 20 and and all of that, and it's all coming out. It's like flooding out, like it's last time maybe it's flooding in it. But I I I get the sense that there's so much more that I'm just I'm gonna really value in the future. So thank you for all of that. So, what what would you say happens when organizations automate communication? Just thinking about this, right? Before they build clarity or trust into the underlying system.

SPEAKER_01 34:25

Yeah. That's such a great question.

SPEAKER_00 34:27

Because I think that's can I just say one more thing about that?

SPEAKER_01 34:30

Yeah.

SPEAKER_00 34:31

So as you said anybody in and around AI-driven organizations, especially ones that are building their own software, are infusing their processes with agents and skills and any number of systems that are uh trying to make them more efficient. So I'm placing myself in that person's role. They're on a team, an agile team, maybe, they have a job to do, they're they're racing at the the speed of light with their coding agents. What are they missing here, right?

SPEAKER_01 35:05

Yeah. I think some of it is understanding what the technology does. And, you know, again, in in like where's the right tool to deploy. And I think in that, human communication definitely is something as well. So the way that I, you know, I kind of think about all of these different concepts, and you know, there's agents and there's all these, you know, different things. It's like, what is that really? And some people are like, oh, it's it's a virtual employee. The way that I think about it is that you know, is kind of like a car. Because as humans, again, our goal is to modify

Automation Without Removing Humans

SPEAKER_01 35:45

our environment for our species survival, right? And we do that through technology, right? So, this technology, like, how do we want to modify our environment? So that's the goal, that's the destination, right? Is figuring out so you know, if we think about it then, we've got the car, right? So I think of that as kind of the agent where it is a vehicle to get information like from one destination to another much faster than you know, say an you know, we could before, than a bicycle. We may have been operating at bicycle level before, where it's like we could only go so fast. Now agents they allow us to move much faster and go to our destination faster. We also have the road. I think of this as MCPs, right? Where that's the connection point, right? That is what enables a car to move quickly. So if we have a fast car but no road, right, it's not gonna be optimized. We also have directions, and so this is skills. And you and I have had different conversations around skills. I think they are one of the most underutilized forms of technology when we're looking at LLMs. They're the turn-by-turn directions, because otherwise you just have a car that is just going, but it doesn't know where it's going. So skills are such great instructions, you know, especially for automating customized business processes. But the other thing is that we have the driver, right? Or we have the passenger, right? If we're looking at self-driving cars, right? Or we have the person who like plots the destination, right? And says, here's how I want this thing to go. And if we forget that, we get car crashes, right? So we have to recognize that like I am a huge human in the loop, you know, fan, where you have to have humans assess is this system doing the thing that I want to so that it can achieve its goals. And so the automation isn't the problem. Automation can be great because it helps us achieve things faster. It is when humans extricate themselves and say, Oh, I don't have to be part of this process anymore. I can just put it completely on autopilot. Okay, if it's working fine, you know, great. But they're putting the system on autopilot. And because we're working with a complex system, things are dynamic, things change, and things break. So you have to have people involved just to manage the entropy of the situation because you know, nothing stays the same, especially when we're looking at software systems that interact with our environment. And what LLMs do is they make that entropy happen much faster. So, you know, to me, it's again about recognizing like what are humans good at? Humans are really good at assessing the goals, humans are really good at figuring out assessing is this on the right path? And there are a lot of different ways that you can automate that to achieve your goals faster, but you know, just completely eliminating humans, I think that's a recipe for disaster.

SPEAKER_00 39:03

Mm-hmm. For sure. So the the clarity there, can can we double-click on how they can build clarity into what they're asking for from these agentic systems?

SPEAKER_01 39:16

Yeah. Skills I think are great, right? Because what skills do is they force you to really think about kind of those turn-by-turn directions. And you can't be clear about an automation unless you're clear about the manual process. And so the clearer you are about what you actually want the system to do, that's where human collaboration comes in. Because it's, you know, like the automation breaks down, and and this is regardless of whether or not we're using, you know, LLMs or something else. But the best way to automate is to have a really clear business process that's working manually, and then you optimize for it. So that's where humans really, you know, because again, it's like, what's the goal? What are we measuring? How do we know if it's effective, right? And figuring out and articulating in a very clear way, here's what we want for our organization is this, this, this, and this. And then the power comes in when you're able to use skills and MCBs and agents and all this stuff to do that faster, or at least at the rate that is appropriate. You know, sometimes you just need a Cloudflare worker that like pings once a day. Sometimes you need a full-on pipeline that's just going to be, you know, automatic all the time. But again, this is like where you get token optimization too, where it's like, how what is the appropriate level of you know, of tokens that you need in any given context? But yeah, I think that's where the humans, it's it's really articulating what is the thing that we are optimizing for.

