CR080: AI-driven hyper-automation with Nicole Onuta, AI business transformation expert

As AI scales and layered agents become more integrated, businesses face questions about just how automated they can become and what this means for efficiency, cost, and innovation.
 
This week, Dave, Esmee and Rob talk to Nicole Onuta, AI business transformation expert about the growing impact of generative AI on the software development lifecycle (SDLC), diving into both the opportunities and challenges that arise, how AI scales in an organisation and how hyper-automation is fast becoming a practical reality.
 
TLDR
04:40 Confused about whether frictionless CX can be too frictionless
07:54 Cloud conversation with Nicole Onuta
34:57 The Eliza bot in 1966
43:08 AI execution!

Guest
Nicole Onuta: https://www.linkedin.com/in/nicoleonuta/

Resources
Turbocharging Software with Gen AI: https://www.capgemini.com/insights/research-library/gen-ai-in-software/

Hosts
Dave Chapman: https://www.linkedin.com/in/chapmandr/
Esmee van de Giessen: https://www.linkedin.com/in/esmeevandegiessen/
Rob Kernahan: https://www.linkedin.com/in/rob-kernahan/

Production
Marcel van der Burg: https://www.linkedin.com/in/marcel-vd-burg/
Dave Chapman: https://www.linkedin.com/in/chapmandr/

Sound
Ben Corbett: https://www.linkedin.com/in/ben-corbett-3b6a11135/
Louis Corbett:  https://www.linkedin.com/in/louis-corbett-087250264/

'Cloud Realities' is an original podcast from Capgemini

Creators and Guests

Host

Dave Chapman

Chief Cloud Evangelist with nearly 30 years of global experience in strategic development, transformation, program delivery, and operations, I bring a wealth of expertise to the world of cloud innovation. In addition to my professional expertise, I’m the creator and main host of the Cloud Realities podcast, where we explore the transformative power of cloud technology.

Host

Esmee van de Giessen

Principal Consultant Enterprise Transformation and Cloud Realities podcast host, bridges gaps to drive impactful change. With expertise in agile, value delivery, culture, and user adoption, she empowers teams and leaders to ensure technology enhances agility, resilience, and sustainable growth across ecosystems.

Host

Rob Kernahan

VP Chief Architect for Cloud and Cloud Realities podcast host, drives digital transformation by combining deep technical expertise with exceptional client engagement. Passionate about high-performance cultures, he leverages cloud and modern operating models to create low-friction, high-velocity environments that fuel business growth and empower people to thrive.

Producer

Marcel van der Burg

VP Global Marketing and producer of the Cloud Realities podcast, is a strategic marketing leader with 33+ years of experience. He drives global cloud marketing strategies, leveraging creativity, multi-channel expertise, and problem-solving to deliver impactful business growth in complex environments.

What is Cloud Realities?

Exploring the practical and exciting alternate realities that can be unleashed through cloud driven transformation and cloud native living and working.

Each episode, our hosts Dave, Esmee & Rob talk to Cloud leaders and practitioners to understand how previously untapped business value can be released, how to deal with the challenges and risks that come with bold ventures and how does human experience factor into all of this?

They cover Intelligent Industry, Customer Experience, Sustainability, AI, Data and Insight, Cyber, Cost, Leadership, Talent and, of course, Tech.

Together, Dave, Esmee & Rob have over 80 years of cloud and transformation experience and act as our guides though a new reality each week.

Web - https://www.capgemini.com/insights/research-library/cloud-realities-podcast/
Email - [e-mail]

