Agentic AI: From Hype to Real Work Done

00:00:07 Mark Smith
Welcome to AI Unfiltered, the show that cuts through the hype and brings you the authentic side of artificial intelligence. I'm your host, Mark Smith, and in each episode, I sit down one-on-one with AI innovators and industry leaders from around the world. Together, we explore real-world AI applications, share practical insights, and discuss how businesses are implementing responsible, ethical, and trustworthy AI. Let's dive into the conversation and see how AI can transform your business today. Welcome back to the AI Unfiltered Show. Today's conversation features a guest joining me from Maryland in the United States. All the links for what we discuss will be in the show notes for this episode. With that, let's get underway. Daniel, welcome to the show.

00:00:57 Daniel Cohen-Dumani
Thank you, Mark. Very nice to see you today.

00:01:00 Mark Smith
You too. Good to have you on and good to have this discussion. I think, as I just said, you know, joining the call, I feel in the first six weeks, to 8 weeks of 2026, there's been a quantum shift in what's going on in AI, particularly around agents.

00:01:19 Daniel Cohen-Dumani
Indeed, it did. And frankly, Mark, this is something we're all going to have to get used to.

00:01:25 Mark Smith
Yeah.

00:01:25 Daniel Cohen-Dumani
You know, there's not going to be a slowdown. I think this is just a new normal. News every day. And you know, as businesses, what's important is to filter what's important for your business. and try to make sense of the news of the day.

00:01:44 Mark Smith
Before we kick off with that discussion, tell me a bit about food, family, and fun. What do they mean for you and your part of the world?

00:01:51 Daniel Cohen-Dumani
Yeah, that's a great question. I'm a big fan of food. By the way, I grew up in Switzerland. I've been in the United States for almost 28 years now. So food is dear to my heart. I love to cook. I love to cook for my family. I'm married, I have four kids. Older kids, three out of the house, one still in college. And outside of work, I just enjoy walking my dog, playing some golf, and just having fun with friends.

00:02:23 Mark Smith
Yeah, nice, As we kick off 2026, what is top of mind for you? What are you seeing? from your lens on the world? What are the, what's happening with your customers? What are you hearing out of the market at the moment?

00:02:38 Daniel Cohen-Dumani
You know, it's interesting. There's a sense of complete chaos in one hand, uncertainty, and at the same time, excitement. And we can see just the last 8 to 10, 12 weeks, it's just been a whirlwind of big global news. big shift in the marketplace, excitement about AI. a few months ago, we were all talking about an AI bubble, right? That was just the only world out there. And I had shifted to maybe there isn't an AI bubble, right? Maybe that wasn't just that true. We're now seeing with the release of OpenCraw and those agentic, you know, people have gotten excited to understand what it can do for me as an individual, and then what it can do as a business. But then you see report like last 10 days of, this Citrini research report that took the market down last Monday was a great example of complete fear of AI or instilling fear in businesses of the sky is going to fall soon. and sooner than we think. So what I'm seeing is a sense of tremendous excitement the next day, complete fear and desperation and chaos. I think business are more lost than ever. They just don't know what to make of today, what's going to be tomorrow. And I think we're seeing that in the adoption of our product experience where there's super excitement yet decision-making has slowed down because of the amount of information that are coming out every day.

00:04:28 Mark Smith
Wow. Okay. That is interesting. You know, for me, the change has been we've definitely moved from a chat-based experience with AI to now, you know, even getting towards a talent last year, I always felt that the agent story was overhyped. Then OpenClaw comes out and all of a sudden, you know, and I've got to quite a substantial deployment of it running. And I am blown away by what I can get done. Why actual work I can execute on it. And it's blowing me away at its capabilities. How do you think this is going to impact, you know, because I feel like the open claw is kind of like, it's great for tech folks to get on. I don't think a layperson, unless they've got very strong will, can get to the point they need to an open claw yet, because man, I've spent weeks on it and I've rebuilt multiple times. And if I didn't have that persistence, I don't know. How do you see it though, translating into real world enterprise application where, you know, agents of that scale are going to be implemented and then what is going to be the impact on people in the business?

