Lessons from Corporate Innovators

A Deep Dive on Leveraging Generative AI with Marshall Kirkpatrick

Play episode

Episode description:

Explore the transformative power of AI as special guest Marshall Kirkpatrick, Editor of AI Time to Impact, unravels creative ways of integrating artificial intelligence into different facets of your professional life.

We discuss Marshall’s journey and gain valuable insights into the inspiration that propelled him to initiate his newsletter. 

Don’t miss this thought-provoking episode on the Agile Giants Podcast, where innovation meets practicality. Tune in now to discover how AI can be a game-changer and how we can strategically use it to our advantage.

Show links:
X- https://twitter.com/marshallk
LinkedIn- https://www.linkedin.com/in/marshallkirkpatrick
Website- https://marshallk.com/
Subscribe to the newsletter: https://aitimetoimpact.com/

[00:00:00] Sean Ammirati: This week’s podcast guests and I go way back. Join me as we welcome Marshall Kirkpatrick, editor of AI Time to Impact. Marshall started his career as a news-breaking technology blogger, which led him to be the first writer hired at TechCrunch. He then went on to co-edit the tech blog ReadWriteWeb, where we worked together.

[00:00:27] Sean Ammirati: After that, Marshall founded a startup called Little Bird, which he sold to Sprinklr in 2016. He recently launched again, AI Time to Impact, a newsletter dedicated to capturing the most captivating and crucial developments in AI. And I’ll just say, I am really enjoying the newsletter. And just to editorialize for a moment, I would really encourage you to go and actually subscribe to it.

[00:00:52] Sean Ammirati: As recently as last night, there was an issue that came out that the first time I saw about things that were incredibly relevant to what I was doing, I saw on his newsletter. So again, we’ll include a link in the show notes to this, but would really encourage you to subscribe. I also really hope you enjoyed this week’s conversation on Agile Giants.

[00:01:10] Sean Ammirati: All right, Marshall, thanks so much for joining me today. Maybe just to get started. Could you talk a little bit about the inspiration behind launching your newsletter? Sure. 

[00:01:19] Marshall Kirkpatrick: Thanks, Sean. I am excited about a lot of things on the internet. I love to learn and I love to use new tools that I discover to do new things.

[00:01:32] Marshall Kirkpatrick: And so the newsletter, AI Time to Impact, was born of that inspiration in multiple ways. I’m working on some news discovery methods. I’ve been working on news discovery methods, you know, my entire career, as you mentioned. But, continuing to evolve those focused on watching what experts discuss at scale and using data to bring the most important developments in those conversations up to the surface.

[00:02:04] Marshall Kirkpatrick: And AI is such an exciting field right now that it was a really logical place to do that. And I’m finding new tools and new conversations. Every day, rising to the top of the conversation in the world of AI and then writing up short summaries of them so that the readers of the newsletter can use those tools themselves and have some view into what options are unfolding for us collectively in the future as well, thinking about how these technologies will impact people, looking at short term versus long term impacts.

[00:02:46] Marshall Kirkpatrick: And trying to illuminate some of the options ahead of us, both individually and collectively. 

[00:02:54] Sean Ammirati: Yeah, so I do want to get to like discovery news and that kind of stuff, but, but you’ve sort of for a long time been at sort of like the breaking point of what’s happening in technology, but both sort of breaking news at TechCrunch and at ReadWriteWeb, but also I think.

[00:03:11] Sean Ammirati: Putting that news into context for people. And then obviously you did that systematically at Little Bird. How do you think about the moment in time we’re at right now in technology relative to kind of previous transformations you’ve chronicled in your career? 

[00:03:26] Marshall Kirkpatrick: Oh, I think that there are some big similarities and some big differences.

[00:03:33] Marshall Kirkpatrick: I think that in the early social media, big data eras, there was a real tangible sense of excitement that there were a lot of things newly possible that many people would be able to, to publish, to analyze, to, to learn, to instrument, new, new processes. In ways that hadn’t been accessible before. And you know, that had both vertical implications, if you will, on things that were newly possible and horizontal implications, in terms of who could perform those actions.

[00:04:17] Marshall Kirkpatrick: The AI revolution, if you will, is similar in that, it also feels like there is a risk. Similarly, there may be a lot of disappointment. There may be a lot of unrealized potential. There may be a lot of unanticipated negative consequences, and I really want to do what I can to point towards the positive and the opportunities.

[00:04:46] Marshall Kirkpatrick: And illuminate some of the best voices, shining a light on ris in this new transition. I think that the way that it is different, that the first way that comes to mind that it’s different is that the technology, on some level, it is much more powerful. It is more general purpose.

[00:05:12] Marshall Kirkpatrick: It is even more accessible. Social media enabled anyone to do a whole lot of work. And speak to the world, but AI does a lot of work for you and that makes entirely new things possible. 

[00:05:32] Sean Ammirati: Yeah, that is definitely true. Let’s talk for a moment though. First on the similar. So the shining a light on experts, you know, without disclosing like the full list, like who are some experts maybe that you’ve come to appreciate who, like maybe a lot of people aren’t.

[00:05:51] Sean Ammirati: Paying as much attention to, and you’re like, “boy, I would love to shine the light on a couple of these people because you do a nice job”, I think, tapping into the wisdom of experts, not just the wisdom of the crowds, but the sort of wisdoms of experts in these fields. 