CR080: AI-driven hyper-automation with Nicole Onuta, AI business transformation expert
[00:00:00] I didn't expect this to take this turn, I have to say. No, well, what is life if you're not happy? I mean, let's be really honest about it. If you're content and happy, then that's the best place to be. There is no other answer. We should do a different. There's a podcast on this. It's a very interesting topic because actually connecting to Gen AI and into AI overall.
Nicely done, Nicole. I am glad there's a professional on the show at the moment. So looking at that, The desire of us being happy and that we strive so much for it. And this music actually helps us go into also exploring other types of feelings. And actually it's part of game theory. If we want to reach an optimum state, so also in life, but also a state of a system, you have to actually explore all dark states before. It's part of game theory.[00:01:00]
Welcome to Cloud Realities, an original podcast from Capgemini. And this week it's Conversation Show. You're exploring scaled AI in the software development cycle, uh, but also layered agents. And what might happen when you start joining some of these AI dots? Just how automated can a company become? I'm Dave Chapman, I'm Esmee van de Giessen and I'm Rob Kernahan.
I'm delighted to say that joining us this week is Nicole Onuta, an AI business transformation expert. Nicole, it's great to see you today. How are you? so much for inviting me. I'm great. Looking forward to our talk. Do you just want to say a little bit about yourself and some of the work that you've been doing recently?
Yes. Hi, I am Nicola Onuta, AI business transformation expert. I'm actually going to share that my journey with AI started long before I realized it. So back when I was, when I first fell in love with the mathematics, I can still remember being in an intensive math class in secondary school led by this incredible [00:02:00] teacher, Arthur Beluca, who always pushed us beyond our limits.
And his famous line still rings in my ears. He used to say, I work with you 25 hours a day and you still can't answer simple questions. Simple, of course, for his mind. But looking, looking back, I realized how much that early intensive math training shaped my way of thinking and my passion for AI. I graduated with a master in applied mathematics, started working in finance, where I noticed that financial institutions are quite advanced in using AI, especially for risk.
Pricing space is so concentrated, but not so much, uh, in operations or customer experience. I decided to do something about it. So we set up, uh, with a couple of colleagues, we partners, we got executive approvals. We set up on global center of excellence aimed to expand the benefits of AI before beyond risk, uh, within that finance company.
Uh, again, the journey is [00:03:00] not an easy journey. We got a lot of ideas. A fraction of them moved into experimentation. And I think within a four years period, there were five real solutions deployed. They did deliver millions of Euro return per year. So that is my journey with AI and seeing the power of this technology.
Now, I am taking an unexpected turn for myself as well, as I said, I'm, uh, my mission is moving more into AI governance and risk management. Well thank you for joining us today and we are looking forward to digging into some of that history and, and speculating a little bit about what the future might be like, uh, very shortly.
But before that, our colleague Robert is here, uh, as is Ez, how are you guys? You good? Uh, living the dream, Dave. Living the dream. Living the dream. Ez, you good? Yeah, yeah, no, I'm just wondering if it's a dream or a nightmare, but it's a very happy dream, actually. Let's see how this unfolds. It's the rain, Dave.
It changes your [00:04:00] mood. Come on. It's raining again, isn't it? Summer's gone, so. It's also cold, isn't it? Yeah. Temperature's dropped. The wind is chilly. Got to get the fire on now. That's it. That's right. Yeah, I put the heating on the last couple of days. It's a big moment, isn't it, in the house? It's almost like you have a family discussion.
They don't know, should we put the heating on? It is time, and then you walk to the thermostat, and it's like a ceremonial moment in the household signaling winter. I've just been using Boost so far, you know, that I'm not going, I'm not going all in yet. Just the odd little boost every now and then. You dipped your toe in.
I just turned it on and went, I like to be comfortable. Just wear a sweater. Yeah. Robert, what is confusing you this week? Well, what confused me, David, what confused me, it was 7 o'clock in the morning, I was stood on a train platform, uh, and I accidentally opened a global banking account. I hope you haven't said that, that's what's confusing me.
That's an unusual way to start the day, isn't it? [00:05:00] You know, we talk about digitalization and digitization, and we talk about changing the way we interact with consumers, citizens, customers, etc. And this is the perfect example. I'm there, I just go into my banking app, so I've authenticated, and they say, oh, are you interested in this?
And it was like, button click one. Oh, this sounds like interesting. Button click two. Oh yeah, I am getting interested. Got distracted. Of course, I'm going through like, I think the swipey screen telling me about it. Button click three actually said open accounts. And I did, and it opened it. And I'm stood there and I think to myself, Oh, that's happened.
And, uh, as you reflect on it, though, how low friction was that? Of course, I'm authenticated. So they know who I am. They know everything about me. They think this is a product that I need is actually a product. I need three clicks later. I've got one. And I thought to myself, That's actually a cracking example of digitalization, of creating a new interaction model that says, I don't actually have to fill in a form to do this.
They go in, buy your persona, this probably fits. And you [00:06:00] go, yeah, all right. But it was like this surprise moment at seven o'clock in the morning on the train station platform where I went, oh, I've got a new bank account. They Instagrammed the experience. It's like, confused. Yeah. Hashtag confused. Hashtag global bank account.
