Agentic Organizations: The New Era

0:00You give them more agency, you trust them more, you give them some guardrails, and you give them some data and you give them a goal to achieve and then say, go ahead and plan and execute this work and give you agency to do it on my behalf. Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters. I'm your host, Matt Page, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI. Today, we're going agentic with Alan Wenshaw, author of Agentware. And Alan is a process re-engineering and organizational change expert and really has always been on the forefront of emerging technology.

0:42AI is definitely in that category. He has experience with PwC, Deloitte, Scotiabank, as well as starting some of his own ventures. So really a wealth of knowledge in the AI space. But welcome to the show, Alan. Thanks for having me, Matt. Great to be here with you. Yeah, excited to chat. And I want to get started here. Define agentic for us, because I think this feels like a newer term in this age of AI that maybe some of the listeners just aren't as familiar with. What does it mean and how does it play into this whole Gen AI craziness we're experiencing right now? That's a great place to start.

1:23So the agentic word and the agentic organization as a term that's coming about now in this Gen AI era is really based on the idea of using AI agents to get work accomplished in organizations. So now agentic being an adjective, if you will, and agentic meaning, you know, how agentic is this organization? Well, how are they deploying this new AI technology? How are they deploying AI agents? And so there's a spectrum that's pretty much evolving. And we're in the early days of organizations, you know, thinking about how to deploy these AI agents to get work done. But that's where this word agentic sort of rises.

2:18And, you know, it certainly has to do with, you know, the nature of the future of work and, you know, the future of organizations and how you're managing this new technology. Yeah. And you think about it like, you know, you're almost giving the AI agency to do something in effect. Do you think of a real estate agent or travel agents, the traditional ones you think of? Similar kind of thing, which starts to become interesting on a couple fronts. A, because, you know, we're humans, you have to, you know, actually procreate and make humans to do that. AI agents, you know, very much easier to create in a sense and replicate.

2:57So it changes kind of the power dynamic in a sense. It's almost like I like to equate it back to like the industrial revolution, right? Going from having to like handwrite books and you would go to like a monk or something like that and they'd have to write you a book. And then you got the printing press and all these things that just accelerated it. I feel like there's some parallels there between the two and kind of what we may be stepping into in this world of AI? Well, and we'll no doubt, uh, the, the parallels are very, very clear, uh, for, um, for that kind of, you know, the explanation you gave around, you know, the, the movement or the acceleration of the technology into the organization, the parallels are clear because, you know, we have, well, we have athletic agents, we have real estate agents, they do work on our behalf, right?

3:48And so, some people are calling the early days you know assistants or even co-pilots um this this new software which which i've dubbed agentware for lack of a better yeah well um uh seems to be really resonating by the way uh so now we have uh this new kind of almost well it's software but it's almost like a version of an employee that you can direct, but is very digital, right? And you can now give agency. And this is something that we talk about as I kind of explore implementing these agents with organizations. And, you know, where at one point they're, you know, early on, they could be an assistant or they automate.

4:41And so they're kind of side by side assistant or co-pilot and then you give them more agency you trust them more you give them some guardrails and you give them some data and you give them a goal to achieve and then say go ahead and plan and execute this work and and uh you know give you agency to do it on my behalf and i get and i effectively i trust you to do it right um and come back to me with with the results so not that much different from what we as as people are expected to do in a typical role in an organization if you think about it and you just said you just you just referred to it uh i think rightly you know we've we got work to do uh today in organizations and and uh you know we could be um first of all augmenting and then potentially handing over some of this work uh completely to to some of these autonomous agents, which is where we're headed, frankly.

5:40Yeah, I like that kind of spectrum you mentioned from it's augmenting to shifting to autonomous. I feel today we very much have a one-to-one type of relationship with AI, whether that be ChatGPT or some other tool. And I think over time that will evolve from very manual, kind of synchronous, unintegrated, one-to-one, nice to have to the future where it's like, okay, it's automated, it's asynchronous, It's fully integrated with all the things, all the systems you use. And it's a many to many relationship and agents are components of that. It's not just one human to one agent. It could be one human to many agents.

