Your Competitors Are Already Using Autonomous AI Agents. Are You?

What happens when AI doesn't just respond to prompts but takes initiative, makes decisions, and collaborates with other AI agents to solve complex business problems? Simon Torrance, a global authority on business transformation with over three decades of experience, opens our eyes to the revolutionary world of agentic AI – and why it matters right now.

TLDR:

• AI with agency can make independent choices, take actions, and solve complex problems
• Agentic AI differs from automation by creating end-to-end solutions rather than just following processes
• Five key elements of an agent: LLM capability, persona, reasoning ability, memory, and action capability
• Leaders should understand the technology, develop a strategy, and identify high-value, low-risk areas to start
• Approximately 50% of knowledge worker roles will be disrupted by agentic AI
• Mental model shift: recognizing AI can do work we currently believe only humans can do

Forget what you thought you knew about automation. 

Agentic AI represents a fundamentally different capability: artificial intelligence with genuine agency that can act independently, reason through problems, and execute solutions without human intervention. 

As Simon explains, this technology allows organizations to create synthetic workforces that can dramatically expand operational capacity at a fraction of the cost of human teams.

The examples Simon shares are nothing short of remarkable. A team of 20 AI agents completing complex risk analysis in one hour that would take human consultants six weeks. Ten AI agents running an entire insurance operation – brainstorming on Slack, managing claims, ensuring compliance – and delivering higher customer satisfaction than human teams. These aren't future possibilities; they're happening now with technology that's already a year old.

For business leaders, the implications are profound. The strategic risk isn't just falling behind technologically – it's being outcompeted by rivals who triple their operational capacity while slashing costs. 

Simon makes a compelling case that we're approaching a watershed moment where AI capabilities will match or exceed human cognitive abilities in many domains within the next decade. This requires a fundamental shift in how we think about work, organizational structure, and competitive strategy.

Ready to explore how agentic AI could transform your organization? 

Then lets chat - ☎️ FREE introductory meeting: https://calendly.com/kierangilmurray/30min

Chapters

• 0:00 Introducing Agentic AI
• 2:06 How Agentic AI Works
• 7:44 Real-World Case Studies
• 15:20 Strategic Risks and Ethical Considerations
• 20:55 Future Implications for Business and Society
• 26:16 Closing Thoughts and Resources

Transcript

Mr Kieran Gilmurray: 0:00

AI just doesn't 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's Kieran. I'm a Globally Recognized Authority in AI, automation and Digital Transformation. My guest today is Simon Torrance, a Globally Recognized Strategist and Innovation Advisor with over 30 years of experience helping board CEOs and entrepreneurs unlock growth through advanced technologies, and entrepreneurs unlock growth through advanced technologies. He is currently one of the leading voices on agentic AI and today we're going to go deep into what agentic AI is and why it actually matters. Simon, welcome to the show. Thanks very much. Pleasure to be here, Simon. Let's jump straight in what is agentic AI and what's it for?

Simon Torrence: 1:03

Well, as the name implies, it's AI with agency. So, just as we, as humans, have agency, we can make choices and we can act independently. That's what agentic AI is. It's AI that can make choices, act independently, work out how to take those actions and then solve problems in new and different ways.

Mr Kieran Gilmurray: 1:28

Well, why is now the time that business leaders need to pay attention to agentic AI?

Simon Torrence: 1:35

Well, because now it's possible. So you know, with the advances not just in LLMs but other technologies that wrap around those, it's now possible for organizations to almost replicate their human workforces or aspects of their human workforces with what we might call synthetic workers, AI agents that are able to increasingly undertake more and more of the tasks that today only humans can do. And that wasn't possible a few years ago, but now it is. And so you, as a leader in enterprise, you are now given an opportunity to augment and supplement your human workforce with a new digital or synthetic workforce made up of AI agents. And because they are software, they're nearly infinitely scalable and they're, of course, much cheaper than hiring more and more humans. So you have a you now. If you have, say, 10,000 human workers, you could in a few years' time have 30,000 synthetic workers, ai agents, which are undertaking tasks that only humans could do before, giving you much greater operational capacity, the ability to do many more things more effectively, and that is a very, very powerful competitive advantage if you do that before your competitors.