[00:06:04] Marshall Kirkpatrick: Well, there are three people that come to mind right now.

[00:06:09] Marshall Kirkpatrick: One of the most connected nodes in the network and inspiring people in the arena, as he likes to say. As a man who goes by the name swyx, and he is an engineer, but he is sort of the public face of what he calls the AI engineer community and recently organized, an event called AI Engineer Summit. And around his orbit, there are all kinds of exciting new technical things being created and discussed all the time.

[00:06:50] Marshall Kirkpatrick: He’s a real smart thinker about everything that’s unfolding. So I would point towards him. The other person, number two that I would point to is a man named Simon Willison, and Simon is a guy who was a real innovator in the social media world as well. He’s been around for quite some time and he is also a developer.

[00:07:16] Marshall Kirkpatrick: But really explores the implications and the possibilities of new technologies immediately after they’re available. He’s a really good public speaker. And then after he speaks publicly, he publishes lengthy explanations full of links and demos on his blog. And he’s also a really nice guy.

[00:07:42] Marshall Kirkpatrick: So I get excited whenever I see what he’s talking about. And then the third person I would point towards is a woman named Joanna Bryson and Joanna is a British tech ethicist. She’s a technologist, first and foremost. But I was just listening to her interview with Azeem Azhar on the exponential view podcast, and her discussion of the technical considerations around risk.

[00:08:16] Marshall Kirkpatrick: I learned a lot and every time I go back and read things that she’s written or watch her speak for quite a few years, I’ve learned a lot, felt inspired. 

[00:08:33] Sean Ammirati: It’s awesome. All right. Well, shining the spotlight on a few other people as well, which is, which is always useful.

[00:08:41] Marshall Kirkpatrick:Tthose are just three. There are so many, and it’s such an accessible world, too. There’s a place for anybody to jump in. I mean, one of the things that is notable about this similar to social media is that there are entirely new voices that are showing up, adding a lot of value and quickly growing, visibility and communities around them.

[00:09:06] Sean Ammirati: Yeah. Totally agree. So to this point, to the other point you called out though, this sort of, you know, human and computers and computers doing, doing more of it, maybe as a way to unpack that, you could talk a little bit about like creative ways you’re using AI in your professional life today. 

[00:09:25] Marshall Kirkpatrick: Oh, there are, there are so many.

[00:09:27] Marshall Kirkpatrick: I feel like whenever I do anything and I haven’t asked an AI for advice. It’s a shortcoming on my part. It’s a challenge to just consistently go to ChatGPT, go to Claude, go to Perplexity. And expand the, the cognitive power available for any kind of task, but especially strategic tasks. I feel like, you know, if there’s certainly a lot of little things, you know, reformatting things, extracting data.

[00:10:12] Marshall Kirkpatrick: Very tactical stuff that is faster and easier now than ever before. But I find it a lot more interesting and a lot higher leverage to, to have conversations with these AI platforms about strategic questions. And they’re just such a, a well-informed sparring partner, right? I mean, preparing for this podcast, I certainly asked for advice from ChatGPT about the best ways to prepare for a podcast.

[00:10:56] Marshall Kirkpatrick: And in my experience. It’s one thing to ask an AI a question. It’s another to ask the AI what questions you ought to ask yourself. So that’s one angle I would take. And then, I find that whenever I get a response regarding something important, you know, and different possibilities that are available to me in a project that I’m working on, it can be really powerful to, similar to the 5 Whys model, if not literally using five whys, just drilling down layer after layer after layer, decomposing a task or an option or a strategic perspective into ever smaller parts of the AI never gets bored.

[00:11:49] Sean Ammirati: Aaron from Box calls it a late-night brainstorming buddy. And I think that is like a relevant part of this sort of, add perspective, different perspectives, complementary perspectives, things like that. And then I think to your point, it’s less interesting, but it is huge leverage to just also automate those routine cognitive tasks.

[00:12:14] Sean Ammirati: If you’re comfortable, maybe one or two more of these sort of adding perspective things, just you know, think times you’ve used it where it’s been like, “Hey, this has really helped me”, you know, add to my perspective, a creative use.. 

[00:12:27] Marshall Kirkpatrick: Well, I mean, one example that, that perhaps feels less creative, but it feels really important.

[00:12:36] Marshall Kirkpatrick: And I think it could be a, a good service to, to others who are listening here, is that when I reflect back on, on all the ways that I have used these tools, one of the themes that has come up is that several times, um, I’ve been able to, to use it as a sort of a, a professional sounding board for interpersonal communication challenges.

[00:13:03] Marshall Kirkpatrick: I’ve been able to, to say to one of these platforms, someone I’m collaborating with is frustrating me. I’m concerned that they’re not, communicating in, in good faith even. Or I think that they have some mysterious, motivation that is not clear to me, and I can’t take it anymore, and then after providing some context.

[00:13:31] Marshall Kirkpatrick: And perhaps some examples, some explanation of my role in a project and the other parties role in the project on several occasions, ChatGPT in particular has told me, I think you’re wrong. I think that this person that you’re communicating with is operating in a very reasonable way, given their role on the project, here’s a way to understand the perspective of that kind of person playing that kind of function, in an organization or in a project and, and how they might be seeing this as a good faith way to, to solve common problems.