00:05:44 Daniel Cohen-Dumani
Yeah, that's a phenomenal question, Mark. you're right, OpenClaw is a very technical play. I mean, it's people like you and I that have time to do those kind of things. But, you heard Microsoft last week released Copilot Task, which is supposedly the equivalent of automating part of your workflow using Copilot. So I think we're going to see more end-user friendly. There's already like at least hundreds of those clothings, toasted, no setup, or human will help you set it up, and it's moving fast. You know, there's one thing important that I'm seeing is a tremendous lag in adoption for most businesses. I feel you and I are in the bubble, and everyone I talk to that are friendly, that are in the tech world, I think we live in a bubble where we see what's happening today. But what I see happening with businesses is they're still two years back. AI is here. I'm trying to figure out how I'm going to use a just simple chatbot on my data for my business. I've tried this, didn't work so well. I dabbled with using this to help me write e-mail. I think this is the maturity of where most businesses are, at least mid-size businesses. by the one that are, a few 100 people to 10,000, 20,000, they're still trying to figure what the plan is. And they're way behind. Yet I think in one sense, we're seeing this dichotomy between the consumer, which is becoming very AI savvy, and the business that are lacking behind. And then we're seeing more and more consumer in the business coming with their tools. I'll give you a simple example. My daughter, my third daughter, who's working in a communication consulting firm, you know, she's bringing her own AI tool to work. She's buying her subscription, doing her own thing, because the business is trying to figure what are we going to do, right? And people are not waiting, right? And I think OpenClaw is a result of that movement of saying, agentic AI is here. Like 2026 is really the year where we're going to see business start knocking off things using agentic AI. Last year was, it's coming, let's have fun and play with it. But I think OpenClaw gave us that, oh my God, we can actually do this. type of moment. So there's a lack of, there's a clear lack of adoption because I feel like business are paralyzed. And they're paralyzed between the noise they hear. There's a bubble. Is it going to burst? Now it's not a bubble. It's agentic AI. We need to move forward. Yet we have a business to run. And I always use that analogy, which is every business today have to be driving a bus on the highway 80 miles an hour, and then changing the four wheels as that bus is running. And we know how hard that is. And I think the analogy is the only way you can do it is you have to slow down. You have to make it so strategic that you're willing to take a few steps and slow down and think about it strategically and say, AI is going to disrupt my business. This is true for every business. Let's think about what we're going to do step by step. And there's a huge, there's a huge disconnect there that I'm seeing. 

00:09:44 Mark Smith
I Don't know if you remember the author from back in the day, back in maybe the 80s, maybe, Ken Blanchard. He wrote a book called, it was called No Can Do, and he talked about a business, about the knowing-doing gap. You know what you need to do, but there's a disconnect, there's a gap between doing and what we're now calling adoption. How do you think organizations get through that knowing-doing gap from, I feel like they're in full paralysis at the moment because some are hearing they need to do something, but they just don't know where to start. Like, I'll give you an example. I was with the CEO, owner of a a big distribution type chain. So he has $8 million of stock on hand at any time within the organization, and he's got a multi-million dollar, $60 million SAP backend as part of it, and Snowflake for the data lake. And he's going, what do I need to do with AI? And you know, from the big question is, could I reduce my stock by $2 million by using AI smartly? And I'm like, mate, these are the questions, right, that owners, you know, that's a massive capital free up, right, in an organization. But it's still like, where do you start on that journey? As in like, this is a company that don't even have a chat experience running in the company. They've got nothing at the moment. Like, But that's the way he's thinking. And I'm like, hey, this is, there's a big gap. And I feel that sometimes the consulting land, and I was just with another company this week in the US actually, got about 1000 staff and everybody that's pitching them is just pitching endless consulting, but kind of like no outcomes. Like, will you hang your hat that you could do, deliver this type of outcome for our business? And I'm like, there's I think even a fear with consulting companies at the moment in the space. I spent three years at IBM up till just recently, and once again, saw that the knowledge inside the organization was not moving as fast as what the market was.