Mr Kieran Gilmurray: 3:09

I suppose, Simon, we've semi-heard this promise before with robotic process automation and intelligent automation, but why is a Gentic AI different?

Simon Torrence: 3:21

Well, because it's not just about automating processes. It's really having, I guess, workers undertaking tasks end-to-end, without having to be stuck in a particular workflow or process. So, for example, an agent can, or a set of of agents can, brainstorm solutions to problems and then undertake actions to to make, to activate those solutions, whereas of course, an automated process doesn't do that. It just does what the process, you know, the in the input and the output, uh, you know tells it to do. It doesn't come up with new ideas and then activate them.

Mr Kieran Gilmurray: 4:03

Yeah, although many would have loved if it even did that. But there's the brittle nature of what was first edition RPA or more tasks. So where is a Gentic AI headed over the next 12 to 24 months? Because this sounds like an accelerating technology.

Simon Torrence: 4:21

It is absolutely Now. It's already here, but in tiny, tiny, tiny pockets. So, like many technologies, you often have the situation where the future is already here, but it's not evenly distributed. So, and then later, I think, on our discussion, I'll give you some real life case studies that bring it to life, that showing what is possible today, even a year ago. Live case studies that bring it to life, that showing what is possible today, even a year ago.

Simon Torrence: 4:53

But if you look one year or two years out, as you've just asked, we're going to see what I'm what we'll describe later on at much bigger scale, because more and more companies work out how to adopt this new approach, and they do so. So, in parallel with that, you've got the technologies underlying agentic AI getting stronger and stronger by the day as well. So you've got things that are already working incredibly well today, which will scale and they'll be even more powerful because technology is developing as well. If you just tip into maybe three to five to seven years out, then a lot of people think we're heading towards, if you like, human AI, intelligence, getting, you know, almost up to the same levels and again, that will, you know, multiply the potential of this even further.

Mr Kieran Gilmurray: 5:40

So how does agentic AI actually work and how is it different from generative AI tools like chat, gpt or co-pilots?

Simon Torrence: 5:48

Yeah, so it uses LLMs as part of its I guess, its content generation and learning capabilities. So there are five key elements to an individual AI agent. Five key elements to an individual AI agent. One is that it has the LLM itself and it can actually use, get access to, multiple LLMs.

Simon Torrence: 6:10

Llms are good at different things. It has a persona. So, a bit like myself, I have a persona, I have role, I have tasks, I have behaviors some good and bad and I have a personality and I have knowledge and access to data. And we give an agent the same. We give them a bit like a job description here's your job, here's your persona. We tell them this is how we want you to behave. Here's some data that you have access to to help you do your job. So that's the second point. We've got the LLM, we've got the persona.

Simon Torrence: 6:45

It then has the ability to reason, so to take a task, break the task down into subsets and then go through that process to achieve those, to go through that process. On top of that, it has memory. That's really important. So it doesn't try and apply the same solution to the same problem over and over again. It learns from previous experiences. That's really good. It has short and long-term memory. Then, finally, it has the ability to take action. It is connected or it connects through APIs to other systems internal and external systems, also to other agents themselves to take action based on the task it has, the way it's decided to achieve the task, the memory and so on, and the intelligence of the LLM in the center. So those are the five elements that make up an individual agent, and then, of course, you have multiple agents that can interact with each other and interact with humans.

Mr Kieran Gilmurray: 7:47

So if I'm a board or a CEO, what do I need to do to start experimenting or building relationships with the Gentic AI?

Simon Torrence: 7:57

Yeah, well, first of all, you need to understand what I've just even said, you know, to appreciate that this is a thing that exists today, that some companies in small pockets today are using it and achieving incredible results just with technologies from a year ago. So that's one thing to even appreciate, even understand what it is and why it exists today and where it's going to be in, say, three years time. So that would be step one, you know, just to understand it. Step two is to then create a strategy for where could we, how do we want to deploy this within our organization? What's our strategy for taking advantage of this new asset to create competitive advantage in our market?

Simon Torrence: 8:47

So that is really important. Which domains of our business do we want to apply this to now? To learn quickly, maybe high value, low risk areas we can learn quickly and what are the areas that we really need to focus on to scale up, to maintain the advantages that we have in the market today? So those are the first set of things that need to happen, and there's many other things we can come on to later on about how you then scale it up, create the right guardrails and governance mechanisms for this.

