Shankar Balakrishnan 0:00:06.3:  

I'm delighted to welcome you all back to this plenary panel on AI potential and reality. You all know me, but we've got a very exciting panel, and a very deliberately chosen panel ahead of us. We've got a very diverse representation among the elite speakers we have. We have representation from a GSI, we've got representation from a hyper-scaler, from a customer, and from a vendor. All of them have one thing in common, their passionate experience in AI. We are going to be having this panel discussion for about 30 to 40 minutes. There will be opportunities for the audience to ask questions, so please do keep the questions coming, and please do keep the session engaging. Thank you. Panelists, thank you all for joining me on stage today and thank you all for being present in this event. I'm going to ask you all to introduce yourselves. Before that, I'm going to also frame the first question, and hopefully, you can introduce yourself and then start answering the question as well.  

 

Shankar Balakrishnan 0:01:11.3: 

If I take myself back to the last 6 or 9 months, and recount some of the conversations with FDs and CFOs I've had, one of the often-heard themes is this, look, 18, 24 months back I had this brilliant initiative proposed to me by someone, who said if I can roll out gen AI, I can realize 20 per cent to 25 per cent efficiencies. First, while we're here, I've got actually more people in my organization because I had to go and hire a bunch of data scientists and Python developers to write these AI models. I've had to pay licenses on top of it. So where is the efficiency? That's a common question we have. How would you respond to this? Maybe Mateus, I'll start with you, quick introduction, and then your opinion. 

 

Mateus Begossi 0:01:58.4: 

Hello, is this on? Good. It's a great question. I'm Mateus Begossi, I'm the Finance Data and AI Lead for EMEA in Accenture. I have about 20 years getting CIO teams and CFO teams to agree on things, and design things and deliver things together in digital finance roadmaps. In terms of the ROI, I think it's there. Okay, it's just a matter of how you achieve it. Accenture so far has north of 3000 gen AI and agentic implementations done so far. We're seeing a typical equation of value in these is around efficiency, yes, okay, headcount reduction, but also process cycle times, productivity, quality, and risk, and even NPS scores, right, from the finest functions through the business. So there are significant metrics that are being achieved. The question, I agree with you, is the approach though because we're finding that when you can easily get to a use case success and a proof point success, that's often not enough to actually release efficiencies or to actually change the process throughput and the speed of the process. 

 

Mateus Begossi 0:03:18.1: 

It's a conversation we've been having, and there is an S curve somewhere where people are experimenting and testing things, but there are people, there are leaders which are far along the S curve, and they have already done that in the previous year, the previous budget cycle, they have invested and discovered this. Now they're baking in more material business cases that actually assume proper end-to-end transformation, multi-dimensional impacts across the organization. So I don't think it's an if, it's a when for most companies on that journey. I would suggest if you're struggling with proof points, that's okay, everyone is, but some people have been through that, so be a fast follower, that would be my position here. Is that okay? 

 

Shankar Balakrishnan 0:04:04.0: 

Yes, great.  

 

Phil Le-Brun 0:04:04.8: 

Shall I go next? Sure. Phil Le-Brun, Amazon Web Services. I have the privilege of working with a team of ex-CXOs. I mentioned this morning I came from McDonald's, after many years. My job, to answer this question, with boards and C-suites, which is how do I use technology? What is technology? What's data? What's generative AI? Whatever it may be. It seems like we're almost answering the wrong question. It's a little bit like if we rewound the clock 120 years or so, someone saying, 'This electricity stuff, what's the ROI?' We're talking about a general-purpose technology here, with machine learning, generative AI, AI, it's permeated every part of the Amazon business because the mindset is quite simply how can we focus more time on the things that matter? I do agree there's so much efficiency in organizations. One study says 80 per cent of people's time is wasted in organizations. Another says, a developer spends one hour a day developing. So there's a massive amount of inefficiency. I don't think the answer is just technology. 

 

Phil Le-Brun 0:05:06.9: 

It's a bit like asking personally, how do I become wealthy? I know, I'm going to stop spending money. I'm going to have a miserable life. I think the organizations that are doing a good job here are looking for those efficiencies, but they're redirecting people to things, to areas that are going to grow their business. The danger we're in at the moment is we released a report in April that said we're living in a two-tier economy in the UK. We've got start-ups that are using AI to fundamentally rethink products and services. We've got enterprises trying to drive efficiencies. Those enterprises need to get back to whatever makes them good, so drive those efficiencies, but think about your customers and what's actually going to drive profitability rather than just savings in an organization. 