00:11:59 Daniel Cohen-Dumani
Absolutely. My thinking regarding how the business thought has to be with the first recognition. of the fundamental difference of this AI revolution is compared to any other technology revolution. You know, a lot of people say this is the internet all over. Well, you and I know it's times 1000, right? At least and accelerating every day. But this is the first time that we have a technology revolution where we don't know what's going to happen. What is the long-term impact? Nobody has any idea. As a matter of fact, if we're chasing this concept of AGI, of artificial general intelligence, nobody can tell you when it's happening. And the consensus is in six months or in 10 years or never. This is how wide of a consensus we have. Yet we have to make decisions as businesses. So the first thing is to understand, one, AI is going to disrupt your business. That is a fact. So there have to be an acceptance of that fact. The second is, all right, so how is my business strategy going to evolve based on that fact, right? So we know you're going to be disrupted. How are we going to be disrupted? And consultant can come in and help you do that, right? And the outcome has to be, all right, we're going to be disrupted. This is what we think is going to happen. There's a lot of knowledge out there about disruption of AI in any kind of business, right? So business owner can learn from that. Second is, all right, we know there's disruption. Here's our strategy. And from that point on, how do we get there? And you're right. It's not about doing a five-year strategy. It's really doing about six months to a year max and say, what are we going to do in the next six months to move the needle? And in the case of your business order, it could be, yeah, you know, what is your biggest business problem? If it's cash tied in inventory, yeah, that's the place you want to focus. That's your business problem. Focus on that one, right? Maybe your problem isn't your employee experience that takes a long time to find stuff. It is the cash in your $8 million of inventory in stock. And then what's where you identify a business problem to solve? It could be one or two. That's where you can start moving the needle. And that could be bringing consultant in the door to help you do it. Or, you know, what I've been flow over the last three years of what we've seen with AI is back to this idea of consumer coming into businesses is the fact that you and I have been in technology for a long time. The premise of technology adoption would be from the top to the bottom. The CEO make a decision, the VP taking on, and then they say, we're going to build a new ERP system. Hey, there we go. And everyone go forward. I think what's different in this case is innovation has to come from the bottom up. There has to be a strategy that say, we're going to do things, but we have to let the people closer to the problem experiment and be the leader in that transformation. Because otherwise, what we're essentially asking people to do, and if that comes from the top, it's not going to work. Find a way to make yourself irrelevant. That's what AI is all about. So if the CEO say, let's embrace AI, the message for everybody is find a way to make yourself out of a job, which doesn't work so well for most people, except the message has to be find a way to make your job easier. So you're not working 16 hours a day and on weekends where you can do the same work in four hours and spend 4 hours doing very human like type tasks, you know, which you talk talking to your customer, talking to your folks. if you're a manager, coaching your people, spending real time coaching them, things you never have the time to do, right? Because you're doing busy work. So I think bringing the folks in the bottom of the food chain closer to the problem to be the one driving it is critical. Top, bottom, bottom up, both at the same time.

00:16:33 Mark Smith
I like that. I want to discuss with you organizational memory. And I feel one of the things that was big in 2025 is the concept of memory with AI. So in other words, outside of our single session, it remembers a lot more information. We saw the major providers in the market provide a settings area where you could put in things about how you wanted it to respond, things that you wanted to know about you. And then over time, you could start building it. Organizational memory is that at a massive scale though, right? Because it's all parts of the organization. How do you talk about organizational memory and how do you implement it?

00:17:16 Daniel Cohen-Dumani
Yeah, great, great question. And that's part of my new passion for the last two years has been thinking about, you know, with AI coming, and I think we're seeing with Agentic AI more than ever. Context is the key to having reliable agents. And without context, an agent is nothing. And context is memory. And for an organization, context is your organizational memory. In a way, it is a memory that is tuned to your organization and is specific to your domain of expertise. So I'll give you an example. If you're a manufacturing organization, you are manufacturing products, you have clients, you have distributors. So creating an institutional memory that understands the concepts that humans understand, not just tokens and tokens after tokens, is critical to enhance the quality and the accuracy of authentic AI. So an organizational memory is a think about a database of all of your organizational knowledge. digital knowledge or even of tribal or non-tribal knowledge put into a database that is structured in a way that an agent can reason against it. So how do you create such an institutional memory? It starts by understanding where is your knowledge stored? And if it's not stored, how do we get it stored? one of the big advances that I'm really the most excited about over the last two years have been the ability to stop taking notes. As much as I love it, I hate it. I've always been someone that learned by writing. But when you write during a meeting, you don't pay attention, right? Because you're taking notes. And now I find myself being able to truly have a meeting with full attention because I have my note taker taking care of that. That is a way to capture knowledge, by the way. And it's the way that organization is to think about if you know someone is retiring or going to leave your organization, put them in front of a computer with someone asking them question and just record it. Don't ask someone to document their work. Just talk to them, right? And that's where a human-to-human connection plays well. And then the AI will take care of structuring those notes and doing what you need. So the first thing is capturing knowledge, right? So if you're a knowledge intensive organization like consulting or professional services, that's part of your DNA. You are capturing knowledge. That's what you do, right? But those organizations do a really poor job at organizing that knowledge or reusing it. So an organizational memory is just nothing else than a system that goes out and pull knowledge where it is. your document management system, SharePoint, Google Drive, whatever, your CIM and ERP and system of record, and then all those other application between like meeting recording or wikis, web application, project management system, task management system. You pull all of that, you extract that information, you organize it, you classify it, you enrich it with AI, and you store it into a digital brain of your organization. How we do it, we use knowledge graph technology to do that. is what has emerged as the next big thing to give context to an LLM, aka an agent. So a knowledge graph database, but at the minimum, a strong structure of representation of your knowledge and memory gives that to other people. So it is not an easy task. because you have to, every organization has so much content that nobody knows what's good and what's not. So you have to be a little careful. And of course, AI is doing a great job at helping us sort for all that maze of millions of stuff everywhere.

