Beyond Chat GPT: How Agentic AI Will Transform Business and Work

The technological landscape is shifting dramatically beneath our feet, and agentic AI stands at the forefront of this transformation. Unlike today's familiar AI tools that merely generate content or provide information, agentic AI operates with genuine autonomy making decisions, taking actions, and learning continuously with minimal human oversight.

TLDR:

• AI agents are applications in an AI-enabled world that perform specific tasks well while continuously learning
• Agentic AI works best in teams or multi-agent systems where each agent handles distinct aspects of a problem
• Now is critical for implementation with LLMs providing cognitive power, increasing workforce AI literacy, and competitive advantage possibilities
• Organizations should start with high-value, low-risk use cases rather than delaying implementation
• Real-world examples include processing thousands of KYC documents in minutes with higher accuracy than human teams
• Future of work will see humans focusing on relationships and creativity while autonomous agents handle mundane, robotic tasks
• Large organizations maintain competitive advantage through existing relationships if they free employees from routine work
• Business leaders should identify what their most talented people should be doing versus what they're actually spending time on

As Chief Revenue Officer of Aibly  and a recognized authority on agentic AI, Jim Marshall offers a refreshingly clear perspective that cuts through the hype surrounding this technology. 

"An AI agent is simply an application in an AI-enabled world," Jim explains, highlighting that these systems truly excel when deployed as teams tackling complex problems together. 

Each agent specializes in specific tasks, processing information with remarkable efficiency before passing results to the next in line creating workflows that outperform traditional approaches in both speed and accuracy.

The timing couldn't be more crucial for business leaders to engage with this technology. With sophisticated language models providing unprecedented cognitive capabilities, cloud computing delivering the necessary processing power, and a workforce increasingly comfortable with AI tools, organizations face a critical decision point. 

Those who embrace agentic AI now stand to gain substantial competitive advantages, while those who hesitate risk falling behind rapidly. 

As Jim Marshall aptly puts it: "The best time to start was six months ago. The second-best time is now." 

Whether you're a C-suite executive evaluating technological investments or a professional curious about the future of work, this episode provides essential insights into how agentic AI will reshape business operations, customer relationships, and workplace dynamics in the months and years ahead. 

Listen now to understand not just what's possible, but what's already happening in this fast-evolving space and discover how your organization can benefit from the agentic AI revolution.

Chapters

• 0:00 Defining Agentic AI
• 2:55 Why Business Leaders Must Pay Attention
• 4:58 Future Predictions for Agentic AI
• 7:04 How Agentic AI Actually Works
• 10:42 Implementation Strategies for Companies
• 12:27 Real-World Applications and Examples
• 16:02 Mental Models for AI Success

Transcript

Kieran Gilmurray: 0:00

AI doesn't just boost productivity it redefines value. Agentic AI isn't just smart it acts with autonomy and purpose. This technology is poised to transform how businesses grow, compete and innovate. If you're in the C-suite, now is the time to pay attention. Hi, my name is Ciarán. I'm a globally recognized authority on AI, automation and digital transformation. My guest today is Jim Marshall, a consultant, business leader and entrepreneur. He has spent 20 years building, buying, scaling and selling businesses. Jim is the Chief Revenue Officer of Aibly and has helped shape the next generation of AI-augmented workspaces, where people and agents don't just coexist, they collaborate. He is currently one of the leading voices on agentic AI. Today, we're going to dive deep into what that means and why it matters. Welcome, Jim. Hi Kieran. How are you?

Jim Marshall: 0:55

Very good, very good.

Kieran Gilmurray: 0:57

Look, I'm really looking forward to this, because this is the hottest topic that has been around for quite some time, so let's dive in and get some answers from an obvious expert, jim. What is agentic AI and what is it not?

Jim Marshall: 1:13

So much hype around at the moment. So what is agentic AI? I'll start with what an agent is, because I end up in lots of conversations where that first step isn't really understood. And, in my simple terms, an AI agent is simply an application in an AI-enabled world. Now that capability of that agent stretches all the way through some really unhelpful analogies, like an AI agent being a digital human. I think that rhetoric is not helpful. But AI agents, those applications in an AI enabled world, really come into their own when they're when they're deployed in teams.