 

Dany Krivoshey 0:05:51.1: 

Nice comments, thank you so much. I'm Dany, I'm chief noise officer - sorry - Chief Digital and Transformation Officer, in Unilever. I agree with you. How many finance people here? So I will not do a mistake. Go away. Guys, let us go back slightly. If you ask Steve Jobs what is the success of iPhone when we develop that for three years? People say, no, but there's not profitability on that. Who will doubt that? Who will doubt cloud solutions that we have now? The same happens with AI. What is, for me, a critical question when the CFO is asking, where is my profitability? Let us go back to the problem statement what we had and what exactly the value realization what we want to put inside. Now when people ask me, where is my savings? I'm always coming back to track and fix. Okay, guys, do we have the right adoption? Yes. Do we have 100 per cent data? Okay, fine, great. This is, by itself, already the first type of value realization, but when you have at least six months of change management of people, and managing people. 

 

Dany Krivoshey 0:07:04.7: 

So when we're asking, what is our ROI on investment? We often need to look after our milestones, are we doing that right? Are we designing the right solutions for our customers first? Customers first, are you creating the solutions for them? Then you're looking after are you creating right adoption, are you creating right value proposition for the business, etc.? Now finally, we also need to ensure what the business case itself, and when we're speaking about business case I'm not just speaking about financial results, are fully agreed with the business and owned by the business. Eventually, it's like asking can this bottle deliver value? No, it's people who deliver value. If we agreed on this business case and we're managing these metrics, let us execute. This is where ownership comes to account. It requires a really holistic approach, rather than where is my profitability? To answer on these questions.  

 

Leigh Romeo 0:08:03.9: 

Thanks Dany. Hi everyone, Leigh Romeo, VP, Head of AI at Anaplan. Been at Anaplan for quite a while now. One of the veterans, so over seven years. Seen the platform come a long way, and now the advent of AI. I do have to say, it's very early days. It's like the early telephones right now with AI. We're all learning. We're at the start of that journey and every day our development teams are learning new methods, new ways to go about things. When we talk about the value, you hit on it, how are you measuring this? For me, it comes down to the use cases, being really specific around having small, very focused use cases, so that you can measure those outcomes, those tangible benefits. If we're talking about gen AI because any time someone says AI these days they mean gen AI, but there's other flavors. Right? Machine learning has been around for quite a while. Optimization. Linear programming, that kind of thing. With gen AI, can you measure how much time a finance analyst has saved by being able to create a report? That is an ask we get every day from finance teams, can you just create a report for us? I don't want to spend hours creating a report for each [?GO 0:09:32.7] that I have. It's going to take me days to do this.  

 

Leigh Romeo 0:09:36.2: 

On the machine learning side, for example, how much time could you save with your finance team just creating that report using an automated method, using algorithms, versus the old method you used to use with five of you, over a month, taking nearly 100 hours? I think it evolves. I think the use cases are quite interesting because, sure, you can have some quick wins with small use cases, but as you get into large enterprise, those use cases get much bigger, and they do need to be a lot more complex to really move the needle when it comes to enterprise.  

 

Shankar Balakrishnan 0:10:15.1: 

Interesting, look, very good perspectives. I'm probably going to dig deeper on a couple of the points you mentioned. It's clear maybe the industry needs to learn to think about, the businesses need to learn to think about AI differently. Let's turn this around. Where are instances where you've seen AI make a significant impact on the P&L, on the likes of employees, on the productivity metrics? What are some of the good examples you've seen? More importantly, what made them successful versus other pilots? What is the secret [?sauce/source 0:10:49.7] to making this work? Phil, I'll let you go first, and then Mateus. 

 

Phil Le-Brun 0:10:53.8: 

Sure, I think Dany started with one of the points, which is solve real problems. It's the old skit about generative AI is the answer, now what was your problem? Those organizations who are doing that are really understanding what their customers want. Travel and hospitality, reinventing how I book a holiday. Natural language, I've got two kids, my wife likes to do this. Yes, how does a customer want to interact? Don't reinvent a crappy chatbot. The manufacturing industry using it to generate synthetic data for testing products, for instance. Consumer packaged goods. Fashion. There's good use cases where they've all fallen in love with the customer problem. Amazon itself, we've been using machine learning for years, across everything from forecasting to logistics to running robotics within a fulfilment center. Again, it all starts with let's figure out what problem do we need to solve?  