Yeah, yeah, but I was like, you know, when you go and then there was, I was, I was a bit annoyed with myself to go, I should have really read that, but then I thought to myself and went, actually, that's a really good example of new digital experiences. So I do wonder what, uh, what fees you've signed up to on this.
Yeah, I did actually check that quite quickly, Dave, and it turns out it's, it's not charged. Yeah, that's, that's lucky. Now, global bank has an unusual thing to have. So what, uh, what, what have your global operations, with it, Robert. It's as simple as, if you're a bit clever about the type of bank account you open, when you travel, I travel quite a bit for work, so you're in different countries, it's just more efficient with transaction fees and everything else and currency, so it's just a thing that lifts the burden.
That wasn't You're using money That wasn't as exciting an [00:07:00] answer as I was hoping for. It's not, you wanted to say I needed it to, you know, charge up me Volcano, Underpin the smuggling operation. Underpinning the smuggling operation, yeah, all of that. Unfortunately, no, Dave, it's just practicalities about avoiding transaction fees overseas.
Well, excellent. Lesser confusion this week, more of a very satisfactory digital experience I'd say. It was a surprise, but the end of the surprise was actually quite good. So there you go. But I, for the computer, I was ultimately confused when the, when I, I would say you looked a bit sort of like shocked.
Certainly still in the blackboard, he's going, I'm actually confused by your, your, your excitement. Like usually we go down to Yeah, we usually go, we came into the meeting, because we were actually in the same meeting that day, and he came into the meeting and he goes, guess what's just happened?
And there you go. Anyway, let's dive back into the ever accelerating world of AI. Let's start with digging just straight into the [00:08:00] report. So, Nicole, I believe you've been looking at turbocharging software with Gen AI. Just set out, set out the boundaries of the report, what you were interested in looking at before you started the work, and then did you, you know, what did you discover?
So just, just set the scene for us to start with. So related to the report, it looks to give insights into what is the value add of GenAI for software engineering, also to set standards, benchmark with different companies and understand what is good to do to start and what is good to do to scale. So this is a high level summary of what the report provides.
Right. So let's, let's just take each one of those together and then we'll perhaps come to the conclusion. So starting with it, what are the sort of steps and the advice that Basically is as simple as giving your devs access to tools or is there something more structured in there? Yeah, so what we see companies doing and most of the companies, big corporations started already with pilots.
So [00:09:00] in pilots, the first thing to do is indeed to contemplate the risk base. And if we are okay to pilot with this technology, again, there's not a robust governance or risk framework around AI. Most companies do not have. that. So we're looking to, to have working groups understanding the risks and approving the pilots.
Most companies already over this stage running pilots and they are now looking to understand how to scale generic device for software engineer. Part of the pilots, what has been discovered is that generic for software engineer does deliver productivity gains. What we see is that the technology provides like good suggestions.
checks errors. Um, and this productivity games are estimated around seven to 18 percent improvement across the software development life cycles. I was talking to a friend of mine who runs a Google systems integrator and we were chatting this morning and actually stumbled on this topic and he was saying they were, they were seeing 30 [00:10:00] percent productivity uplift in some aspects of it.
Now, I don't know how much of that was sort of a hardcore application dev versus, you know, platform creation and cloud tool usage, but the, it seems like the efficiency upside is very high. Yes, actually, um, this is one of the more established use cases for Gen AI. And it also depends on the task and depends on the type of software engineering job that Gen AI is applied to.
From my experience, when I've seen even In translating legacy code from one language into another, we could even achieve an 80 percent productivity improvement that was the most powerful application. So it really depends on what are we talking about? What is the group focusing on? Again, software engineering is is broad, so it's from requirements, gathering and design, implementation, testing, et cetera.
Very good point about that. It's incredible. improvement for legacy transformation, because it's always held us back as the complexity and the amount of [00:11:00] effort and everything else. I mean, is this Not to mention the age demographic of people who wrote the original code. Yeah, exactly, right. I remember back to Ron Toledo's that we did, and he said a father and son had worked on the same code base, if you remember that.
Do you remember, uh, yeah, and at this point now, is this, is AI and this software assistance via AI, going back to the MEC 2, it feels like it might be the start of the first of the deletion of the legacy, slaying the legacy dragon. It's like there's enough in it now to say, right, now is the time to get this sorted.
It's an interesting question. So again, from my experience, what we've worked is translating R into Python, for instance, and just works really good. Um, now the problem with legacy systems, as you say, is yeah, resource constraints, but also the fact that these languages are old and they are not known widely.
Um, so larger foundational models like open AI, Gemini aren't always effective for these specific legacy [00:12:00] systems because they're not trained on this older languages. So there is a potential for smaller, customized large language model in this space. Again, I've not personally been involved in such a use case, but what is now discussed.
is that smaller large language management models can be deployed within a company's own cloud and trained specifically on their legacy system, making this, uh, this process feasible. But again, this is in itself an investment. So depending on the size of the company, the size of application, all these factors need to be considered before concluding whether this is the right direction.