00:21:35 Mark Smith
It's interesting because as you're talking, I can see people in, you know, Microsoft land, which I spend a lot of my time in talking, going, oh, we just pointed at SharePoint. And because that's where all the documents in OneDrive, that's where it'll be. But you're talking about, what I understand you're talking about is quite different. It's actually extracting out almost a librarian type function. It's almost like I can see new, like some new AI job roles for people here to actually be the curator. They might use agents as their, arms and legs as part of this, but actually curate a central intelligence system of the business that is not your day-to-day working files of here to there, but actually a corpus of information about, it could cover everything by around your culture. how you operate in regards to your customers. So a lot of the stuff that working in a business, the stuff that's not in a job description or in the templated document, you're saying moving that into a central system so that agents act with intent based on that core second brain or brain of your business, memory of the business.

00:22:52 Daniel Cohen-Dumani
Yeah. And, you're seeing new world coming out such as system of intelligence, right? Which is this common representation of knowledge in your organization or institutional or organizational memory. Those are synonyms.  I read a very interesting article last week about, you know, there's been all those talk about SaaS is going away. People don't need those systems anymore. And it's true to a certain degree, because when you think about a CRM system, a CRM system is nothing else than the state representation of where you are in conversation with a client. And guess what? All of those piece of data is stored somewhere else already. So why in the world are we asking sales representative to spend their day going updating a system where they've already had the meeting? that may have been recorded. They may have received a contract or sent something to the client, which give you the stage, right? And can a system with intelligence replace a CRM? Because at the end, all you want is track where you are.

00:24:06 Mark Smith
Yeah.

00:24:06 Daniel Cohen-Dumani
And if a system can just see, I'm reading your e-mail, I'm seeing who you're talking to, I've seen the recording of meetings you have, I've seen what document you send, I know that this opportunity is at 90% of the cycle. Don't need to update my CRM anymore. I have a system telling me that. And I think what you're seeing with AI and those system intelligent and agentic AI is saying, maybe we don't need as many of those system of record that we used to have and have people manually updating this one and then integrated this one with this one or manually updating that one, which is what, hey, guess what? Many organizations, they still do that. Doing this system, going that system and so forth, right?

00:24:50 Mark Smith
I smile because I've spent 22 years of my career selling Dynamic CRM from day one. And I was in touch with an old colleague friend of mine in Brazil last week. And I said, I don't think we need a traditional like that product. And he goes, what do you mean? I said, I don't need a CRM to have a head or have an interface. As far as I'm concerned, it can be headless. As long as the data is captured and And my next step, I can bring it up in whatever interface I'm using when I need it. Why do I need to make sure that the first name and last name pair fields are filled in? Like that I've gone and typed it in? Or, you know, or that did I format the phone number correctly because the CRM required it to be formatted a certain way? It's just like, that's all relevant now because I know through the agent systems I have, I could tell it to build a CRM based on the operating model of an organization. And it won't be full of all the redundant stuff, because one of my things with Dynamic CRM is that it is using a metaphor of 25 years ago, you know, of how it represents data. It's like CRMs haven't really evolved. that heck of a much in the last 25 years. And they're full of legacy of how things used to be done. And it's all weight. And I'm just like, I know, because in the last week I prompted one into being and doesn't have a head and it works very efficiently and it works. And I'm just, yeah, I think the whole SaaS thing does change a lot. I mean, obviously easier for an SMB, SMC, or mid-market type business, harder for an enterprise, but I see it going more and more that way, as in that we don't need to have a keyboard screen interface forms over data type of experience anymore.