[00:14:15] Marshall Kirkpatrick: And it has shifted my perspective, 180 on some challenging communication. And I am so thankful for it. 

[00:14:26] Sean Ammirati: I love that example. I’ve never tried that before, but I’m going to now. And I love that example. That is awesome, Marshall. That is really cool. There’s, 

[00:14:36] Marshall Kirkpatrick: I’m so glad there’s something about the emotional distance that, that it affords that can be useful in many different ways.

[00:14:45] Marshall Kirkpatrick: I’ve certainly also been in cases where I wanted to provide someone feedback and as a way to, to sort of depersonalize it and take some of the sting out of providing some critical feedback, I have said, let’s see what the AI has to say, and we’ve marveled at its wisdom and critical perspective together.

[00:15:10] Marshall Kirkpatrick: And I’ll be darned if some of that isn’t real good advice that we can agree. I mean, similar to the way people say that the most effective negotiation often occurs when you position yourself, not just as advocating for your own personal interest, but you know, collaborating with the person you’re negotiating with in order to satisfy the challenging third party who is not in the room, let’s come up with a solution that works well for them there’s something like that that AI can, can be a, an impersonal Oracle. 

[00:15:46] Sean Ammirati: This is amazing.I always learn something when I’m talking to you. This is amazing. I love this idea. So, I could push on that more, but I think that’s good.

[00:15:57] Sean Ammirati: I like this. So using it, that’s very creative. I will try that out. So maybe just shifting to a little bit more kind of business processes now. How do you think about kind of optimizing the business processes for, given this, this new capability as well? 

[00:16:12] Marshall Kirkpatrick: Yeah, I think that there’s a lot of potential there, because these, I mean, these large language models have billions of parameters.

[00:16:27] Marshall Kirkpatrick: They have just ingested so much written material and have weighted understandings of the relationships between words and phrases and concepts that they just offer so much. So from a process optimization perspective, I really like, well, first of all, just to ask the AI, to assist me with optimization of the process.

[00:17:06] Marshall Kirkpatrick: and ChatGPT at least will say, okay, yeah, you haven’t told me what you want to optimize yet, but here are, are 10 good steps to start. With any kind of optimization process and, you can go through, and answer, all of those or some of them, you know, and it will do a call and response. Being progressively more informed, by what you have said in the past.

[00:17:30] Marshall Kirkpatrick: One of my favorite things to ask ChatGPT is to write out a five-step process for something or a three-step process, depending on how much time I want to take to read the response.

[00:17:55] Marshall Kirkpatrick: I’ll say, okay, five step process. And then I’ll say, all right, step number two there. Break that down into five substeps. And it will, and they’re generally pretty good, and I’ll say, okay, take step number 2. 3 and break that down into five sub-steps, and it will just go progressively more and more granular, and sometimes I’ll push back and say, I don’t know that that one doesn’t seem so great.

[00:18:20] Marshall Kirkpatrick: It seems a little duplicative of what was done before, or I don’t want to do that. I want to do this instead, and it’ll revise those and, much like ideation. What you’ll end up with is this incredibly rich detailed map of possibilities, most of which, you either already knew it and now have received some validation for, or you don’t want to do and, and no problem.

[00:18:48] Marshall Kirkpatrick: It was, you know, the cost of generating those ideas was super low. But some number of them, some percentage of them are going to be high leverage, novel, unforeseen, opportunities, steps in a process or details that you had not accounted for. And the cost of generating that well-qualified list of possibilities is $20 a month, for a ChatGPT plus, you know, subscription, and it’s available immediately on demand 24/7. So I find that that’d be really exciting. 

[00:19:34] Sean Ammirati: I love that question for you, just as I was listening to you to describe that. So one of my theories is that chat is a wonderful interface to illustrate the power of these technologies. But probably not the interface that we experienced a lot of this two years from now.

[00:19:53] Sean Ammirati: And the analogy I often make Marshall is like, so, so my father was a programmer before VisiCalc existed. So before spreadsheets were a thing and a big part of his job used to be like updating financial models programmatically, not now, but in the seventies was updating financial models programmatically.

[00:20:17] Sean Ammirati: CFO’s team comes to him and says like, “Hey, if we sold 5 percent more of this and 3 percent less of that, could you rerun what our finances would look like?” And then VisiCalc came along, right? And all of a sudden, everybody in the finance department was like, “Holy cow, computers are really good at numeric computation, right?”

[00:20:37] Sean Ammirati: Which obviously he knew already he had, you know, been doing this as a programmer. But I think ChatGPT and chat as an interface, in general, is really helpful at showing us generative AI as a UI. As I was listening to you talking about this processes, like, I can’t help but think I don’t think the right UI to do what you just described two years from now is a chat interface.

[00:21:01] Sean Ammirati: Like, I feel like that’s an important enough function. I have a suspicion that when we think about business processes two years from now and things like that and process optimization, this is more likely to happen in UIs that are not. Call on-response text-based chat. I don’t know if you’ve thought about like what you wish the tool would look like.

[00:21:20] Sean Ammirati: And we’re beginning to see some things in computer vision today around this. But have you thought about if you are given a magic wand, what the UI he would interact with for business process optimization like that would look like? 

[00:21:33] Marshall Kirkpatrick: I have not, I mean, I love mind mapping,  I would love to have found an automated way to populate a mind map in great detail.