I mean it does, it does feel to me like the, the potential of that is very high for the first time. I think it stands on the shoulders of a couple of things. It stands on the shoulders of getting our heads around, I think, how we decompose large legacy these days and things like microservices architectures help with that, you know, taking the risk out of some of that decomposition.
But then when you've got, if you've got code translation in the way that you guys have just been describing it, [00:13:00] I do think it provides. A useful route map to be able to not only do the conversion, which before was actually difficult now and impossible because of the language constraints and demos and all that sort of stuff, but not only does it fix for that problem, but it will probably fix that problem very, very quickly.
So it will do that with a very high level of productivity attached to it and efficiencies attached to it. So it could end up being cost effective as well as possible. And at the moment it's, we're still balancing those two. Those two pieces, I would say, and there's a bit in there around, you know, a lot of people are starting to rewrite their legacy because they're fed up with it and they think I'll just, you know, it's input output, but there is another side to it, which says I don't really want to change my legacy.
It does a job, but I need it on a supportable, configurable platform that I can secure in a modern way, so I might not be changing the code base. And in this scenario, that's an excellent way to just deal with the headache of old things. So at least you're [00:14:00] modernizing to a degree. It's not the perfect answer, is it?
But it at least gets you out of a hole. And again, all the heavy lifting done with the intelligence. Yeah. Well, let's return to the story. I think that's an amazing use case. Actually, for me, it's like right up there in terms of one of the best first uses. But Nicole, you talked a little bit there about sort of getting into it and, and the early stages and how you address that.
Talk to me then about the effects on the, on the wider software development cycle and how you might start to think about scaling use of AI in teams. So, in terms of scaling today for software engineering, what, what I see happening is the organizations are taking step by step, a step rollout plan. So they are looking at a couple of aspects.
One of the aspects is indeed what type of task is a specific department busy with. The second item that is very important for considering rolling out is what type of environment are they using? Because that is really important for determining what type of [00:15:00] system is best for that environment. Again, for having the, the, the maximum productivity gain is good that this, the assistance and, and, uh, the gen AI, um, uh, technology is embedded in their way of working.
And the third important item that is looked at is training. Of course, these models are still hallucinating. They are providing, uh, pieces of code that actually do not run. Because there is also the copyright concern. So training, understanding exactly how to use the suggested pieces of code, it is an important part of the scaling plan.
How would you advise teams handle that? So what kind of good practice have you seen? that teams have been using, do you see it down to like each individual to root that out? Or do you see being able to use some of the, you know, like standup culture is the, is the good practice that you've seen or good tooling that you've seen that can help resolve some of the issues?
So it is a combination of training and also activated, activating communities in this [00:16:00] space where people can support each other. They can have dialogues. on, uh, and share solutions to common problems. Communities play an important role in this. Also what we notice for beyond productivity gains. So using this technology for software development, uh, delivers a more happy software developer.
So connecting on this topic is actually leads also to a higher engagement. And we've discussed this a few, I'll bring my stats out again, David, the exact 20%, if you remember, which is the, uh, uh, happy person is 20 percent more productive. And then you've got the product productivity gain on AI. And, um, 20 percent on average robber is that every single person is exactly 20 percent exactly 20%.
Love the way you use it. Good in it. And Esme, you had the thing about intrinsic motivation, which is the mastery purpose and the autonomy. See, I did remember I used that the other day as well. I've learned. And then you get this point where you go, this is a really key part [00:17:00] in what we're doing to create a happier ecosystem of people who all come together.
It's sort of like the foundation of digital is psychological safety, happiness, autonomy, blah, blah, blah. You got a list. And jumped by Van Halen. and jump by and I just, if I had a business, I just play that through the speaker system all day. That'd be it. It's just, you would never get away from it. You'd never get away from it.
That would be a psychological experiment and everybody would be super happy. I guess 21. But this is a really key part about happiness and we keep coming back to it, but it's so, so true that the happier people are just. better at everything that goes on in life. So, yes. And I think this, this taps on the flow in organizations.
It might be something that's been triggered with me the last couple of days that I was, you know, feeling a lack of flow between departments, people, life in general. Um, so I think if these, agents or bots or AI is able to help you get into that flow to remove bottlenecks, because that's what [00:18:00] we also do with lean.
And, you know, we tried to pinpoint waste, but if they are actually able to, to get away or to remove all those obstacles, I would become more up here. To be honest, do you see anything happening there, Nicole, like end to end in processes that it can actually, you know, move work forward instead of a human being that needs to see that it's get stuck somewhere?
In a scale environment, yes, definitely. So if we look at the long term vision, right, for scale, again, we're going to say we are not yet there. And to be honest, there is still a question because this GNI is a hype and organizations would like to see more in terms of return on investment. And we are kind of at a tilting point where there's, there's been a lot of investments, but we did not see this investments fully coming back and return on investment.