6:18It could be an agent to other agents, right? So it starts to have this like network of agents that could proliferate over time. Quick break in the pod. If you're listening to this podcast, chances are you've been thinking about how to actually use AI inside your business. And that's exactly why we built the AI Opportunity Finder. It's a free tool that helps you uncover high impact, tailored AI use cases based on your business, your goals, your pain points, and your industry. No fluff, no generic use cases, just real ideas that fit your business and the ranked by ROI potential. It takes about three minutes to run and it's like having your own personal AI strategist for free.

6:57If you wanna try it for free, check out the link in the show notes or go to hatchworks.com backslash AI-opportunity-finder. I'll just build on that because this idea of kind of having a network of agents, that's really in question right now. In some regards, you know, you could consider an agent or something like, as we're speaking about it, a piece of software that's easily replicatable, right? So some organizations have already come out and you can find this online that an organization like Klarna not too long ago said that they had like 700 employees in a contact center, which is now by autonomous customer experience agent.

7:47You know, is it one or is it many or is it just one replicated? It's one replicated, you know, multi, you know, working in parallel. well um but yeah that's that that really is at the heart of this right because you said you know once you have one really good version of a customer experience agent let's say that that can access uh you know a person's data privately and you know interact with them in their natural voice or you know in multimodal meaning that that you can go through text or you can go through voice and and from that way it's it's going to be a very uh kind of familiar experience for someone that's uh that's you know reaching out to a contact center in in today's world and this is these are scaling up in 2024 so it's happening now and i'll just sort of share why don't i just um kind of lay out some things that are happening in the world to make this all possible right now because yeah please do we're talking um you know earlier there there's there's like big trends and uh you know factors kind of all converging right now and first of all that the the gen ai kind of um requirement is to have good infrastructure right so that the cost of gqs and that infrastructure is going is rapidly decreasing so that's one either the models we're using are getting more efficient uh you know week every month you've got some new models coming out and so you know you got open source models you've got closed source models those that's a nice combo when you get the uh reducing cost and the improved ability there right absolutely so you know they're they're they're also getting cheaper to run and they're getting better so so there's like that's two and three and now the quality of the call them the ai agent frameworks even because there's their frameworks for example um you know framework like crew ai that allows you to basically set up a number of agents that can work sequentially autonomously and you're giving them programmatically you're you're setting them up to do work for you those frameworks weren't around uh last year they are now yeah and then the kind of final unlock here so you've got a couple of other unlocks that you're making uh all this unstructured data available to the agents to access so it used to be when when i was a you know executive at a large financial institution everything had to be structured data so we couldn't really use it if it was unstructured now that we can we can tap into unstructured data and give it give the give this new software these new agents access to it and finally there's so much going on in between organizations through like apis and and other api services that these you know one agent of one organization is going to be speaking to the you know purchasing agent is going to be speaking to the you know to the uh you know the customer or taking agent at another organization for supply chain.

10:59So it'll be it'll be all these factors kind of converging. And they are this year to, you know, to accelerate the kind of the the era of the agent, which which kind of you might but you might ask. So what happens with with knowledge workers in effect in these organizations, I think is going to be there's going to be tremendous change. Yeah. And I think just a shout out for another episode, if anybody's interested on the AI orchestration topic, we talked with Alexander DeRitter, the CTO at SmithOS. But to your point, these new platforms are arising to help orchestrate the AI and the agents. But yeah, let's transition there because one thing that really struck me in preparing for our chat and whatnot is your idea of the agentic organization.