[00:21:48] Marshall Kirkpatrick: Mermaid graphics is a format that the ChatGPT will do some output in, but it’s not great. I’ve seen some people who, and I have had a lot of fun taking mind maps that I’ve made, exported them in OPML format, put them into ChatGPT and said, this is a mind map on the following process, please expand on it.

[00:22:15] Marshall Kirkpatrick: And then had a bunch of proposed new nodes, that’s you know, that’s fun. I’d love to have something like that in an automated way. We were talking about a story that came up just yesterday in the AI Time to Impact newsletter about lang chains exploring some asynchronous interfaces where they said, what would be possible if, if you didn’t expect an immediate response like a chatbot but instead gave the system five or ten minutes?

[00:22:51] Marshall Kirkpatrick: Or even longer to do extensive analysis and discovery and summary and synthesis and write a long-form report. And LangChain just made a template for that available yesterday. So I agree that there is a lot of new possibilities. I, you know, oftentimes I don’t appreciate what’s possible until I can get it in my hands and really start using it.

[00:23:22] Marshall Kirkpatrick: Voice didn’t seem terribly interesting to me until I started listening to, speeches in Spanish on YouTube with Whisper from OpenAI open on my phone. And asking those speeches to be translated, into English, then pausing them and asking for historical context around what the speaker was referencing, and that, that felt amazing.

[00:23:56] Marshall Kirkpatrick: I remember when I first saw Twitter, I thought it was stupid. And now I’ve spent the last 17 years of my life building a career largely around it. So that was a good humbling experience. I don’t know what future UIs will look like for AI. Yeah. 

[00:24:16] Sean Ammirati: I think the LangChain example is a good one, right?

[00:24:20] Sean Ammirati: The report template that they just came out with. I would say by the time this is published, that’ll be. Three or four weeks old, probably, but, but as of when we’re recording, right. But as of when we’re recording that, that came out last night and that actually. The article I referenced in the intro, that was the article we were talking about.

[00:24:37] Marshall Kirkpatrick: Will you tell me, Sean, can I, can I ask you on your, on your own podcast? 

[00:24:46] Sean Ammirati: Fire away, Marshall. Yeah. 

[00:24:48] Marshall Kirkpatrick: In that case, do you have a vision for UIs that would be powerful.

[00:25:00] Sean Ammirati: I think there’s going to be a bunch of them, right? I think just, just like you could make the argument Excel led to the entire sets of categories of enterprise software.

[00:25:14] Sean Ammirati: So ERP. All the FPNA software, CRM, right? Like it was putting that in business users hands and helping them. And so we ended up with a bunch of UIs around that. I think you’re going to see the same thing around generative AI. And I do think LangChain’s doing a nice job to push the envelope, but still it’s fairly correctly given their business strategy.

[00:25:36] Sean Ammirati: It’s fairly Python-driven today. Whereas I, I think we, we need to think about sort of tools that consumers and business users could use. It’s one of the reasons why I’m really excited about this sort of no-code tools on top of LangChain. There’s unfortunately a lag from the tool to being kept up to date with the LangChain library.

[00:26:01] Sean Ammirati: But I’m excited about those tools because they do a nice job of like letting people use them. As it relates to the business process thing that you were mapping out. Part of what was coming to mind is like memory and state of memory is going to be different than memory in a chat environment there, right?

[00:26:17] Sean Ammirati: So if I were using something like that, I, I’d want it to remember like, okay, this is my business process for these things. And then I can come back at any point and expand it as well as what I would really love in that case is to have that visually mapped out. And have some type of agent that’s constantly looking for ways to automate and improve that and giving me suggestions on a regular basis, right?

[00:26:43] Sean Ammirati: So, this is not the same thing, but to give an example of this, we had a corporate client we worked with who has a voice of customer database of basically things that their customers wished were possible in the market that they play in. And most big companies have this, right? So it’s like, oh, I, we know that in our market. There’s customer pain on these.

[00:27:07] Sean Ammirati: You know, 50 areas, and what we did is we helped them build an agent that scanned academic journals for new capabilities in the field that they were in and then compared those capabilities against those needs. Right? And so this is what I mean. By like different UI, like that’s not really a chat UI.

[00:27:31] Sean Ammirati: Now, I suspect that company would have never let us be interested in doing that had they not first experienced generative AI in a chat-based UI, just like Larry Ellison probably has a hard time creating the categories of software he creates if Dan Bricklin doesn’t introduce VisiCalc, right? 

[00:27:55] Sean Ammirati: So this is an important point, but I think the UI is going to continue to evolve and I had not thought about the business process mapping one until you were talking about, which is why I just sort of wanted to riff with you on that for a moment. 

[00:28:06] Marshall Kirkpatrick: Yeah, yeah, you know, I when I think about processes, in particular, an optimization of processes and the opportunities that get surfaced there.

[00:28:20] Marshall Kirkpatrick: One of the models that comes to my mind is, from a different context, and that’s from one of my favorite books, Joshua Waitzkin’s book, The Art of Learning. So, Waitzkin was a childhood chess prodigy, who got burnt out on being the world champion chess player, and went into martial arts, and then became a world champion martial artist.