Jim Marshall: 1:51

So you might hear that referred to as multi agent systems or agentic workflows. But the beauty of agentic AI is that each agent is performing a very specific task well and learning from it. So we think about Gen AI being very good at creating content, whether that's text or images or even code. Agentic AI is different in that it operates more independently, is able to make decisions and act in real time with minimal human intervention. What it is not is a digital human. There are things that a gentic AI is very, very good at and better than humans at. There are an awful lot of things that humans are an awful lot better at than ag entic AI awful lot of things that humans are an awful lot better at than agentic ai.

Kieran Gilmurray: 2:48

Wonderful, wonderful, wonderful. Well, why is now the time that business and leaders need to pay attention to agentic ai?

Jim Marshall: 2:55

oh, good question. Um, now, now is the time. Uh, it it's. I was fascinated to find out that the can opener was invented about 40 years after the tin can, and I imagine in my mind I have all of these unopened tin cans and I think that now is the time because we, with the power of large language models in a gentic AI, it feels like we've been given the can opener models in a gentic ai. It feels like we've been given the can opener, we've been given the doorway to really access that, uh, huge, um, an ability to get to tap into that cognitive power.

Jim Marshall: 3:32

So, so, one, what's available to us? Um, secondly, our, our people are, and I think there's two fronts to this. One is our people are already playing with this stuff. So as organizations, we need to be ahead of the curve. And the second thing is, as we talk about a generation past that were digital first not you and I, I fear digital learnt we're now having a generation that come into the workforce that are going to increasingly be AI first. So we have that huge power of LLMs and the cognitive power at our fingertips. We have people that are increasingly embracing and wanting to leverage that power. And then, from an organizational perspective, strategically, we have the ability to drive really significant competitive advantage. So most organizations, I would suggest, have a co-pilot or using GPTs, private GPTs or chatbots. Some organizations are relying or increasingly using AI for decision support, but the real winners are going to be those that embrace agentic AI. So now is the time, or maybe the second best time, because the best time was about six months ago.

Kieran Gilmurray: 4:52

So where is agentic AI headed, if you can predict it over the next 12 to 24 months?

Jim Marshall: 4:58

OK, but I have a really lousy history of predicting the future, so bear that in mind. It's a fool's game, obviously. I liked and we've spoken about this, kieran, I liked in your book where the levels of agentic AI maturity were laid out in levels. So I think we're increasingly seeing organizations who are embracing agentic AI moving from sort of level one, perhaps using some fairly simple augmentation to summarizer agents, for example, being able to process large amounts of complex information and provide analysts with discrete, high level information, through to perhaps more self-guided execution and reflection, so research agents looking at prices across various markets and then being able to execute dynamic recommendations. Where else do I think it'll go?

Jim Marshall: 5:56

We're seeing two types of customer broadly, two types of customer at the moment One which is going to be a customer that's going to be a customer that's going to be a customer that's going to be I've seen what you've done with agentic AI and translations. Can I have one of those please? So, so, a kind of not deliberately, but a bit of a lazy purchase. And then the other, perhaps more forward thinking approach has been I've got this problem or this opportunity. Do you think it could be solved or partly solved by agentic AI? By agentic AI. I think what we'll see in the next 12 to 24 months is a much broader recognition of the power of agentic AI. If I deploy multiple teams and have the ability of those teams to speak to each other and process large amounts of information in real time, then we'll start to really unlift the bonnet on how we can change our organizations using agentic AI.

Kieran Gilmurray: 6:52

So how does agentic AI actually work then? How is it different to chat GBT, co-pilots, gpts, and what does that actually mean for the future of work?