11:52but in principle, like organizational change is extremely hard, right? It's not an easy thing to do. Just any type of change to human habit, especially organizations, which are huge groups of humans is difficult. Uh, what, what does this change look like? I think one thing that stuck out to me when we chatted, you mentioned that we don't have like the soak time that we used to have in past transformation. So A, what does it look like? What's going on? And then how should organizations be approaching this kind of organizational shift in transformation that's going on? Yeah. Yeah. This is a really important topic to address.

12:39We have had many different iterations of you know new software new erps coming into organizations even new sas uh platforms that that organizations use and you'd always have enough time to train people up to use it to to transform people's roles um but but there's no doubt that this is happening at a rate that we haven't seen before and so uh what what you can do today and and i will say uh you know why Why I wrote about this even in the book was that I've worked with some great HR leaders in the past to work on change management initiatives when some new technology would change an entire division and they have a new set of processes to work with a new system.

13:32But now this is sort of happening in entire organizations all at once. so it's a technology that is being made available and it's easy to use um in most cases and you know that this there's a you know super simple versions or there's more complicated ones but generally it's happening all over the entire organization and we really need and this is an important point we really need hr leaders to be involved in this uh interesting yeah reason i say that is because well, I say that because I'm actually talking to many HR leaders, as it happens in my network, that aren't really tapped into the kind of accelerated nature of this new technology coming into organizations, let alone their own organization.

14:26And so I feel like someone has to bring them to the table or they have to come to the table really quickly because what's going to happen and uh i mean i i'm i'm not coming in as a doomsayer although i ask a lot of hard questions because you know in sort of halfway through the book yeah the organization's implementing ai agents they're implementing an entire or a lake and and they're saying you know we're we're transforming people's roles and you know half the organization i can't keep up we can't learn it you know mess necessarily fast enough or uh you know it's it's a it's a case of these uh new tools kind of learning faster than than the people can learn and so uh what do you is it possible that we're going to have you know to transform people into kind of higher value you know roles 100 will there will it also mean that you know there's going to be a you know decent amount of displacement i mean And anyone at some AI expert level is raising that as a point.

15:36So, you know, my recommendation is HR needs to help people transition and get new skills, upskill them, number one, and really help them understand, you know, where this new world is headed. I mean, that's my view, that that should be a responsibility of a responsible organization bringing this technology in. That's a really interesting respect. And I think you're probably the first person I've heard mention it to that degree, because people assume, oh, it's technology. It's like, you know, the CTO or CIO or something there. But yeah, it's it. So what does happen to the knowledge worker, in your opinion?

16:17Like, how much displacement is going to be happening here? And it's almost like the dirty little secret of AI in some ways, you know? I think it's out there. So again, the question of, you know, let's set aside, you know, optimist, doomsayer, or however you want to describe some of these people. but you know the they're the people that i listen to as being real ai experts and you know i'm i'm deep in the space but they're saying 50 of knowledge workers are going to hire a hand of either job transformation or displacement within three years that's a little and that's that's a think about anyone that is dealing with anything digital like you know doing research research insights where whereas data scientists um up until up until last year I'd say you know there there was a long run for data scientists uh role or job and now I have friends that are data scientists that are having you know some difficulty finding roles because organizations are saying oh well I'm using this check GPT thing I'm giving it all my all our data I mean I'm oversimplifying of course because there's more you know private versions out there uh and and Give it all your data, and it tells you what all the insights are.

17:43It tells you what you should be concerned about. It does the charts and trends and all the data science stuff. Another case in point is a senior person at JPMorgan, not too long ago, at a Harvard Business School conference online, and I caught this because she said we used to have people come in, and we required them to learn Python. And so that was a skill that you had to have. Coding was an important skill, getting to STEM. Now, now she says, not needed. We don't need you to learn that because, you know, we're going to have this new AI software able to code anything we need sort of on the fly.