[00:28:46] Marshall Kirkpatrick: And then wrote a book about what those two experiences had in common around building expertise in, in either context. And I, I read the book almost once a year at this point, and it’s fun and, and inspiring, but there is one part of the learning process in particular that he talks about there that that feels really relevant to me in particular when it comes to the drilling down, decomposition and the segmentation capabilities of these LLMs and that is what he calls the soft zone of being in a martial arts competition and being more aware of the moments of transition in between moves than his opponents are aware, so that he can then insert actions on his part into those moments of transition to gain a competitive advantage.

[00:29:56] Marshall Kirkpatrick: And it feels like, these AI s are just a natural partner to do that in a collaborative as opposed to a competitive way to just help zoom in deeper and deeper and deeper until we find levels of granularity that are green fields of possibilities, especially in a, you know, in a, in a social context where, I don’t know, maybe that’s taking it too far.

[00:30:30] Marshall Kirkpatrick: Maybe other people are thinking about those things as well, but, um, that’s at least help service oversights, you know, things that they may not have foreseen and perhaps opportunities to gain competitive advantage and come up with new ideas as 

[00:30:45] Sean Ammirati: I don’t think that’s taking too far. And in fact, I, I think. You wouldn’t know this because it hasn’t published yet, but the episode right before you is Karl Ulrich, who’s the Wharton professor who did the ChatGPT versus Penn student test on brainstorming, right? So he has 200 ideas from University of Pennsylvania students, and then he has 200 ideas from ChatGPT, and he compares the quality of the results from the two different universes.

[00:31:17] Sean Ammirati: It’s kind of a classic academic study on efficacy. But, but in the episode, he, he, he sort of points out, although he teaches at Wharton, he’s never taken economics and he’s not an economics guy, but he sort of says, like, if I were to put on my economics hat for a moment, right. And sort of pull on the sort of prediction machine framework a little bit, right.

[00:31:36] Sean Ammirati: When you make things like this so much less expensive than they ever were before, it just changes how you think about the things that are next to that. Right. So, you know, they came up with 200 ideas. For effectively a dollar an idea. Now, if you think about like, what does it cost a company to come up with an idea?

[00:31:59] Sean Ammirati: Right. It’s probably literally a thousand times more expensive. It’s probably a thousand bucks for a company to come up with an idea. If you think about, you know, design thinking workshops that might come up with 10 or 15 ideas, what’s the cost in human capital and service providers and resources to do that, right?

[00:32:17] Sean Ammirati: So if you make that a thousand times cheaper. It changes the value of the things that intersect with this. And so to your point on this soft zone, when there’s things that you used to think of as really hard and really expensive that are now inexpensive. Thinking about where those intersect with it, we are going to have to resort our economic models around this because now things that historically we thought of as, you know, not that valuable because it was prohibitively expensive to do the things next to it may not be anymore.

[00:32:53] Sean Ammirati: And I think, you know, this sort of learning concept of soft zones is a helpful way to think about it for sort of broadly knowledge works. I like you calling that out as well. I thought that was, that was really smart, Marshall. 

[00:33:08] Marshall Kirkpatrick: What an interesting variation o theinnovator’s dilemma, perhaps.

[00:33:15] Sean Ammirati: That’s right. I think it’s clear the argument that AJay and his co-writers make in the book Prediction Machines as well, right? Is that we commoditize this, so value flows in different ways as well.

[00:33:32] Marshall Kirkpatrick: I want to, I feel like when we were growing up, Sean, presuming that process is done, it’s not done.

[00:33:42] Sean Ammirati: All right. But, uh, I’m definitely not done growing up for sure, just to be clear. 

[00:33:45] Marshall Kirkpatrick: Yeah. Right on. Yeah. Me either. I feel like I used to, to hear people say often, you know, in the future, it won’t be about what, you know, it’ll be about your ability to ask the right questions. And I felt like that’s been true for a long time, but it feels more, more true now than ever before.

[00:34:07] Marshall Kirkpatrick: And in that kind of experiment that you’re describing, I imagine that really. There’s a lot of leverage available in, uh, in creating the right AI system that whether that’s the right prompts in ChatGPT or using an entirely different model with different documents, have you, to set up the AI to most effectively generate the largest number of high quality ideas to make sure that perhaps they can catch some outliers and not just revert to the mean of the most predictable new ideas.

[00:34:51] Marshall Kirkpatrick: And yeah, the ideas themselves are one thing, but building the idea machine and operating the idea machine and knowing what questions to ask feel like exciting, perhaps newly valuable questions.

[00:35:07] Sean Ammirati: 100%. It’s, it’s the co pilot model, right? Like you got to give Microsoft credit. They a hundred percent nailed the analogy there, you know, I’m going to get on a plane tomorrow. It’s equally scary to me that if they came on and said, there’s nobody in the cockpit or the navigation systems not working. I want to get off the plane equally quick with either of those announcements, right? So the analogy, they really have nailed the analogy there, 

[00:35:38] Marshall Kirkpatrick: Oh, well, I’m sorry to interrupt, but, um, so what are, what are the emergent qualities that, that become passable then when we’ve got both a good pilot and a good co-pilot  collaborating?

[00:35:51] Sean Ammirati: Yeah. So I’m spending a lot of time thinking about that as it relates to innovation. So I don’t, I think it’s different for different categories of work, but I also, I have this hypothesis that innovation will actually be the blueprint we use for a lot of knowledge work in the same way that, you know, when I dropped out of school at Carnegie Mellon 20 years ago to do machine learning and the kind of classic sense of machine learning.