Jim Marshall: 7:04

Agentic AI is built on successes in other areas of technology. So our compute power, brought about by really rapid developments in cloud. We've already spoken about the plethora now of large language models at our disposal. This increased sophistication and availability of apis. So, um think, I like to think of the, the large language models, the llms, um serving as the cognitive engine. Be careful to be drawing on those human analogies. So the, the brain of the, the agentic workflows, um and and so we can deal with vast data sets. We can um contextualize information easily. We can reason and predict with increasing levels of sophistication.

Jim Marshall: 7:55

Agents then have those tools at their disposal. So API calls pointed research at the internet or file structures, and then a multi-agent team can use those tools for very specific tasks and then pass that task on to the next agent. So, for example, an agent might be specifically looking at reading a complex bid, whilst the second agent might be looking at the existing file structure of previous bid answers and matching those to the bid questions, bid answers and matching those to the bid questions. A third agent may be looking at internet or other file structure research to supplement answers where it feels less confident, and then we might be able to give the human in the loop a degree of confidence in the answers and loop back and learn. So that ability to set up or to solve complex problems or create opportunity by having agents working together is is of enormous value.

Jim Marshall: 8:53

Um, what does it mean for the future of work? I have to be really careful about not going into a kind of matrix type answer on this and and and and calling doom. I mean, there there are some, there are rightly some fears. Um, what I? What I think, how I think about this in the context of the future of work, is, if we wind back to that industrial revolution and we look at how many people today would volunteer to dig a field, I think you'd have very few sticking their hands up. We regard that as a job for a machine and those jobs for machines meant that people went to work in higher value jobs, perhaps in banks and insurance companies, and look back to that old footage with the pride as they set up at their desk with their quill pen and wrote a ledger. And you know this is a high value task.

Jim Marshall: 9:43

And now you would probably describe writing a ledger as quite a robotic task. Would probably describe writing a ledger as quite a robotic task, um, so so, so a lot of. Although we're working with a high degree of sophistication, a lot of the activities that we're replacing are quite mundane, robotic, um, pieces of work. So, in terms of the future of work, I think, I think and hope we can augment our very talented people with a genetic AI to allow humans to be more human in the workplace again, to focus on relationships, to focus on original thoughts, to focus on being truly creative. Yeah, there are obviously going to be some significant changes to workforces off the back of this, though, and that it won't it won't come without some, some negative connotations so what?

Kieran Gilmurray: 10:35

what does a board or ceo need to do to start experimenting or building with the gentic ai?

Jim Marshall: 10:42

it's going to depend on the, the level of maturity of the organization. For, for those that are still sitting there thinking, I know, I know I need to do something, but I don't know what it is um, I would, I would encourage them to get involved in um, in, in in groups and and talk shops ably. We, we run a program called agent thinking and that really breaking down the some of the barriers, some of the things we've talked about today. What's, what's the nuts and bolts of agentic ai, how do we allay fears? And then a bit of a round table on where use cases in the organizations for opportunities and problems might exist. There is a degree of pace that's required.

Jim Marshall: 11:23

We've talked about competitive advantage and the need to get on the wagon, but I'd also encourage CEOs and boards to look at small steps being better than no steps and small steps being better than big steps that result in decision paralysis. So I would encourage organizations in those in those small steps to look at areas where they can potentially drive high value but low risk use cases, so proof of concepts that don't heavily compromise the existing IT infrastructure and then bringing smart people on the journey. Boards and CEOs that don't think that their people are already operating in the wild west of unsanctioned GPT usage have probably got the blinkers on. I know there's ways of locking these things down, but not embracing it doesn't mean that your people aren't embracing it, so get on top of that pretty quickly.

Kieran Gilmurray: 12:20

Can you share some real-world, cutting-edge examples of agentic AI in action to bring this all to life?