18:29And same thing with Jensen, Juan over at NVIDIA recently. And don't bother learning to code because we got that covered. Our GPUs are so good that they're now enabling these AI models to be absolute super coders. It's an interesting perspective there. And just another shout out to an episode. We have one with our CEO and executive chairman talking about how it's changing the nature of software development. because you bring up an interesting point. Like, what is code? It's a language, right? So can it not just be AI just be writing the code? I think one thing that always comes to mind when this topic comes up, and Alan and I actually are connected through Reuven Cohen's community he's building, which is amazing if anybody's interested.

19:24But I look at somebody like Reuven and just how he understands the foundations of software development and architecture and things like that, that I still think is a very critical skill versus, you know, maybe you wonder how much code he's actually writing. It's more he knows how to connect things. He knows the foundational principles. And I wonder in the future if AI can just write the code for us and do all these things, does the next generation kind of like miss that foundational piece that you kind of the hardening and learning over time of like architecture, or is it just a different path of how we should be educating and learning as humans, uh, to be able to better direct AI?

20:14Like I kind of have that, that interplay in my head of the two scenarios there. Yeah, that makes a lot of sense. Uh, I think they're both right. Uh, you're going to have to definitely continue to understand how to architect it, how to pull it together. Although Ruben, I think himself, I'll have to go back when he said this recently, that even his role, the kind of role that's playing right now might not be around in a few years. Look, I coded as a kid. I've been very techie as I've implemented large scale systems and worked with startups and all in the the SaaS space and so I understand the the coding side of things and the architecture side of things but look now now you've got real examples of kind of not needing not needing to know that depth and you're right then it's going to do will we need it or will we or it's just a new world where you don't need to know that anymore but yeah you just say a few words and and the architecture will be built up again, I think it's, uh, suffice to say that, you know, there's, there's a lot of unknowns right now and it's, I will say that it's easy to, to, to get caught up in, oh, everything's going to change overnight.

21:39I, you know, I think we have a couple of few years, but, but that's not like it used to be where we had we had you know many many years to to re-educate people and to decide you know what they needed I've got um yeah one I think I've already kind of alluded to the fact that I've got I got a lot of computer science uh young folks coming up to me to say what do I do and I said well I can't what advice are you giving them uh when they do come up to you the right advice anymore I mean, you have to understand how to manage these AIs and to get the most out of them, no doubt. No doubt. And I'm going to say that one of the topics I've discussed with a senior person lately is that leadership can't be automated.

22:35And I'd say can't yet be automated. Yeah. Yeah. Right. Right. Because, you know, these, you know, yeah, leadership, I think management is getting to the point of being automated because I can, I can define someone that's a manager as someone who can define roles, goals, tasks, skills that this, these new AI agents will have in the organization. and then have many of them kind of working for you, if you will, and manage them. But on the other hand, the frameworks are such today, and I mean, this, not to get too techy, but I mean, the frameworks are there. No, go for it, yeah. You can literally say to one agent that has a QA role, if you've received an output from this, you know developer agent that uh that this created the output and you've now reviewed it you can send it back to previous you know of worker agent software call it what you will but we don't have to deterministically describe that in software we can just say you can delegate back And the framework will say, okay, so here's enough of a prompt.

24:00The prompts are generated so that this particular agent, again, to use the word excessively, but this particular agent will then say, okay, I've now analyzed it and I'm sending it back because of these issues. So now you've got multi-agents working together. This is where we are today with 2024 frameworks. Really, we didn't have this in 2023. We do now. And I think the systems are only going to evolve. I mean, we will look back at today and think of how rudimentary things are from where it will progress. You can equate it back to early days of the computer to the internet to mobile. to keep going on, right?

24:53But it's interesting. You mentioned the Reuven comment where he's kind of putting himself out of a job too. Like, even if there is that person that understands the architecture, how things work, they can do a lot more with a lot fewer people once they start employing these agents. And we've been playing around with it at Hatchworks as well. And it's really interesting, especially, I think a lot of the QA use cases, Those are like, you know, easy ones, unit testing, things like that, that you can kind of hit early on. But what about, you know, back to the organizational change component? What's the initial, how should organizations start?