[00:36:26] Sean Ammirati: We went to Wall Street because predicting stock prices was a pretty good place to figure out, you know, how to realize economic value for predictions, right? So the, you know, the old verbs of machine learning were predict, classify, cluster. Now you’ve got this fourth verb of create. I suspect the create verb is going to be, we’ll figure out the innovation, the economic model there, primarily starting in innovation.

[00:36:52] Sean Ammirati: In innovation, I think what you’re, you’re going to see is, New models of how people innovate. So to, to make this real for, um, people that kind of think a lot about startup innovation, we might need to think about a different model than startup accelerators, which by the way, there’s precedent for, you know, the, the, the startup accelerator model came along.

[00:37:20] Sean Ammirati: Right after cloud computing and mobile computing changed the cost model of creating startups by roughly an order of magnitude, right? Like I remember selling all the servers I bought for my second startup to go to a 500 a month credit card bill. It completely changed the calculus of doing and what Y Combinator and Techstars deserve a lot of credit for coming along with saying, Hey, the economic model has changed by an order of magnitude.

[00:37:50] Sean Ammirati: Instead of spending a bunch of time writing a business plan, let’s spend time creating prototypes and then we will get a bunch of people to look at those prototypes in sort of a gamed dynamic where we’ll get good terms for you and so don’t spend time on Sand Hill Road saying synergy and conference rooms, build software, do demos and we’ll set up A situation to make that work and viable for you and obviously today, plenty of non-software companies are going through accelerators.

[00:38:25] Sean Ammirati: But the point is we change the process of building startups predominantly because of this enabling technology of cloud and mobile. I think the same thing is going to happen with generative AI. And so you could imagine a very different funding and start creation model. So, for example, maybe instead of starting with the prototype, maybe you start with the problem space.

[00:38:45] Sean Ammirati: And you build 50 prototypes in two weeks and then you look at which of those 50 prototypes actually work out and then you build five higher fidelity prototypes of the most promising ones. And so now maybe I’m not interviewing for your ability to create prototypes. I’m interviewing for your ability to empathize with problems and pick big, compelling problems.

[00:39:09] Sean Ammirati: And, and so you could end up with a very different, you could end up with a very different model. Uh, as an example, like that’s an example on the innovation front again, I think there are other categories as well, but we need to remember that like we should be looking for these categories against that verb create just like 20 years ago.

[00:39:29] Sean Ammirati: We looked at these categories. against the verbs predict, classify, cluster, which was basically all of machine learning in the enterprise till about two and a half years ago, or 95 percent of machine learning in the enterprise till two and a half years ago. Wow, 

[00:39:45] Marshall Kirkpatrick: that’s, that is fascinating and really exciting.

[00:39:50] Marshall Kirkpatrick: It makes me want to, and maybe I’m, maybe I’m rushing things and I’ll, I’ll read your Your book, presumably, uh, or, or watch all of the episodes of your show. Uh, but, uh, what, what’s another, okay. So that, that dynamic there is, uh, thanks to generative AI prototypes, that the cost of prototyping, uh, drops, uh, dramatically unless we do a whole bunch of them.

[00:40:19] Marshall Kirkpatrick: Yep. Okay. So that’s one, you said one way that things could shift against the verb create. Uh, but one of many, will you, will you be so generous as to share with us another 

[00:40:28] Sean Ammirati: one? Sure, I’ll do, I’ll do one more for sure. Uh, so a, a different type of create that I think is interesting as well is in this sort of how do you think about scaling up the things that you’re creating, right?

[00:40:45] Sean Ammirati: So I tend to think about it as ideation, early idea validation. And then scaling the idea is sort of the general commercialization innovation process, and we’ve talked a bit about brainstorming already with the prototyping. I gave you a little bit of a of a flavor on this idea of validation. So the thing about scaling up.

[00:41:05] Sean Ammirati: Okay, so one of the things. That many of the best technologists struggle with is how to think about the right revenue model, how to kind of create the right revenue model against the value that they’ve created. And I think an underappreciated part of entrepreneurship is what I would describe as revenue model product fit.

[00:41:28] Sean Ammirati: So we talk a lot about product market fit, but there’s, and I think you and I were actually both in the room, uh, when this happened years ago with the Google founders, but I will never forget in my life. John Battelle going at Larry and Sergei at the Web2 Summit. Like, how are you going to make money? How are you going to make money?

[00:41:49] Sean Ammirati: And, you know, John is a, I would argue, one of the best interviewers in the tech business at that point. So he was doing it in, in a way that only, probably, he and maybe Tim O’Reilly could get away with. But, but with fairly aggressively, right? And I remember at one point, Larry just exhausted, says, look.

[00:42:08] Sean Ammirati: Worst case scenario, we’ll run punch the monkey ads on google. com. Because if you remember, like at that point, Yahoo had those banner ads with the monkey across the top. You know, click on the monkey and it would send you to the page, right? Now, turned out, Google correctly, and a variety of people deserve a lot of credit for this, was careful and they waited.

[00:42:33] Sean Ammirati: And they came up with a revenue model that really fit with the business that the end of the product value that they were delivering. And today that’s created a machine that prints money in Mountain View, right? So good that they didn’t go with the punch the monkey ads, I think is the is one takeaway from this.