 

Phil Le-Brun 0:11:48.8: 

I think one of the anti-patterns, if I took your example, Leigh, is what I see organizations often do is automate things that shouldn't exist. So we have 100 reports today, let's generate 100 reports no one reads even faster. Digging deep into it - I lived through this - my team used to generate 7000 unique reports at McDonald's a year, and we were told we weren't fast enough. When we actually democratized it, when we gave the ability to generate this to the people who were meant to be using the reports, the number of reports dropped significantly. So it's also tying accountability for the use of the tool to the request to, so breaking down some of these 19th-century silos we so like in our organizations.  

 

Mateus Begossi 0:12:32.7: 

I think that's great. I've been working with a few clients more closely, especially in life sciences and consumer business, right, so for an example, right, and always the finance team. We're looking at when it's properly done, and I'll tell you what properly looks like, but when it's properly done, you're starting to really see that, which is things like, yes, 30 per cent to 50 per cent FT potential reduction somewhere, right, but that's fine. We're looking at things like 95 per cent controls automation. We're looking at working a close on workday three. We're looking at forecasts done on a cycle time in nine days. Okay, we're looking at 95 per cent touchless invoice. It's actually 100 per cent, but they don't want to declare 100 per cent. There's always a reason, right? So there's really tangible impacts when it's done well, when the conditions are met. The challenge is to get to those conditions to actually do this well. I think it starts with proof points, as Leigh was saying, right, proof points and experimentation. There's an institutional maturity that comes with that.  

 

Mateus Begossi 0:13:37.1: 

What we are finding in common with those companies, which are now banking on these bigger business cases, and they're really big business cases, is things like a history of digital literacy in the business. A cultural readiness to adopt and experiment, and to work within guardrails of these new technologies. We see that. They began early and they were deliberate about it. It's not about giving a chatbot to everyone and let them figure it out. It's actually thinking what is the best way to build from this type of thing? So have a strategy early on. Finally, they had a modern data landscape. They had made investments in modernizing their backbone data. They had a significant mature data engineering and data management practice in place. So these are conditions that are making the leaders today. You don't have to be a leader. Right? Being a fast follower is pretty good because then you see what works, what doesn't, but that's common across those. Right? Is that fair?  

 

Dany Krivoshey 0:14:40.4: 

For me it's fair, absolutely. You need to ask the audience. Guys, are you tired, by the way? You're sitting here forever, I'm just sitting here 15 minutes with chairs. Why I'm asking, what I just told, what I just did is the companies which are really successful, they deeply care about people and consumers and customers. This is very clear, Leigh, this is, the [unclear words 0:15:07.2] is if you really deeply care about the problems of your consumers and opportunities, this is the only focus you need to do. The second thing is absolute change in mindset of how organization runs. It's developing digital talent and digital literacy, which then moves to how we collaborate together to make solutions work. It's a cultural shift again, which when absolute governance over the data, this is where you're making data as a discipline. These are companies which are really successful here. Finally, disproportional change management investment. Now I can give you some use cases. I discussed about that already.  

 

Dany Krivoshey 0:15:49.5:  

We have thousands of customers around the world. How much time does it take us to respond to the customer query today? Around the world, yes, B2B customers, less than one minute. When customers ask a question, there is an agent which is replying to the customer by taking the data which is one source of truth, and instantly replying to the customer within less than one minute. Something what was done three weeks previously. So have reduced the number of people supporting the customers? No, we actually increased the number of people because we want them to do something more valuable. So but we see what the difference to a customer experience is absolutely different. I'll give you another example. When customer places an order in Unilever International, it's usually around 60 containers from 4 to 5 different sourcing units around the world. Try to manage the order. Guys, one order is at least five different pages of Excels, which you need to split in and you need to find which sourcing units to do. We process an order today within one minute because we have OCR and we have agentic OCR and everything is driven by the data. So once a customer places an order, it's one minute in the system. This is when we see really the power of AI.  