SPEAKER_00 40:59

All right. So you you've definitely spent like over a decade, 15 years, you say, like working with legacy code and systems. How would you say organizations carry those legacy communication patterns and how would they hold them back?

SPEAKER_01 41:15

Yeah, I there's there's a huge discussion kind of in the legacy code refactoring code craft community, because I think this is one of the first areas that was jettisoned from the, oh, we don't need this anymore. We can have LLM, we can use Claude or Cursor or you know, Copilot or whatever to replace and do the automatic refactoring. But I think that, you know, my my thing is that in the next couple of years, you're gonna see that, oh no, we actually do need experts who are in here. Because again, it's that what do we automate? How do we think about what the best path is to do that in order to get our business? And I think this is where the communication skills for software developers then become so important because it's not about writing in a programming language anymore where you are typing out the dollar signs and the percentages and all of those different programming language syntax that feels like an alien language to 95% of the population. Now you're able to collaborate with domain experts in real time. And a great example of this, I spoke at the big agile conference last year, and there was a hackathon, it was called Hack for Good. I was organized by April Jefferson. Like I have seen so many of these hackathons where it's

Legacy Communication Debt And Fast Feedback

SPEAKER_01 42:40

like good intentions, but then things end up in prototype purgatory, right? There's nothing that actually of value that gets deployed and used. But what I saw was that in three days, they had brought in an organization that specialized in eradicating child abuse and neglect, right? Yeah, let's do that. Awesome. And so by using a methodology called FastAgile by Quentin Cartel, which is more about making feedback loops real time. What I saw was that the speed at which you could get features out in terms of prototyping made it so that the domain expert could immediately give feedback. So the problem that we focused on was that there was a really, really useful intervention, but it wasn't being used because the data entry for the therapists who were like in the field was so burdensome that they just didn't use it. And so they were at risk of losing a grant. So we developed an iPhone app of okay, what about this? What about this? And she was like, This is great, but I would want the numbers in this way because that's kind of how it works with this. And I would love to have it in colors and I would love to see some like haptic feedback. And it was like, okay, Now try it. So over the course of three days, there was something that was like actually working. And now it's being used. It's being rolled out to, you know, other organizations. More grant money is going in to, you know, things that are more effective. And you know, we were all in tears because, you know, the you know, the domain expert who was leading this was like, this will save lives. And to me, that's why I'm like, eh, it's not always bad. It's about recognizing that you know, humans interacting with each other to figure out what is the goal, right? Different ways of using AI can help us have those better conversations. And I think one of the biggest shears is between developers and domain experts. But this is why human communication becomes so essential with software developers, because the you know, the skill of having to communicate with humans and translate kind of those implementation details and to more higher level features and things like that, it the feedback loop was so much slower, but now it's becoming so much faster. And that's why that human communication skill is becoming so much more important, not just for software developers, for everyone in the organization too.

SPEAKER_00 45:14

In an AI accelerated world where information is abundant, that is absolutely true. What other human capabilities do you believe are are becoming truly scarce and um but valuable?

SPEAKER_01 45:27

Yeah, I think critical thinking, for sure. Deep listening, right? So being able to take time, kind of think through, I think, you know, having a culture and kind of having the operations to activate our sympat parasympathetic nervous system, where then that way it's like you know, I have worked at different organizations over the years, and some of it's like, oh, we have this beautiful campus and like you can go outside, but like culturally, it's like no, you can't really do that. Like, if you do, then your scene is not working. And I think that's some of it too is looking at culturally and figuring out what is the importance, because a lot of times we think if you're not typing as fast as you possibly can, and if you're not deploying something as fast as you possibly can, then you're not contributing value when sometimes, not all the time, right, slowing down so that you can think critically, so that you can activate that social, you know, goal, you know, idea where it's not just that individual survival, it's the social survival, right? It looks like you're not necessarily working, right? Because of because that's how we've defined productivity right now. But you know, so I think assessing kind of what are those norms and you know, figuring out what are kind of the operational tools that are going to help us kind of get there effectively.

SPEAKER_00 46:53

When you said, I had to chuckle, sorry, and when you said it's beautiful outside, but I can't go because I'll be looked at like I'm not working. I have experiences where people make those statements and I'm like, what is wrong with this, with the culture here, right? Like, why do you believe that for yourself?

SPEAKER_01 47:11

And so then it's like they feel like they can't do it, so they just don't do it, right?