[00:42:51] Sean Ammirati: But if we think about create as a verb for a minute, part of what you need to do once you get from ideation and idea about issues, you need to create the right revenue model to match against. The, the value that you’re delivering and working with different entrepreneurs on this, I’ve seen some really creative prompts that, and really chains of prompts to help entrepreneurs think about this.

[00:43:18] Sean Ammirati: And this makes me really excited because it takes that scale, this, that, that creates skill specifically of, okay, here’s a problem. Let me think about the right revenue model. And it kind of democratizes it in a way where I think creators who are creating magical things. That they just, they’re just not getting the, the appropriate value out of it is, is also things that you could imagine sort of being, being arbitraged away or sort of commoditized.

[00:43:47] Sean Ammirati: So, so that again, the, the cost of that goes down, we’ll use the AJ framework and the cost of that goes down. And so the value of these other things that we thought of as unreasonable before, so that’s kind of a different, different create cool term there as well. Love it. So I, I realized that we’re pushing time here.

[00:44:09] Sean Ammirati: I, I want to ask you kind of two more questions to kind of wrap this up. So, so the first is, um, I think one thing that you are particularly good at is when you see these tools, you change your workflow quickly to adopt them. I remember all the way back to the, to when we were working together at read, write web, you did an amazing job of like building tools to help you break news and then using them and we would, we would give them to.

[00:44:38] Sean Ammirati: Other people on the team and they would never get as much value out of them as you did because they didn’t change their behavior the way you did and probably a bunch of scar tissue around that. But the, but the point is you did a really nice job changing your behavior to reflect the tools that you are using and and in this point in time, I think people.

[00:44:59] Sean Ammirati: Being kind of AI default is really important. They got to change their behavior to be kind of AI first and AI default. Any encouragement you might have for people trying to make that transition to change their behavior? 

[00:45:13] Marshall Kirkpatrick: Oh, that’s, that’s interesting to, to think that that’s what the, what the problem was, uh, back then.

[00:45:19] Marshall Kirkpatrick: And it’s interesting to hear from, you know, from a friend, uh, what the. What the solution was as well. Like what, uh, why was it that I was able to use those tools in a particularly effective way? Uh, no one had told me before it was because I was willing and able to change my workflow accordingly. So thank you for that, that perspective.

[00:45:40] Marshall Kirkpatrick: Um, that’s, that’s really generous. Um, and now I’m going to go and reevaluate my whole life. Uh, you know, I, I, 

[00:45:51] Sean Ammirati: that was not the goal. That was not the 

[00:45:53] Marshall Kirkpatrick: goal. Oh, please. Is it not? I mean, that’s like the, the discursive equivalent of what you’re saying I’m capable of doing with technology. Let’s so, so to let’s do that, you know, here.

[00:46:04] Marshall Kirkpatrick: And, uh, as a result of this dialogue, uh, I mean, that’s great. I mean, seriously though, to, to level, to zoom out a little bit from that. I mean, finding out what works. And finding, you know, positive feedback loops can pretty quickly become the fat path of least resistance, if that’s not too cliche to say, um, I mean, I, I feel like I found things that, that worked well for me, that people praised me for.

[00:46:41] Marshall Kirkpatrick: That, uh, that helped me, you know, make money and get attention and break news stories, you know, and as a journalist and I, I dug deeper and, uh, and just kept going down those same directions. And there might’ve been other paths. Uh, there might have been more breadth that, uh, that I could have pursued, uh, and not been such a, you know, a T shaped or a pie shaped, uh, you know, person going deep on, on a handful of, of tools and approaches in particular.

[00:47:18] Marshall Kirkpatrick: And now that we’re entering into this new This new era with AI being so accessible, uh, I mean, I, on some level, I’m, I’m doing the same thing I’m saying, you know, when something works real well, for example, uh, on a lark, I decided once to ask Chad GPT to explain a complex concept to me, uh, using a baseball analogy, uh, because I love baseball.

[00:47:48] Marshall Kirkpatrick: Yeah, it did a great job of it. And I said, Oh, I’m going to, I’m going to do that again. And now I’ve done that probably a hundred times. Uh, but I’ve tried other things that really didn’t work as well. And so, you know, so having a spirit of experimentation, uh, going where the, where the early results are and investing in winners of, of strategies, um, I think are, are good.

[00:48:14] Marshall Kirkpatrick: I mean, generally speaking, when I think about. Uh, habit formation. I, I really like BJ fogs, uh, approach, uh, make it small, tie it to an anchor habit and celebrate each time you do it. And so sometimes I, I will say after I. Uh, do a debrief on a project that’s kind of a, uh, established anchor habit, but I’m pretty good at doing, then I’m going to, uh, I’m going to do something.

[00:48:47] Marshall Kirkpatrick: I’m going to pick up my, my chat GPT mobile app. And, uh, I’m gonna maybe, uh, narrate or, or OCR in my, my notes and ask for it to extend on my analysis or to reflect on it or give me feedback or have it. And then I feel happy and celebrate in my mind when I, when I do it. Um, but I, I mean, there’s just so, so many different ways to, to, uh, to try to build a, a new habit of using a new technology.