00:26:59 Daniel Cohen-Dumani
Right. And I think the first thing we're going to see is agents filling your CRM system. That's sort of step one. That's the easier one.

00:27:08 Mark Smith
It's already happening, yeah.

00:27:10 Daniel Cohen-Dumani
But a few years down the road, you're going to say, do I really need that system? Can I store it anyway? I'm just storing data, right? And can the agent tell me who I am? Because the goal of those system of record is to give you a current state of where you are in sales, in finance, in whatever, right? In selling your projects. But if AI can figure this out on its own for agentic modes, then you don't need those system of record. They'll be maintained within the agentic world. You're still going to need an organizational memory of sort, right? You're still going to need those, right? But there's going to be less and less need to go enter data in systems left and right. That's going to change. But that, I mean, you and I think, you know, there's always the optimist that say it's coming soon. This is, we're still a few years away from that. Massive shift, for sure.

00:28:04 Mark Smith
Yeah. Oh, totally. That's what I've got to keep reminding myself. I'm where I am because of my proactive engagement. That's not anywhere near the marketers. I was doing a presentation in January to a user group in Australia, and the adoption rate is around 3% in corporate world of AI. And it's after three years, right? So there's a big disconnect.

00:28:29 Daniel Cohen-Dumani
And we're talking just, okay, let's use, let's have a chatbot to make, to do my job on public data, right? Not even on my private data, just a corporate sanction chatbot. That's where people are still debating now.

00:28:46 Mark Smith
Yeah. Oh, and this is why it blows my mind where I see CEOs of big LLM companies saying white collar work will be gone within 18 months. I'm like, Hang on a second. Do you realize the people that are running or switching on or off these systems are not switching them on at the moment? As we close, talk to me about culture and talent in the context of AI and where we are in the market and where we move forward.

00:29:17 Daniel Cohen-Dumani
It's a great question, Mark. I think there is quite a bit of discussion about how people are going to learn. I was actually having coffee this morning with a friend of mine. And everyone comes to me and say, I'm speaking a lot about AI. They always ask me questions and how do I do it for myself? And what I've been telling people lately is use AI to learn AI, but use AI to learn. Don't use AI to shortcut yourself. Use AI to learn how to do things. And then when you learn it, you understand how it works and now you can use AI to actually do the work. But there's nothing better than saying, how do I set up an agent for me and use AI to do it? Don't be afraid. You have this amazing agent in front of you that can reason with you, that can coach you, right? And if you're being strong enough to use it that way, I think you can be successful and learn. The risk we're going to see is, you know, what we've seen a lot happen over the last two years is AI has become this equalizer. You know, it brings the lower performer up, right? Because now they're as good as everybody else because they can use AI. But the people that are taking shortcuts, and we're going to see that in education too, is If you take a shortcut and say, I'm going to use it to solve my problem, but I'm not going to learn how to do it, is when you stop learning, right? And that is the biggest risk I see from an education, from a mentoring, from a coaching standpoint, is just using AI to shortcut. And that's how I do it myself. When I'm doing something I don't know, I'm going to say, tell me how we could do it. Right, I go to Cloud Code and say, how can we implement this? But let's think about it. And I'm first going through a planning motion of saying, let's brainstorm ideas. Give me three ideas, give me three ways to do it, and then let's talk it through. And that is going to be critical in the way we approach education, we approach learning and teaching in the workplace is using those to your advantage?

00:31:42 Mark Smith
Daniel, thank you so much for coming on the show. It's been very insightful talking to you. I really appreciate your time.

00:31:48 Daniel Cohen-Dumani
My pleasure, Mark. Thanks for having me.

00:31:52 Mark Smith
You've been listening to AI Unfiltered with me, Mark Smith. If you enjoyed this episode and want to share a little kindness, please leave a review. To learn more or connect with today's guest, check out the show notes. Thank you for tuning in. I'll see you next time, where we'll continue to uncover AI's true potential, one conversation at a time.