 

Dany Krivoshey 0:17:07.9: 

The approach eventually is coming, are you customer-centric to understand the problem? Do you have digital intelligence team which is focusing to solve engineering solutions? Do you have the right focus on your data which you're doing? How much you invest in change management across your organization? Once you're building these enablers, then the success is coming.  

 

Leigh Romeo 0:17:35.1: 

Thanks, Dany. I guess a few specific examples around where you would see that value. Some examples where you've got that cultural piece as well, Dany, and maybe some of the ones you can't really measure quantifiably. That would be like this morning we demo'd model-building, and there were a few responses, 'Oh my gosh, you put me out of a job.' It's like, 'No, we're going to make you much, much happier.' You don't want to spend two or three days building out a custom time module or a massive building materials for supply chain, a big, laborious task, when you could get the technology to help you. That is tangible, but it also makes you a much happier model-builder. That's one side of it. One really defined outcome, a manufacturing customer of ours actually using optimization within their manufacturing. They were cutting glass. They figured out the most optimal way to cut this glass, within the size of the sheets that they had to, actually reduce wastes, reduce those costs, and have a better top line. Those specific examples around more mature, I suppose, processes like Mateus was mentioning, versus some of the newer gen AI type use cases where sometimes they might not be so quantifiable.  

 

Shankar Balakrishnan 0:19:17.8: 

Look, that's great. What I'm taking away from - sorry, Mateus, did you have a point? There are really good, successful deployments, and there are really good use cases being found. One of the conversations which I had in the breaks from most people in the audience, especially after your session, Dany, was this was excellent, I'm struggling with the same problem, how do I get buy-in from my management? How do I do this? In your experience, in light of these recent, successful examples, what do you think is the way executives should consider or probably do consider an AI investment? Is it an iron-clad business case, or is it just experiment with it and come back when you have good results? What have you seen work and what do you think should it be? 

 

Dany Krivoshey 0:20:03.2: 

One is hire me as a consultant. No, actually, yes. Guys, it's not about that, it's how I make this happen. Do you think it's easy, what you have shown there? By the way, I'm not a technology guy. I'm a finance person. I was a CFO before, so I know the pain of FP&A and the reporting and on time delivery, I know that very well. It took us five years. How to gain there, people don't trust us to do the changes. No one believes in magic. The only way people start trusting you, if you start showing some results. So you start small, and you're building trust, and people suddenly see the value, and when you're focusing on pain points, what you suddenly resolve, people, 'Actually, it works. Can we do something else?' It took us five years. Currently, the situation is I need to fight back because the amount of demand we had is just huge, it's just piñata everywhere, all these prices coming down. Previously, we were fighting to do one case, but we did it gradually.  

 

 

Dany Krivoshey 0:21:16.8: 

One of the key points here is, one, understand who is your stakeholders, What is their problem to solve? Is this visionary stakeholder, or this is controlling stakeholder? If I'm speaking to CFO, likely it will be a controlling person. What controlling needs? Confidence, right, how you're reducing risk. You need to speak his language. If someone is visionary, like our CEO, Aseem Puri, one of the most amazing CEOs in the world, which really build the business, very visionary, you need to show him how you're going to grow his business. What I'm speaking about here is not about technology success, it's about intelligence of understanding human being you work and live. The biggest unlock of our team was creating empathy environment and educating our team members digital people how to read stakeholders and how to manage them. It's two years' program, guys. It's not PowerPoint, this is how you deal with stakeholders, thank you. No, it's two years of mentoring for our people to get into the mind and understand how to deal with different people. How to negotiate. How to understand what the narrative, what is important for them. If you start from this angle, not from technology, who are you as a person? What's your problem today? How I can help you. What is your risks or what are the things, what ambitions what I can help you with? This is the starting point for that.  

 

Dany Krivoshey 0:22:51.8: 

The second is small success stories, which are building up. Once you're building up, you will see what momentum is rising. Continue doing that. Experiment more, and then you will see, suddenly, more and more people are coming. The third success point here, make HR your best friend. I don't like when people say HR, human resources. Resources usually you consume. Right? Talent development. Change it. It's talent development professional. Once your HR becomes your friend, and you're focusing on talent development, and putting that as part of an agenda of people development in organization, and you're making this part of a targets and making this part of a development cycle, suddenly more and more people are joining you. The fourth one is your perseverance. Just push for this to happen. It's easy to raise your hands. It's much more difficult to continue and believe in what you're doing. If you really believe in that, you will be successful. Sorry, it was a long answer.  