Mr Kieran Gilmurray: 9:19

Can you share some real world cutting edge examples of agentic AI in action?

Simon Torrence: 9:24

then yes, yes indeed. So I'll give you two examples, which I use quite a bit, because they illustrate two different things. A lot of people say, well, agents, they can do certain basic things today. You know, maybe they could help me book a flight or manage my calendar. You know, I tell it what I want it to do and it organizes those things for me. Now those are extremely basic types of applications of an agent. But I'm going to give you two examples that are real today. They were actually done over a year ago and they demonstrate the art of the possible.

Simon Torrence: 10:02

The first illustrates that agents can undertake complex knowledge work, work that traditionally you might pay a Deloitte or an EY consultant £2,000 a day for. And this is an example of a company that does big procurement projects. They're buying very large capital equipment. It would typically ask a consulting company to do a risk assessment and a cost assessment Might take six weeks, six people, half a million pounds to do that, and they do that many times over in the year. A very typical type of analytical project, but very, very high value, complex analytics. Now my colleagues were asked could we do it with AI? And so my colleagues created a team of 20 AI agents agent, several analyst agents, some domain experts, some quantitative analysis experts, et cetera put them together, trained them in three weeks from nothing, and pressed the button to do the work, and in one hour they completed the same task that the human consultants take six weeks to do one hour. And in so doing they also looked at 50 different cost scenarios rather than just one, and they all brainstormed together, chatted together and came up with this solution. It was as good, if not better, than the human team, but done in one hour. And of course the organization now has those assets. Those are replicable agents they can use again rather than paying out 500,000 pounds every time. So that's one example and that's a year old, that example with technologies from a year ago.

Simon Torrence: 11:54

The other example I'm going to give you is even more mind blowing, and this is whereby a company wanted to create a new insurance business. It wasn't an insurance business, we wanted to create one that would complement its core business. Insurance is quite a complicated topic, heavily regulated. The company found it very difficult to hire humans to their company because they weren't an insurance company. They're very expensive, these workers. So they said to my colleague could we create an operational unit to run the insurance business with agents or with AI the term wasn't used then. So we created a team of 10 AI agents again customer service reps, we had actuaries, gdpr experts, we had a manager to manage them and in three months we created this complex team. They cost, of course, a tiny fraction of hiring 10 humans 10 humans in insurance maybe a million euros, million pounds. This tiny fraction and those agents now they still do it now.

Simon Torrence: 13:07

This is from the beginning, over a year now. They brainstorm and discuss and collaborate like we do on Slack. So they're there, they all have a persona, they have a role and they interact and solve problems on Slack related to the operations of running an insurance business, including managing claims, product innovation, et cetera, et cetera, regulatory compliance. So they brainstorm on Slack, they come up with decisions, they activate those decisions and occasionally and occasionally to start with, a lot of humans were in the loop to check things before it went out. Now the humans participate less.

Simon Torrence: 13:48

Some of the human workers have asked to be cloned so they can keep up with the agents because they work very fast. Come back to that if you wish. And what's most important, that team is now delivering higher net promoter scores than the human customer service team. It is more efficient when we've done tests on managing issues coming through and through brainstorming, product innovation. It is achieving much higher underwriting profitability than normally is the case for this type of product. So you've got incredible cost savings, the ability to even launch a business because the company couldn't before, greater customer satisfaction and greater underwriting profit. So an amazing example which has been going on for nearly 18 months now, and I share that, because 99.999% of people have never heard of that before, don't even know what I've just described is possible, and that's just so. That's something, an example of something that exists today, and you could apply the same principles to nearly every aspect of any type of knowledge intensive business.

Mr Kieran Gilmurray: 15:08

I think, having worked in insurance brokerages and insurance businesses for 10 years, you had me at. The robots or the digital workers were actually talking to each other. That in itself was much more impressive than I've seen inside of insurance companies. How do you see agenda AI then transforming platform and ecosystem business models in the foreseeable future?