25:34Is it just with the education of the people? Is there, what elements, like where's the starting point? Because it's just so much to figure out, right? Here's what I sort of embodied in the story that is the book. And I'm not trying to plug it, but that was where my thoughts went. No, please do. Honestly, I effectively have the HR leader saying, this is your opportunity. So this is what the technology is. We're going to expose this to you. And now you have the opportunity to come to us and say, in your role, how do you envision AI kind of helping you in your role today? So make the existing people in their roles.

26:24And this is actually proven territory where we would bring in new technology previously. And as I said, there was more time involved, but it was the same sort of process where, okay, now we're going to re-imagine these new processes with this new technology. So I've kind of set out the scenario where an entire division or let's start with the division is is asked to uh come to the table and reimagine their their roles um with uh with AI empowered processes and you know I what I will say is that the hard question then next arises to hold on a second, are we now talking ourselves out of our own roles?

27:18Yeah. And it's like either you want to be part of that change and this is your opportunity, the finance people, the marketing people, the salespeople, the logistics people, these are new tools. They are coming here. They are they they they look like this. They can behave like this. How do you want to use them? And I would engage with, you know, with the organized, you know, with with people at every level. And you can do that through the existing, you know, call it the existing organizational structure. um and and look uh i think that i'm kind of alluding to the fact that it's difficult because we've gone through a previous version of kind of outsourcing right um in in the past and it was outsourcing to call centers like in in uh lower cost countries and saw north american um folks kind of training up their replacements that were going to take over their roles because they were lower cost in the lower jurisdiction, right?

28:27It's not that dissimilar to think, you know, this is now, you know, a capable piece of software that can do your job. And we know that it's coming. Why don't you want, like, here's your opportunity to get ahead of it. So, so this is going to be a very intricate balance for an HR person to say, you know, here it is, educate yourself, come to the table and start using this because those that, those that are using it will make our organization more competitive. and therefore, you know, we'll have a better chance of survival as an organization versus those organizations that are really, you know, using the technology and, you know, for both cost efficiency and, you know, better quality, you know, it's a competitive dynamic.

29:20And so it's all of a sudden kind of brought, you know, organizational competitiveness to the to the fore and you know i think it's uh i think it's going to be the case so very much that you know here's your opportunity um i think that's more call it even responsible and humane because i i envisioned you know those conversations that that are in the book that say um i'm just not ready to do this and you know them you know i'm i'm self-selecting i'm i'm going to move on from this organization because it's like the change is too fast or, you know, I'm just not able to keep up. And so some people will kind of natural, there'll be natural kind of attrition.

30:05But I think make no mistake about it, that somebody using AI is going to be more effective and more valuable than somebody not using AI today. So I'm jiving with the HR role responsibility and driving that. But what about and kind of empowering the user to do that? But two things. A, HR isn't always the most technical in nature. And I guess, to your point, it's more about empowering the people within their domain to use AI to apply it. But what responsibility is it of the organization to either give people a starting point, whether it's access to tools or some amount of training? How do you see that playing out?

30:52Because I've seen in some organizations, there's like, you know, it's like cold starts where people try to do it and then they hit roadblocks. And, you know, any thoughts on that? The scenarios that organizations made it? I thought you were, let me see if I can play that back because I, we're sort of asking whether you know what are the responsibilities of the organization and in fact think about an organization that's going to that's going to uh treat this uh transfer role transformation you know very humanely and and possibly and you know doing what it can to to upskill as many of their existing people as they can um and and you know treat everybody respectfully and you know you're you know you're you're you're in this organization because you know we believe that you've got something of value to you know to deliver uh versus the organization that's going to be seen as uh hey sorry we've got a new form of labor uh there's the door right yeah I'm going to say that I would certainly buy from and work with the company and the organization that has treated this kind of whole transformation more humanely.