SPEAKER_00 47:15

Yes.

SPEAKER_01 47:16

So it might not be that anything is just wrong, but we're just conforming. So Edward Schein or Edgar Schein, I really like his work on what culture is, and so he defines it as three things it is shared behaviors, shared expectations, and shared learning. And so, you know, if we don't explicitly say, please go outside, it is important to our culture, right? This is how we get creativity, right? Then, you know, there are explicit, you know, expectations, but then there are also implicit expectations. And so if we don't see the behavior, then we might imply it's not useful or it's not allowed, but it's not actually not allowed. We're just we could go outside, and then someone's like, oh, that's okay, maybe I'll do that too. Yeah, but some of it is just defaulting to this is the way that everybody operates around me, and I don't want to risk being seen. Amy Edmondson has some great work in her book, Teaming Around Image Risks, where people don't want to be seen as ignorant, disruptive, negative, or there's one more.

SPEAKER_00 48:24

So, what's the difference between a team that communicates efficiently and a team that communicates intelligently?

SPEAKER_01 48:32

Yeah. Again, I think that this is kind of not mutually exclusive because I think that a team that communicates intelligently does communicate efficiently. So if we look at again that Shannon Weaver model, right, the idea of what effective communication is, is I am getting an idea from my brain into your brain as quickly as possible with that with at the highest fidelity possible. So without losing information, right? And that means that conceptually, I need to figure out like what is the thing I want you to I want to say, and then what is the thing I want you to understand. But there also is that like, how can I get that message there quicker? And so there are things like feedback loops. There are times where, you know, especially when things are very complex, you want people to be on site because you need all of that rich information. You need people to be bumping into each other and having those different conversations. But for this, something like this, it would be really hard for us to schedule a time where we were in the same studio, right? So we're able to get the same fidelity without having to be in the same physical location. So again, it's like how do you structure your you know communication infrastructure? And so, how do you look at the channels and you know, so Slack and how all of that is organized and built so that you can get messages across the network efficiently?

SPEAKER_00 50:01

I really need your brain power on the synthesizer that we're building. I think about that a lot. Whenever I'm building out that system, I'm like, what would make this better in these particular areas and developing this new workspace for the 60-day roadmap, the mapping exercise that we're doing here in AI Ready? It has me rethinking some of that structure and how we can leverage different work streams and things like that, and how

Critical Thinking Listening And Culture Norms

SPEAKER_00 50:30

those work streams can coordinate amongst themselves and more operationally across the board. So it's fascinating. We're actually recording a podcast.

SPEAKER_01 50:40

Yeah, but these are conversations that we're already having about the communication infrastructure. Right. So for example, you know, there's a lot of information happening over on Discord, right? That's the channel that the organization has chosen. The idea isn't necessarily that we have to move where those conversations are happening, but we do have to be able to, you know, capture those conversations so that they can be analyzed. So that's an example of building some automation where what's a technologically appropriate way to do that. And then you start having conversations around privacy and you know, speed and how much and things like that. And again, that's where the human conversation becomes really important. Because if you just go, oh, let me just scrape all of it, it's it's like that's where some of the problems come in. Using these tools to enable us to have better human conversations so that we can be on the same page around, okay, what is the actual goal? What are the priorities? And then what are some of the automation you know techniques that we can use to achieve our goals faster?

SPEAKER_00 51:43

Yeah, yeah. Okay, cool. I want to circle back to how communication develops in organizations. And, you know, leaders know that it's it's super important for their teams, themselves, their their thought processes. Now tell me how. Sorry, I keep getting distracted.

SPEAKER_01 52:02

You're fine.

SPEAKER_00 52:03

Yeah, I was thinking of your camping trip and the board meeting, and my mind is going in many different directions. All right. I'm like, I'm seriously overwhelmed with this conversation, and I'm gonna get that a lot. I I love it. I do. And I'm also becoming cognizant of the time that that is it's gonna exceed an hour, which I haven't done before, and I don't know that there's like a real rule around that, but I imagine it does lose its effectiveness.

SPEAKER_01 52:34

Yeah. Well, we want to wrap it up. We can totally do that now. This is probably a good place.

SPEAKER_00 52:40

Yeah.

SPEAKER_01 52:40

Unless there's like a burning question that you've got.

SPEAKER_00 52:43

I really, yeah, I really don't, but you've been so thorough, is the thing, right? It's like, gosh, like most people do not answer so effectively and thoroughly. It's just they just don't have as much to say and they don't have all this industry knowledge and yada yada, like different experiences. Yeah, I'm I almost feel like I'm a I'm I'm in your audience when you're speaking, you know? So it's like Yeah.