So the long term vision, so how can we get more benefits? So I look at a couple of dimensions. So, um, as you say, expanding adoptions. through [00:19:00] all spaces. If you look at the software development cycle, so again, requirements, getting requirements, gathering design. So to truly scale, we can expand the adoption beyond coding and testing into more challenging areas like requirements, gathering and design.
Design in particular poses a challenge even in the Uh, if, even if we would use the most advanced models right now, like current multimodal models, uh, because of course design is, has to do with solution visioning, requesting understanding context, the creativity. So if we expand the adoption to all spaces, we are hitting challenges, the challenges of performance.
However, if these challenges, if that, point, we, if we reach the point where these models become even more advanced, then we would be able to expand. Uh, so that is indeed a petition, but this is more on the long term vision. And the long term vision is, yeah, creating this more symbiotic relationship between AI and developers where AI acts as a trusted advisor.[00:20:00]
It's a partnership where AI handles more operational, repetitive work, and the software engineer gets more time for high value, adding tasks, um, more creative tasks. Um, so for instance, if you think about a construction place, right? Um, so the cons, the, the software engineer can, can focus on, um, thinking how to design the house while the AI partner, trusted advisor can say, yeah, can, can, can create the bricks, can decide what materials to use in the bricks.
So it is indeed creating this, this, this, uh, symbiotic relationship in the long run. Uh, also what I see, and again, these are not. These are visioning points. I also see a space of increasing productivity with what we call layered agents, layered agent systems. So right now, we do need to be very [00:21:00] critical about the outcome.
But a significant long term benefit lies in increasing productivity through layered agent systems. So imagine having a system where one agent writes the standard code, another creates the proposal for improvement, a review agent checks the work, and a final agent runs simulations in a test environment.
So it's a layered approach. It will allow for faster iteration cycles and continuous improvement. What's crazy is in that example, is if, if each of the agents you describe at the moment that was being done by a human, you know, that would have like an hour's or day's lag in that process that you've just described.
I'm presuming that the use of agents in the way you described, that's going to be seconds worth of turnaround to create an iteration. Right. We still gonna rely very much on the human in the loop, but nevertheless, the checks that are needed are going to be minimum. So, and, but for such a vision to create such a vision again, yeah, we need to, [00:22:00] what is needed is this scalable infrastructure, what I call for long term success.
And a company that invests in building that infrastructure where from an AI perspective, we train the AI. So an infrastructure that does the model training, does the data pipeline instead of for continuous learning. So organizations that are able to build this scalable infrastructure will able to capitalize on generative AI potential return a bit faster.
So Robert, this is the coming together of agentic intelligence and the mech suit. concept. I mean, I know we joke, but everybody loves a tightly coupled, overly configured ERP system that takes six months to redeploy and doesn't actually do what the business need. Yeah, but it's this bit where the agent, if it's, if it can be retrained, et cetera, it can adapt much more quickly to what actually.
Happened or needs to happen. So yeah, it's just the maturity bit, isn't it? Around the how do you get it up? So it does what you need it to do. And then the second part of it is trust, which is if you're [00:23:00] handing over decisioning or flow to a, an algorithm, a fancy one of that AI. You have to be able to prove the efficacy of what it's done.
And there are certain things that that, but we see the models starting to have the, the capability to show us it's working essentially. So then we can say there is an audit trail on why this decisioning was made. But yeah, it's going to arrive. I just think it's that balance between the first organization to do it successfully at scale edge to edge, it's either going to be an almighty success and then there'll be a huge rush or they'll destroy their company in three weeks is that thing, isn't it?
It's a high risk but high reward approach for the first that properly rewires their organization with this type of technology approach. Are you saying it's at the same level as when they switched on the CERN accelerator and they thought they were going to create a black hole? Well, they did it with the atomic bomb as well.
They thought it would ignite the atmosphere and destroy the earth. Um, turns out it didn't. And the CERN didn't destroy the earth either and create a black hole. So I suppose, [00:24:00] yeah, but I think it's, I love that. Well, I think those other two examples are a bit more dramatic than, um, your business process stopped working.
I do want to, we may well come back to this subject a bit later on, but Nicole, I want to pick you up on the point you made about having the right scaled infrastructure. So what are the core components of that for you? So the race killed infrastructure again is is a cloud infrastructure, and I'm going to come back to because this seems maybe, uh, a bit further in time.
But if you look at innovationists, they said that the next disruptive idea is, it's not going to come from one field, but applying, uh, learnings from one field into another. So if you look at finance companies, right? So in finance, uh, analytics is, it's been for some time used and, uh, advanced analytics within risk, within capital models is used for some time now.
So in this space, the regulatory requirements are very tough. [00:25:00] The only thing is for most banks and financial institutions, they apply after the fact. So there is a development process and there is. a validation process that follows. So again, looking and contemplating what is happening in that process. So after development, there's validation, there are checks, the checks, look at the data, what data has been used?