32:19I don't think we've had to have this conversation before because of the speed that I believe and others believe that we're seeing the impact going to have. Yeah, that's really interesting. I guess, to your point, organizations could just be looking out for the mechanisms and the outcomes that they're focused on, which is shareholder value or whatever it may be. And that may not, the human component may not just be part of that equation. That's, yeah, that's really interesting. This is the point you're, you're, you're bang on because I'm sure you and I, I'm just going to make an assumption, but I mean, I went, you and I went to our MBA programs or I, I know I did in a time that I, in the nineties, it was read is good.

33:09A shareholder value is all we're trying to drive for. But then I was really fortunate to have worked with some top organizational design change management folks. So when I grew some sort of larger changes and, and, you know, process driven change of technology driven and then process changes in organizations. And, you know, I was on the consulting end many times. And then, and then as an executive, I, you know, I led some of that transformation too. And, you know, you can see the difference, or at least I've always felt that, you know, these sort of changes need to be treated with, with utmost respect.

33:51Cause I've seen, I've seen the other side and I don't like it. Yeah. Yeah. It all comes down to the, the incentives, the incentive structure, I feel like in a lot of ways, because people may say they want to approach something one way, but if the incentives don't marry with it, it's kind of a, it's like a, you know, heart transplant type of thing. And maybe the body doesn't accept it, you know. So now you come back around to the agentic organization, right? How much agentic, an agent, where are you implementing and how quickly in your organization? it's going to happen quickly because of the competitive forces at play yeah yeah i mean i don't think there's any other answer other than that and then it's right i mean and then it's how do we yeah how do we treat the the the human beings that that are that are a more i'm going to say more expensive form of labor knowledge workers okay because that's really that's really you know I'd say who we're speaking about anyone kind of moving anything digital around financial words.

35:06Yeah. Yeah. The knowledge workers never really experiences this type of scenario. And you look at it from a business perspective, you know, in AI that works 24 seven, doesn't need PTO, doesn't cost nearly as much. I mean, so what happens to humans? Let's say this plays out and the AI, generative AI is as productive, useful, integrated into everything we do as some people say. What happens to humans in your opinion? Well, it is really, it's a very serious matter. And I just, I chuckled because I just heard Elon Musk speak about this, because it's a really laughing matter. But he said, we're moving into a world where work is going to be optional.

35:58I heard that, yeah. Yeah, right. But it was sort of the other way around. Work is going to be optional because you're not going to be able to do that digital knowledge work as well as an AI can do the work. So somebody will determine that it's optional, whether you determine it or whether someone else determines it for you. But work will be optional. I'm going to say I hope that we find that we're creating new business models. We're thinking in ways to save the world from climate change. We're focusing on health care solutions. ends. And yeah, and not everyone can do that, however. So it's difficult to provide a real answer because when someone of Elon Musk's stature says, you know, work will be optional.

36:57So what's the alternative? Recreation? Well, how are we going to fund that? Right? Yeah, we had a... In the US, there was a presidential candidate that was all about UBI right Andrew Yang you know in the last uh in the last run yeah right there's there's no doubt that that's going to uh that that's going to come and and uh you know come up again that that whole discussion I don't know how it's uh I don't know how it's going to be funded because yeah a little bit above my political pay grade right now but uh it's it's kind of the natural progression of of the discussion, right? Yeah, it'll be interesting to see how it plays out.

37:40The only prediction I can give is that it's going to be messy in the process as we try to make this transition. But to do one shift on topic before we wrap it up, one thing I think is curious, this is not necessarily, I mean, it could be in the organizational change, but I think with generative AI, there's this element of personalization, whether it's with product services, experiences in organization, out of organization, a level that we've never had before. We've always been promised this level of personalization. It never seems to quite get there, but with generative AI, it's like, well, okay, maybe this actually is possible.