Jim Marshall: 12:27

Yeah, so lots of examples. And one of the difficult things about working with agentic AI is the potential applications are almost limitless across an organization, whether it's in back office operations or in marketing, in new business, in pricing, in procurement, finance, you know, the list goes on. So we're really nervous about presenting use cases because of that, or have one like that and if, before you know it, you've gone down a rabbit hole of being known as the people who do translations with agentic ai or the people that do, uh, um, big documentation, um, I think some of the best examples are some of those quite mundane, robotic tasks that are then replaced with a really clever, quick solution. And I'll give you one that we've, we've, um, we've worked on, which we're really proud of, and it was a regulated organization that was really struggling with, uh, the volume of bank statements that being are being sent in for, know, your customer, and the problems were several fold. One was just the amount of work that had to be done. Secondly, just how boring that work is. And then, thirdly, just the high propensity for error when using humans in mundane tasks because it's boring.

Jim Marshall: 13:42

So we work to create a number of extraction agents and we're able to prove. The level of extraction on the documentation was far surpassed our customers' expectations. So extracting information, for example from photos of screenshots, so your extraction agents then passing that information on to a Know your Customer agent. Your extraction agents then passing that information on to a know your customer agent. That know your customer agent was able to compare information on, for example, a bank statement or a water bill with safe list and unsafe list. That was then able to be passed on to a research agent which was able to look at any areas of the document that had not previously seen or recognized and then could go out to the internet and make some deductions about what they might be and whether they may or may not be safe. And then the fourth agent is a reporter agent presenting back to the human in the loop.

Jim Marshall: 14:37

We believe these are all good records. We believe these few should be looked at and some middle ground where the human in the loop increasingly agrees or disagrees with the agent and pushes that information back to the system of record. So you're looking at processing. In this example four to six thousand documents in a month and those being processed in a matter of minutes with a higher degree of accuracy, a great level of extraction, but also, really importantly and unlike the previous process, an auditable system of record about why those decisions were made and the ability to report off them. So, like I say, quite a mundane piece of in in terms of what, what was happening, but a really good example how, of how each agent had a very specific task and was and was achieving that with a high degree of accuracy and then learning from it. So the process, that continual improvement loop, um uh, yeah, really powerful for the, for the, for this business and at one stage, as you said, that was probably quite a cognitive or intelligent task to do.

Kieran Gilmurray: 15:49

Now we recognize technology can do it and therefore it becomes mundane. What's the one mental model you think leaders need to shift if they want to succeed in the era of agentic AI?

Jim Marshall: 16:02

In terms of the mental model, I would say and this might seem like a strange answer I think I would still start with our people and I would start with what do I want our very best and most talented people to be doing, and then look at what are all the tasks that they are actually doing, which we are, which we are, which which we're less comfortable with, and that would be, that would be, my mental model. I think that sort of leads on to how I, how I would see those businesses transforming. So a lot of commentators, who know a lot more about this than me, are predicting and I think probably rightly that we're going to see some very large businesses emerging that are actually run by relatively few people. I think we're going to see challenger organizations that are AI first having some awesome and rapidly improving tooling and agentic AI to really challenge the market. But I don't think it's all doom and gloom for large existing organizations.

Jim Marshall: 17:17

I think one thing we've talked about where those organizations may have a huge advantage is the relationships they already hold. Advantage is the relationships they already hold. So if I was a large organization, I would be really focusing on customer experience and relationship and setting my people free to focus on those relationships, whether they're supplier or end customer relationships, and I think we talked about 12 to 24 months. I think that's the window to make that happen. I think large organizations that are relying on people to do mundane tasks and are not focused on customer experience are going to struggle in the next 24 months.

Kieran Gilmurray: 17:54

I think they will. I think they probably already are as well. Unless you're using your labor for the right purposes, they won't be there for very long. Jim, this has been a masterclass. Today, we learned that agentic AI is very, very real. It reasons, it acts, it adapts on behalf of your business and its people. The opportunity is enormous, but if you're not on board, then you are in danger of getting run over by this rocket ship. For those watching, jim's latest work on agentic AI is a must follow. I'll put the links in the body of the text and the post and the podcast and the webinar and everything else down below. Thank you everyone for joining us today. Jim, thank you. That has been excellent.

Jim Marshall: 18:33

Lovely to see you, as always, take care. All the best, kieran.