Simon Torrence: 15:29

Yeah, well, I guess it's the same for any type of business, whether it's a you know it's a traditional vertically integrated business, or a platform or ecosystem business. Essentially, it's giving you more workers to do more work for a fraction of the cost. That's the principle that cuts across everything here. So if you need more workers to look into fraud, you don't have to hire humans. You can get teams of agents to do that. You can get teams of agents that represent different silos of a business to come together to constantly monitor things that are quite difficult to do within traditional silos. So I think the principles apply to any type of business. It's quite interesting Now.

Simon Torrence: 16:13

Amazon is just a lot of you think about. A platform business like Amazon is just launched a, an agentic advisor platform business like Amazon. It's just launched a, an agentic advisor. So you know you, you, you express an interest in something and it will go and look at, look over the whole of Amazon and come back to you with recommendations about what you might be interested in, and then pretty soon, I think maybe even you can now you can ask the agent to buy it for you and ship it to you and all that sort of stuff. So the principles apply, it's being able to have many more, if you like, workers working to achieve the objectives that you have.

Mr Kieran Gilmurray: 16:50

So what are the key risks in the strategic, the operational or the ethical risks that we should be planning for or should be aware of if this is live now?

Simon Torrence: 17:00

yeah, there are many risks. I mean I. The strategic risk would be your. Your closest competitor adopts this at scale before you do. You know that is extreme. You know that for me that's probably the biggest risk. You know you. You decide I haven't. This sounds too complicated. Don't want to do it. If your competitor suddenly triples their operational capacity in three to five years' time, how are you going to compete with that? For a fraction of the cost, you can't compete with that. They'll lower the prices. They'll be much more efficient than you. That's the number one, the biggest risk, I believe you. That's the number one, the biggest risk, I I believe.

Simon Torrence: 17:43

Now there are other risks, um, related to technical risks, of course there's. You know ai in general is much more well. It's at danger of being, uh, hacked again by, by cyber hackers etc. So you need to put in plenty of guard rails there. There are ethical risks as well, because you're you, you. If you're interacting with customers or with other employees, you need to do that in a responsible way. I mean, agents have no morals or ethics unless you give them ethics, those morals and ethics to guide them.

Simon Torrence: 18:14

So I'll just give you an example in the insurance case study, we gave the team the task of reducing claims costs so that increases profitability, and their initial approach was to just not continue the policies with certain customers because they were high risk. Now that's slightly illegal, stroke, unethical in insurance. So we had to tell the agents you know there is a. You know we have to pay a certain number of claims costs. That's part of what insurance is. But they've got it down to almost nothing.

Simon Torrence: 18:55

So so you, because they just work 24 hours a day, 365 days a year on the objective you set them and they test lots of ideas until they achieve that objective. They don't have any of our not not any ethical and moral boundaries, but they don't have any of our time boundaries or the need to have to sleep. So they, that would be. That's another one. And then you have the risk of just not having the skills in-house to to take advantage of this technology. That's a major, major risk, as you attract people who can manage a agentic workers and that's a new skill. Um, and there's all kinds of other financial risks of making bets in the wrong place and so on. So plenty of risks related to this, and that's why controls and guardrails, both technical and ethical, are really, really important if you want to scale this up that said, though, you can program that into a genetic AI, which may actually give you an advantage over the human world.

Simon Torrence: 20:01

It's a great point you make. In fact, again in that example, we put in place agents to check the work of the other agents, to make sure it was ethical when we had a GDPR agent that was making sure everything was compliant with that ethical. When we had a GDPR agent that was making sure everything was compliant with that. But also we gave an agent the values of the company, the, the that we wanted the company to adhere to, and they checked and acted as the conscience for the rest of the agentic team. So you're quite right, you can build I mean, you can build in quality assurance agents as well, but also you know security and ethical agents to make sure they're acting accordingly.

Mr Kieran Gilmurray: 20:42

Will agentic AI amplify or reduce inequalities in between companies, workers or potentially even countries that use agentic labor for all sorts of different roles and uses?

Simon Torrence: 20:55

Yeah, it's a great question. Well, certainly between companies. I think it will create significant inequality. As I said, if one company really adopts this before the other, it will have a massive competitive advantage.