[00:49:22] Marshall Kirkpatrick: And, uh, and doing so thoughtfully just offers so much potential. I’ll toss out two more real quick role models. Um, I really like, uh, there’s a Forrester analyst named Rick Parish, uh, that says that when you, uh, he has a model of discipline, uh, where he says, uh, a practice that you’re doing, uh, with discipline, uh, has.

[00:49:47] Marshall Kirkpatrick: Four qualities to it rigor or following a documented best practice cadence, uh, and coming back to it consistently coordination where you’re coordinating that practice with other stakeholders and other practices that, uh, you know, in the larger context. And then the fourth quality is accountability. Um, in a, in a corporate context, making sure that someone senior who’s accountable for the project.

[00:50:18] Marshall Kirkpatrick: Uh, but in my personal context, that might be keeping a log and, and setting goals and you know, something like that. And so, um, when I, when I think about trying to innovate with discipline using ai, um, I, I think of what should be the process I follow, what cadence should I, uh, do this practice with? How can I coordinate this with other parts of my life and work and, uh, how can I.

[00:50:45] Marshall Kirkpatrick: be accountable to myself, uh, to, to actually give it a good faith, a period of experimentation. Uh, those are some things that I do and the last thing I’ll, I’ll share is that I, I do feel like there’s a, just like we’re not going to automate everything, but, uh, the hybrid models are really exciting. I think there’s a lot of offline thinking processes that can be really I find AI often powerful, not even as a literal tool, but as an analogy and, uh, and as a way to think about thought, uh, to learn about artificial intelligence as a way to reflect on the structure of my own non artificial intelligence.

[00:51:32] Marshall Kirkpatrick: Um, and one of the things that I, I do sometimes using those analogies and, and sometimes otherwise is, uh, I’ll sit down with pen and paper and just set aside literally five, maybe 10 minutes, honestly, five minutes for me, uh, is more than enough. And just write free form, uh, going deep thoughts on something I’m working on.

[00:51:53] Marshall Kirkpatrick: Um, and, and I consistently then go back and see one, two, three of the sentences I just wrote are like real keepers. And things that, uh, I wouldn’t have thought of had I not spent even that five minutes of deliberate focused attention on. So when thinking about how to effectively apply a new tool or technology to my work, I mean, take five minutes and, uh, and just focus on, uh, imagine what could be done on a particular problem or challenge that you’re working on, uh, or with a new tool that you’re learning about.

[00:52:29] Marshall Kirkpatrick: And that, that may. Be really effective for people as well. 

[00:52:33] Sean Ammirati: Yeah, I love that. Um, I think implicit in what you’re saying to is that like this stuff is often learned by doing. It’s an experiential learning discipline. And so taking this approach and having these experiences is valuable. Realizing we’re kind of actually at time.

[00:52:53] Sean Ammirati: There’s one last question for you is beyond that, beyond this experiential learning anything else that you would. Encourage people to think about doing as they try to think about the future world that we’re heading into with AI and planning for the future that’s coming. 

[00:53:10] Marshall Kirkpatrick: Well, I’m in the news business.

[00:53:13] Marshall Kirkpatrick: And so I, it is talking my own book a little bit, but, uh, I think that there is really a powerful, uh, there’s powerful leverage that comes from staying up on the newest developments in a, in a field, uh, not just as, you know, inside baseball, not just as drama, but especially in technology, this is a, a stream of new tools.

[00:53:44] Marshall Kirkpatrick: being made available, uh, to, to pick up and use in your work to make existing work more efficient or new work possible. And so, uh, the, I think that the folks who consistently follow new developments in the field, uh, are, are going to have both a, a practical access to new tools and just feel inspired with the, the sense of possibility.

[00:54:15] Marshall Kirkpatrick: That comes from witnessing everybody’s innovation all going on all the time. Um, that will help fuel your innovation and breakthroughs in your own work. 

[00:54:28] Sean Ammirati: Yeah, I couldn’t agree more. Um, so again, I said at the beginning, but I really, I’ll say because it’s less promotional coming from me than you. I really do encourage everybody to go subscribe to Marshall’s newsletter.

[00:54:40] Sean Ammirati: Like it is. It is really good, and he’s a friend, and I, I like Marshall, so I am biased, but I’m telling you, I’ve shared it with lots of people who don’t know who Marshall is, and like, it is, it is quickly becoming a must read for so many, so would encourage you to do that. Marshall, outside of the newsletter, which we’ll promote, what are the best ways for people to kind of keep aware of all the different projects and things you’re working on.

[00:55:11] Marshall Kirkpatrick: Well as my as my wife, who sends her warm regards to you. Yeah, I’ve said today it they don’t even call it that anymore Yeah, the old X.com Yeah, yeah, there is certainly that. I’m finding more and more, uh, opportunity and connection and fulfillment on LinkedIn lately than, than ever before. My personal website is marshallk.com

[00:55:37] Marshall Kirkpatrick: And if you are excited about stuff like this, shoot me an email. And I’d love to chat and hear your perspective on these matters. This has been so fun, Sean, to get to hear and catch up with you on how you’re thinking about, all of these, again, our thinking is, is similar and different in some really, really exciting ways. So thanks. 

[00:56:05] Sean Ammirati: Thank you, Marshall. So this is great, fun episode, and I hope everybody enjoyed it as much as I do. Thanks so much, Marshall.

Join the discussion

More from this show


Episode 1