 

Shankar Balakrishnan 0:24:01.0: 

No, it's a good answer. Phil, Mateus? 

 

Phil Le-Brun 0:24:07.6: 

You summed it up well, Dany. I think a lot of this is about psychology, how do you bring people on the journey? There's a lot of fear around this. Is it going to replace my job? We do a really bad job in most organizations answering the most fundamental question, what does it mean for me? No one actually cares about your transformation, your opportunity to increase your return on incremental capital invested unless people can answer, do I still have a job? So that change management piece, I look back on my career, and I was more a psychologist than a technologist, how do you bring people along with you? Skills is critical, and I'd start at the C-suite as well. Out of curiosity, how many of you dislike people and don't understand money? Of course, it's ridiculous. You don't say I've got a CFO and a chief HR officer. I'm not going to ask you the same question about technology and data, but the reaction we get it is very different. The average C-suite, outside of the IT function at the C-suite level, less than 24 per cent of executives have enough technology knowledge to do their job. Not to do my job, to do their job. At the board of directors, 6 per cent to 13 per cent of board director members have enough technology knowledge to do their job.  

 

Phil Le-Brun 0:25:14.3: 

So reskilling, this is about shadow of the leader, reskilling starts at the very top. If your CEO hasn't written a prompt, and yet is declaring they're going to be a generative AI company, something's a bit off. We know already across just EMEA, across Europe, Middle East, and Africa, the biggest single barrier to the adoption of technology, digital, generative AI, whatever it may be, is simply skill sets, from the board on down. So you're turning yourself into a learning organization because everyone's after the unicorn, everyone wants the data scientist who's transformed two global organizations with a GPA of four for £60,000. How do you actually embrace the fact…? We talk about agile a lot and such like. Agile means learning. What works, what doesn't work. The rapid experimentation. Lowering the cost of learning. Having that mindset that you're not going to write a five-year project plan around the use of generative AI, but you're actually going to figure out what problems to solve. Then you're going to figure out iteratively how to solve them, giving your people the skills necessary to do that while you're on your journey.  

 

Mateus Begossi 0:26:21.1: 

What was the question again?! I think to modernize, to achieve this, right, I think - amazing points, by the way, so just building on that. We agreed not to build on things. What we're finding is, start with value. Okay, chase value. At the beginning, value is taking risk away from something. At the beginning, maybe value is streamlining an operational process. Right? So compliance, operations transformation. That's a CFO dogma. Start with value, chase value. If you really want to get to a point where you have no significant value unlocked, right, measurable, you need to have an end-to-end process thinking. We call that the art of reinvention. Right? You reinvent the process. Think from the right to the left. So think about the outcomes and what good looks like. Bring together, right, very deep process expertise. You cannot ignore that. Bring very deep, pragmatic AI knowledge, and capabilities and confidence, which you've built by doing the experiments, etc., or you hire, but the point being you need to have an end-to-end process reinvention mindset to be able to actually unlock a lot of value. 

 

Mateus Begossi 0:27:38.6: 

You don't start from that on day one. You do need to go through the process of building internal confidence and shoring up your enablers in data and change. As you're thinking that end-to-end process philosophy, you also consider all the dimensions of the problem, not just the technology play, but also the process, the risk, the compliance, but also the people side of it. So I think that's the recipe for us so far for the biggest successes has been that one, yes.  

 

Leigh Romeo 0:28:09.3: 

I was just going to add, where do you start? Well, don't boil the ocean. I mean don't bite off more than you can chew. Don't go out with a solution then look for the problem. Always start with what is the problem that we're trying to solve? Even if it's a small thing, that can have a big impact. One of those ones I've been thinking about as you've been talking is change management. If you've got a new application or something that your end users need to learn, is there a way that you could use this technology to help them learn, rather than say, here's the documentation, or go away and use the training course or whatnot? Remember, Dany, we had this discussion a couple of months back, you said, 'I just want a video on my dashboard, so my users…' I said, 'What's wrong?' 'Well, they just don't know what to do. So it would be great if I could just have a video up there and we can teach them exactly what's on the page and how to perform their task.'  