Simon Torrence: 21:07

Now, if you think about a, let's think about a country as a whole. We've got some other considerations here, because for most, you know something like I think 50% of workers in the West are what we'd call knowledge workers, like you and me and many people who work as white-collar workers within organizations. I think it's about 50%, now maybe even more 60%. We estimate that about 50% of those roles will be disrupted by agentic AI, by AI agents. Over time, probably 15% to 20% of people's current roles can be replaced, and so you've got another 30% odd that could be will change. I will have to learn, you and I will have to learn how to manage AIs, manage people and manage AI and people working together. That's a skill we don't have today.

Simon Torrence: 22:04

So if you've got that disruption to half of half the working population, that's a major, major issue, and so governments will need to think about that issue. They'll need to put in place reskilling capabilities and, just as we saw with manufacturing, where manufacturing was hollowed out in the West, you know, sent to cheaper places abroad, leaving, you know, I guess, people who felt left behind, people who used to work in manufacturing felt left behind, not having a feeling of place in the world. That creates a lot of disgruntlement, of course, and we see that in politics today, and so I think political leaders need to anticipate this, you know, today, and start thinking about how they're going to mitigate and manage this, working with industry, because industry will be driven hard to adopt agentic. It's not going to stop. So government needs to understand how fast that's going to happen and work with industry now to think about what the implications are.

Mr Kieran Gilmurray: 23:15

Well, those could be quite significant, because we have seen the hollowing out of manufacturing and if we see the hollowing out of so-called knowledge worker industries, we really are in trouble economically at a time when we need the middle group, not the reduced birth rate or the accelerating elderly. We need that middle group to derive more value for all of us. So what's the one mental model you think leaders need to adopt or shift if they want to succeed in this era of agentic AI?

Simon Torrence: 23:45

Yeah, I think it's to recognize that AI agents are capable of doing the same or similar types of knowledge, work or much of it that today we think only humans can do, and that's only going to get more and more the case. So the notion of when super intelligence occurs, when humans and AI are at the same level of cognitive ability people used to think it would be in the second half of the century and then it got a bit closer and closer and now there is consensus that it's about in 10 years' time. Some of the leaders of the big AI companies say it's in three years' time, but I think there's now consensus it's 10 years' time. If you think it's longer, you're an outrider, so the conservative view is 10 years time. If you think it's longer, you are, you're an outrider, so the conservative view is 10 years time. Now that changes everything. If an AI can do the same cognitive work as a cardiologist or a judge or a consultant like us, or a software developer like us or a software developer, whether it's three years, five years or ten years, that's a major, major shift in the way that the economy is going to work and how labour markets work.

Simon Torrence: 25:05

So, appreciating that and understanding the art of the possible today, like the case studies I've just shared. You can do that today. You could have done it a year ago. Like the case studies I've just shared, you can do that today. You could have done it a year ago. Now, understanding that and preparing for that is the mental model shift that is really needed. And I have to say, in the work that I do with very large organizations that employ tens of thousands of people, today, in knowledge-intensive sectors, the leaders have not repeat, repeat have not processed this yet. So I spent a long time just explaining what we've just talked about and then the penny starts to drop a bit slowly because these are big in, you know, big organizations. But I think that's the big mental model shift that's needed.

Mr Kieran Gilmurray: 25:51

Yeah, I think I've argued for some time now. It's not the technology, it's the limit of our imagination that holds us back, and organizations that are set up and they all are to maximize profit, beyond some of the social cares or some of the altruistic industries. Their job to maximize shareholder value can certainly be made so much simpler by this, but, as you say, it is all the implications in the world. Simon has been a master class. Today, we've learned that agentic AI isn't just about smarter systems. It's about creating autonomous agents that can reason, act and adapt on behalf of their business. This opportunity is enormous, but it also, by the sounds of things, really needs responsible leadership. Simon, for those watching, where's your latest work on the Gentic AI? Where do they follow you?

Simon Torrence: 26:35

Well, so there's two places. One is on the website, so that's ai-riskco, and then, secondly, I publish a lot on nearly every day on LinkedIn. Most of it I do myself, some of it I outsource to AI, but I always edit it. But so if you look, if you follow me on LinkedIn, then I publish a lot of material there.

Mr Kieran Gilmurray: 26:57

Fantastic Look folks everyone joining today. Thank you so much. We'll put all the links down below. Until next time, I'll see you when I see you. Thank you, Simon. Thank you audience for watching.