Is there bias in the data? Is the data stable in time? So all these checks take place after development. Now, if we look at this scalable infrastructure for the long term success, and again, I'm still talking here about building AI models, and this Next version. We would actually in part of model development, we would already qualify and do all these checks that are now as a second step are now part of the second step in validation.
So in the future, we would have modules which are. [00:26:00] Incorporated in the development process, which will automatically check all the requirements and say, yes, you're okay to move to the next stage. Now your data is stable, is representative. You're okay to move to the next step. So once we have a model going through all life cycle stages or steps that are needed for development, we already know it's good.
We already know it qualifies. So this is what I mean in terms of. Technicalities. It is a cloud infrastructure using platforms using, uh, automated pipelines. I was actually talking with some of our best DevOps engineer and they were sharing with me this concept of uh, The super machine. So the super machine from quote, we develop and we and we deploy already all the components that are needed for training a new model or to say bringing it to production.
So all those stages that are needed. There is a super machine already creating the containers. So that is what I mean by the building the scalable infrastructure for [00:27:00] long term success. It's all the checks and balances are built in in the development process are built in all life cycle model development process.
And once something comes. At the end, we know it is safe to use. That's what I mean by that. So it's almost, without wanting to oversimplify, but it's almost the next step of things like infrastructure as code, policy as code, that if you've got the building blocks of, The system build you're trying to do already established, already understood that you can then run an automated gen AI driven process through the middle of that leveraging componentry that already exists in a safe way.
Is that the gist? Yes, and it is not, it is not a terribly new concept. It's just that we now have the technology. Yeah, to fully automate. To implement it live, real time. This is the concept of what, um, yeah, compliance by design, you could call it, or risk management by design. Assurance in a box. Right. I like that.[00:28:00]
It's the inner box term. But it is, isn't it? You're, you're, you're going, oh, I want to assure that everything's safe and ethical and good to use. Boop. There it all is. I was going to go down the hyper automation route off that point. You went to assurance. Well, there is that. But the point, I think the point was that you can assure that your automation is good.
So because your automation is doing a lot more than mechanical transactional. I think it is a joke, but it is an important part of it. And the automation potential that we've been talking about for the last 10 minutes here, when you start to join those dots off, like hyper automation, where you're thinking 60, 70, 80 percent of existing processes are potentially automatable in future.
It's a striking number. And it goes back to reshaping the workforce. Yeah. So what does a company look like after you've done that deep level of change? And, you know, that's a long term thing because it's like hard to change human organizations into automated ones at that scale. That's why I think this is definitely a C suite discussion.
You were talking about a vision, you know, an inspirational [00:29:00] vision of how are these agents gonna, you know, do everything, or I even hear about startups that go for agent first. Yeah. I can imagine. And then if the agent is not able to, we're going to put a human in there as well. So that's, that's a completely different way of looking at things.
That is different. Yes. So, so what we notice right now are these investments. And then we also notice frustration coming up because we are not materializing the benefits to the extent that we wish to materialize. And it's the note that triggered me is looking at it differently or changing the perspective.
Yeah. Okay. Let's remember how we are looking to implement this tech, how we're looking to implement these models. So we are looking to implement them mainly in traditional businesses with system legacy, with maybe data challenges. 'cause of course, the better the data, the, the better the we can apply this technology.
So we, and that's where we are now. Getting stuck. And it makes a lot of sense, right? So [00:30:00] the new perspective is, um, we were actually having a discussion with, um, uh, leadership teams in this space is imagine. So the question that C level executives are asked to think through maybe, Imagine that you would start doing your business model from scratch, absolutely from scratch.
You define the data input, you define how that comes into the process flow, how close to the customer you get with the intelligence. And when we look from this perspective, We might notice that some of the businesses of today could be indeed aging businesses. And there's been also discussions about the next billion dollar companies being run by one or being set up by one single founder.
using AI and, and being able to compete with bigger corporations. And again, this is, this thought was triggered about the remark about a [00:31:00] different perspective. That scaling an organization angle through agents is actually a bit mind blowing. I hadn't thought about that quite like that. That is, that's a, yeah, it's a sensational point.
Can you imagine the ease at which you could scale a business when you're triggering additional agents against demand, for example. There's always that thing, isn't there, where in very complex businesses, say tax authorities or something like that, where the rules, you know, you print them out, there's thousands of pages.
There's often this bottleneck about, you have to funnel in very complex cases to a very small number of people and they take ages to analyze. If you can capture their intelligence to a, A good level, then you can scale expert capability within your organization. And it's always been a thing that these types of organizations know they have to do and know they want to do.