SPEAKER_01 53:09

Yeah, and I think this is kind of what my clients like because then it's not a, oh, you should make eye contact all the time. I think this is kind of where a lot of communication coaches just if you don't have the other understanding of how systems theory and like how all these things work, like you know, you can end up giving advice where it's like, okay, in that context, that might not be the right thing. So for example, one of my clients, the challenge that he was working on was that he had kind of an all-hands meeting for his work. They were gonna bring in like the 200 people on the IT team to a physical location. And he was like, I am going to hate this. I'm introverted, I'm shy, I like don't know what to do when we talk to people. So everything that I do on my coaching is all around experimentation. So what we discovered in kind of the the process of decomposing this problem was that the reason that he was struggling wasn't necessarily that he was shy. It was that he was analyzing the conversation before it even happened. And so a lot of times decomposition is another important thing that we do in software development where we're taking, you know, a problem and we're breaking it down to its teeniest, teeniest, teeniest, tiniest thing that can be tested. So you've heard atomic habits and things like that. So what we ended up doing was like, okay, well, how can you make sure that you don't that you're not spending all of that cognitive overwhelm? And also you like, you know, you've heard eye contact, but you don't feel like you're going around and staring at everybody so that you feel really uncomfortable. And so the experiment that we came up with was just pay attention to the expressions of the people around you. And if you see someone else make eye contact with you or do a head nod or do a small smile, all you have to do is mirror what they do. That is the only thing you have to do. You don't have to come up to them, you don't have to talk to them, you just mirror. And so a few weeks later, he came back and he was like, I don't know what happened, but that was the best time I ever had. I connected with so many people, I didn't realize that there were so many people who were passionate about refactoring, and so now I'm creating this group. And so it's like all of that from you know, we think that communication is this like big, huge thing, but the way that we navigate complex systems is a lot of times by just breaking them down into their smallest component parts and then executing the system. And then you try and it's like, whoa, that actually had a lot of impact. There's again in systems theory, this is called leverage points. Donnella Meadows has a great book, Thinking in Systems, that goes over this. So, but yeah, and I think that's that's part of it is

Small Experiments And A Mindset Shift

SPEAKER_01 56:00

that you know, that feeling of overwhelm is what a lot of people feel, right? And when I'm naming out loud, it's like here are all of the different possibilities and here are all the different things, but also here are all of the heuristics that we can lean on and say, Oh, in order to achieve my goal, I know that leverage points are a thing. So then we can bring it down to what's the smallest problem, but what's the smallest thing that we can test to observe in each interaction? So, you know, I work with my clients where it's every single interaction is a tiny test. Okay, try this thing. Like, don't even think about any kind of conversation or anything, just mirror back. If somebody nods their head, you nod your head back. That's all you do. That's all you do.

SPEAKER_00 56:47

Okay.

SPEAKER_01 56:48

And so then every interaction becomes, you know, a mini experiment. And that's how we learn. That's the scientific method, that's how we explore patterns. And it takes this burden of I'm not a good communicator to, you know, I am learning how to communicate effectively with the people that are around me in the context that I'm in, given the goals that I have.

SPEAKER_00 57:13

That is great advice, by the way. So I want to wrap it up here, Andrea, with one final question, because it takes not just you know listening to great advice, but also having your own mental mindset shift, right? And I know that intuitively and academically because of the you know books and experiences I've had in life. But if you could give one final piece of advice for a single mindset shift for navigating the next few years of AI-driven change, what would it be?

SPEAKER_01 57:53

I truly believe that if we start to look at uh human communication as a system and understand its properties and how it behaves, and we focus on that, given all of the things that, like systems theory and Conway's Law and all of these things, that that is how we create healthy software systems that will help us achieve the goals that we're looking for. And sometimes that is pausing and allowing ourselves to have more empathy. Sometimes that is going super fast because we have a challenge that we need to resolve as quickly as possible. But paying attention and really thoroughly understanding the science of how humans communicate, that's going to directly translate into how we build our AI systems and then also how we automate, how we leverage the tools, and which problems we focus on solving.

SPEAKER_00 58:51

Wow. Okay. That was fun. I thoroughly enjoyed that conversation. Thank you so much for joining me on the podcast today, Andrea. We got a lot of work to do to help this community, and I'm super grateful for your contributions. So without belaboring in any any further, thank you again, and I look forward to working with you on the synthesizer and enjoy your camping trip.

SPEAKER_01 59:19

Thank you.