 

Leigh Romeo 0:29:14.6: 

Well, that, for me, is a really great way to start. All right? Really small use case, you're looking at that change management element, you've got an automated way to show people how to use something. Then you're moving away from having to create reams and reams and documentation that will probably go out of date very quickly. Then again, you've got quantifiable success, the users are going through whatever process, and hopefully, you're making them a lot happier as well.  

 

Shankar Balakrishnan 0:29:45.4: 

I'm going to invite audience questions in a minute. I'm going to ask the panelists one more question, so please, people who are helping with mikes, with the running mikes, please be ready. Phil, one question to you, you mentioned that there's this [?AW 0:29:57.5] support which points to a two-speed economy in the UK, where the smaller organizations are innovating and experimenting, whereas the bigger organizations are focused on efficiency. That's pretty telling and insightful. First of all, why do you think that is? Second of all, given the challenges of productivity in the UK as a nation, which we face, what is the implication if we don't fix this two-speed economy? Where do you see this going? 

 

Phil Le-Brun 0:30:23.8: 

Well, given how much of the GDP comes out of large enterprises in the UK, it's pretty significant. We've got a talent gap. We've got a productivity gap. The two are linked. We've got the opportunity to embrace this technology, to reinvent roles, which is going to require a lot of change management. If I look back on my career, every role I've taken where we've applied technology has got rid of undifferentiated work. It's made me happier and it's made me do more productive things. I think it's essential we address this. Otherwise, those start-ups will be the next enterprises because enterprises cannot possibly save their way to success. There's this massive opportunity, and I think we as employers need to really embrace this, give people the skills, recognize we're training machines as organizations, and really think differently, think big about what the opportunity is here. I agree with Leigh, let's not boil the ocean, but small thinking is a self-fulfilling prophecy. If you look for a ten per cent saving in a process or a P&L, you're going to find it. Try with ten times. How do I make my profitability ten times bigger? My throughput ten times bigger? You have to fundamentally change your thinking. That's what we're talking about in organizations here, it's about changing our mindset from 19th century ways of working, with 21st century technology, to new ways of working. Breaking down the silos. Recognizing that those silos we have are in the way of how work really gets done. That work gets done despite our organizational structure often, not because of it.  

 

Shankar Balakrishnan 0:32:00.8: 

Mateus, possibly, to you, the last question, as you work with other UKNI enterprises, which is a big portion of your job, are you noticing these bigger organizations being more risk-averse, and hence the fear of losing market share and business to the lower organizations? If yes, what do you think is going to change? Especially as we look at the public sector, where there's a massive need for a step up in productivity, what do you think is going to give? 

 

Mateus Begossi 0:32:31.2: 

I don't think people are afraid that much. I don't think the strategic landscape - perhaps disagreeing with you, Phil, but it's fine - I think there's always been the risk of start-ups disrupting business model. I do think it's becoming easier now for something to scale up very quickly and become threatening. On the other hand, I think large organizations do have the engine room, they have institutional momentum to adjust. Right? It's just an operation model challenge perhaps, more than a sizing problem. Are they agile enough? Are they governed, is the board switched on to this? Do they give them enough leash to experiment and play? One thing I notice is companies that tend to move from year to year, experience like the budget, has it been budgeted for next year, can we do it? I think these struggle a bit. I think some agility and resource allocation is probably going to be the name of the game for large corporates. Right? I think the start-up problem has always been there, they just scale faster. Then again, you can also do that. Okay. I think this space of AI, the amount of investment going into this space is massive, there are galactic levels of investment going into this thing. So much value being packed into a small space, and everyone is trying to work through the problems that, as you've seen, the fact it's a non-deterministic type of technology. For finance profession, that's not okay. So how do you overcome that? You can't.  

 

Mateus Begossi 0:34:03.2: 

The problem-solving is happening. We start to see these breakthroughs when this stuff is going to become present in the platforms we love and use, like yours. So it becomes easier and less risky to use. There will be a commoditization of this. We start to see breakthroughs. I think there where you see the scale, right, becoming really measurable in large corporates, which can work with things which are mature enough. I think there's a period now ahead of us which is very stressful and competitive and experimental and risky. I think if we're aware of that, we can just pull through that. Does that make…? 

 

Shankar Balakrishnan 0:34:38.3: 

Makes complete sense.