And it feels like the technology is now arriving where that's becoming a reality. And so you don't get the brain drain on your [00:32:00] organization because obviously all the expertise is held in the people who've been there the longest on the demographic curve. They're the furthest and if you get, you know, a retirement situation goes on, what have you lost?
So as I say, how do we capture that and then how do we deploy it and scale it? And that's another angle for businesses to say, I can keep intelligence in my organization and I don't have to rely so much on the human to get me out of trouble when something very complicated. Could be saying like, could be saying, you know, I don't know, X years in the future.
I'm running, I'm running 14 footsie 50 businesses on my iPhone today. What an amazing part. Well, you could run 30 if it was Android, Dave. So almost, if you look at it from this perspective, it's almost a generational, one generation to go into this. space also because our, uh, the new, uh, the newcomers are just used to interact with the phone.
It's a very, very different system of interaction. Uh, and also if, if you, if I listen also sometimes into [00:33:00] other, the experiences of entrepreneurs or smaller companies, and they do notice the use cases isolated, uh, within the company, maybe in different silos. And then maybe there is another fourth use case that might connect the three use cases existing.
And then. That might lead to, you know, 30% efficiency increase at the business level, but then those decisions are really big level decisions or management board decisions. And they're tough to, to, to, to take because yeah, those, those members themselves are involved in those activities. So that's one thing.
The other thing at one point we will, so some decisions, some more tougher decisions will be. will be needed. Um, and they are driven, uh, for instance, I see them coming faster for businesses that are interacting with consumers because consumers, they will get the intelligence in their, in their phones, uh, quite soon.
Um, so they will have the ability to do extensive, um, say scrutiny of services and [00:34:00] decide based on a lot of information, just using their personal agents. So everybody would have this personal agent. So in consumer, um, focused industries. I see that being accelerated further to stay competitive, a little bit less in, in call cash business type of business models, because there's no rush there, right?
Um, especially business to business, they will be a little bit slower with, uh, with moving into This more strategic decision, uh, decisioning making. So yeah, it is, it is something for the future. The big question indeed that, that remains, and I think we started that a little bit, is well being. And I think that will be, in itself, a topic that, that will need a lot of attention.
And I, I fully, yeah, have confidence that this technology will help in that space as well. So as we touched on it a little bit in the main [00:35:00] conversation there, this notion of layered agents, and I know that's something you've been looking at and exposed to even actually in the course of the last couple of weeks. So what you've been looking at? Yeah, definitely. Yeah, I think we've touched upon quite some topics in the in the previous conversation talking about a trust layer when we're talking about Salesforce, because like a couple of weeks ago, I was at Dreamforce, and I'm still using the same demo that they showed during the keynote.
And I think that's quite impressive, because that's a demo that really showcases AI assistance Virtual agents, one on one connections, communicating with end users in a fantastic use case, uh, for Sax, which is a huge, uh, retail company, obviously. So a lot of people can relate to their own customer journey as well, which is also a very strong point in a demo.
Uh, so what I'm fascinating about, we saw that at the first chatbot already, 1966, I don't know if you know it, one of the first chatbots. was created called [00:36:00] 1966. Yeah, there we go again. Did it have a funny voice? Well, I don't actually know if that's the point that I want to get to. Wait a minute. So Eliza was created like in 1966.
Um, because, uh, and it's 1966 and it's called Eliza. So we humanizing already. a bot back in the days. And it was originally programmed as a, like a parody parody parody to highlight the superficiality of human communication with computers. And I think we've all gone through that flow in the past years that you expected the bots to help you.
And then you came after a few weeks, you came to the conclusion, okay, I need to get through this bot as soon as possible. Give me a human, give me a human, give me a human. That's the, that's the funny twist. though, I I've been there where you go get through the technology fast. I need to talk to a human because they can sort my problem out.
And now it's turning on its head where I go, I don't want to talk to a human because often the technology does [00:37:00] a better job of sorting it out for. Yes. And what you now see, and that's also with the Sophie demo, uh, that, uh, that Salesforce is using, and that's the thing that I'm triggered by or intrigued by is that they use a human face and a human voice.
And we're talking about trust a lot, so where does that leave the bots? Because I also saw in the report, Nicole, that you're talking about augmented software teams, so you have your AI assistant. Is it also going to join sending out GIFs in team chats? You know, is it going to be part of a retrospective?
You'd have like bots that have got like a really good meme game. Yeah. You know, is it going to be that human? Is it going to be like one of us? Do you think it needs to be? Because I find the move to humanize these things a little hokey, but maybe that's just me. Like, I don't, I think I'd almost be happier with, you know, You know, you go on to, you know, whatever.
com you're trying to get a refund for, or, you know, kind of report damage to the item that's been delivered. [00:38:00] If it was just like a super, a super efficient bot that didn't bother with any of the thing, but you knew it was a bot, I'd be like, that's brilliant. That's exactly what I want. I think marketing has a lot to do with it because you also want people to create a connection with your brand and they use the bot to do that.
Depends on the interaction, I think. So where your use case there, Dave, where are you going with this, Rob? Well, well, uh, explore the bounds, but the, uh,
but the, if you want to return your TV, yeah, all you want is an efficient bot that's going to get through it fast. But if you've got a bot that's part of your team and helping you do requirements and everything else, and it's more persistent in your life. Then you probably do want a bit of a personality behind it.
Cause when you interact with it, you might have a bit of fun. Maybe it can tell you cheese jokes and tell you, you should go off and listen to jump by Van Halen. I've already got one of those. I mean, yeah, exactly. Why it's, it's a very human variant. Uh, but yeah, you've got, you've got this point where I say, I think it's the role.
And some roles do, [00:39:00] you do, I would want some personality in my bots. Well actually there's a, there's a study in the Netherlands, or at least there's an elderly home with people that have dementia, and they have a bot, and that plays songs from back in the days, or that it can actually showcase some old pictures to help them.
Where does it get its power from, Esme? I don't know. From the old person. Gosh. You've gone dark fast there, Dave. The human battery point strikes again. No, let me get the case. They videotaped it and they actually saw that the elderly people started to communicate and share experiences with that bot. You know, they didn't feel lonely anymore.
And that's just, you know, that's human connection with a bot. It's the future of healthcare. We will have to get used to having bots in our life. Otherwise, we're not going to have healthcare. So it's coming, isn't it? It is, it is such a wonderful example to observe, because what makes this different, this, this bottle, that is.
engaging in [00:40:00] such a way with other people? What makes a difference from a bot that engages also emotionally in a marketing selling process? And what we notice here is a key thing that it's good to notice. Again, AI is a tool. It's just a very powerful tool. It is the intention behind it that matters. And That's where, yeah, we also need to look in our own psyche and understand our human behavior and how we are being partnered with such a technology can impact the world.
So this comes back also, it's like a full circle towards intent, but also towards software, for software engineering. I do believe also noticing the developments. In Gen. AI, the last two years, it is really astonishing to witness the developments and how fast they go. So I do expect within the coming years, even subject matter experts or non trained professionals, they will be able to deploy.
Powerful models, maybe even end to end with, with a bit of [00:41:00] support. So if we put this technology and this power in the hands of everybody, actually, what becomes important? Yeah, the other thing that occurs to me as you say that though is, is the risks and challenges that are, that are built into that. So I know that you've done some work on that in your, in your report.
What do you, what do you see as the core risk that we should be looking out for beyond? You know, some of the more obvious dystopian flights of fancy, but, um, how do we prepare ourselves for this? Because you're right, the last two years have been epically fast, but I think the next two years might be even more profound.
Yeah, so again, the standard key risks are what we see at this moment is just very pragmatic hallucination, getting wrong answers. We also see there is a risk of over reliance over the outputs, of course, security risks by data leaks or. misusing personal data. So these are kind of standard things. Also what I see more systemic risks that are worthwhile paying attention to, but [00:42:00] because we are, and corporations are definitely geared towards profit and KPIs in this space.
And this is especially in the space of AI. This is where ecosystem wellbeing has to be taken into account as well. It's a very powerful, can be also a manipulative technology. Then you will answer, yeah, there is, in the AI act, there is, there are clear rules against, um, using AI to manipulate, but of course you have to go into the definition of what it is.
So it is all about, I think a big invitation to, especially to leaders is to indeed consider the economic prosperity, wellbeing of their company. in the same time, look at the ecosystem, impact on the ecosystem, and how can we use this technology not only to improve our profits, but also to bring humanity to a next level.
I think that is a big invitation. Very good. Very good. I'm going to bring us to a bit of a close today. Nicole, thank you so much for spending time with us today and sharing your insights. It was a [00:43:00] fascinating conversation that went to some really interesting places. Thank you so much for having me. It's been a great conversation to be part of.
Now we end every episode of this podcast by asking our guests what they're excited about doing next, and that could be going to listen to the new Cure song. I'm dating this episode because it's only just dropped and Rob is excited about it, aren't you Rob? Yeah. You're dating yourself at the same time.
Watch yourself. We're going to have to edit that. Or Nicole, it could be something excellent that's going on in your professional life. So what are you excited about doing next? Well, as we've seen in the AI era, there are some unexpected turns. So even for me, I see here an unexpected turn. I've always been very passionate about execution, deploying these solutions, seeing them making an impact into the world and I've never thought I'd miss. So passionate about risk management. So what's exciting for me next is, uh, shifting my focus to AI and governance and risk management. And yeah, my mission is now to ensure that as we [00:44:00] scale this, the groundbreaking technology, do we keep them safe and sustainable? So if you would like to discuss Van Halen, the cure, or any of these multi layered agents, please let us know, you know, we're on LinkedIn.
We're on X. You can email us even via [e-mail]. We're very eager to So a huge thanks to Nicole, our sound and editing wizards, Ben and Louie, our producer, Marcel, and of course, to all of your listeners. See you in another reality next [00:45:00] week.