Podcast transcript: Why artificial intelligence needs C-suite support

47 min approx | 13 July 2021

EY and Microsoft logo

Simon Hobbs  

This is the EY podcast, Artificial intelligence and the C suite, ideas for leadership from EY, offering clients the expertise they need to harness leading AI services from Microsoft. I'm your host, Simon Hobbs in California. Make no mistake, the industrialization of artificial intelligence is now fully underway. AI chatbots recognise our voices, our faces and mimic our human interaction. AI powers drug discovery, assists in surgeries and coordinates complex manufacturing. Not only does AI construct your social media timeline, but it also seeks to police it and our streets as well. But as businesses invest billions of dollars, the divide is widening between AI winners, and those who still struggle. You're about to meet two leaders who specialise in implementing AI. They believe to win, artificial intelligence must focus on an organisation's big problems, not be buried away in some corner. They believe to win, C suites must give AI projects their full leadership support to build at scale and to build rapidly. Before we go any further, let me be explicit, conversations during EY podcasts should not be relied upon as accounting, tax, legal, investment, nor other professional advice, listeners must consult their own advisors. Joining me now from Seattle is Microsoft's Global Head of artificial intelligence solutions, Stela Solar. Stela, I am so glad that you were able to make it. Thank you. 

Stela Solar  

It's a pleasure to be here. 

Hobbs  

Also joining us from Denver, Colorado, EY's data and AI leader for the Americas. Importantly, from within EY's new Microsoft Services Group, Hugh Burgin. Hugh, welcome. So the Big Four meets the Big Five?

Hugh Burgin  

 That's right. Thanks, Simon, for having me. 

Hobbs  

Yeah, indeed, Stela why don't you kick us off with your assessment of where we are now on the evolution of artificial intelligence? Because, I mean, clearly, AI is real, it's no longer just a concept, no longer a black box. What does AI mean to business as we have it at the moment?

Solar  

AI is very real, you know, as computational power has increased, so has the ability to build richer, more complex, more accurate AI models, you know, models, which can predict things that are much greater accuracy, forecasts, estimations, health care, diagnostics, predictive maintenance, all of these are able to be predicted at just much more higher accuracy. And then over the recent years, there have been some breakthroughs where these models have become so specialised that and so powerful that they've been reaching human parity in areas such as object recognition, or reading comprehension, language understanding, and you can imagine all of these unlock new use cases for organisations across a wide spectrum of what they do and the value they add. 

Hobbs  

When you say human parity. Are they actually as good as humans? 

Solar  

Well, you know, this is a big debate for questions on what human parity means. If you think about human parity, AI also makes mistakes. And so it's about thinking, how does AI perform at a level that is as good as the human? And that does include some errors. And so we're always monitoring the models, how they perform and continually evolving.

Hobbs  

Hugh can you give us some examples of where you see AI working really well, at the moment?

Burgin  

Yeah, I guess there's two points I would make Simon. One is, it's everywhere, every sector, every function, I've seen success from AI, whether it's retail, or whether it's agribusiness, or whether it's telecommunications, that AI is successful, whether it's the HR department trying to improve employee performance, or whether it's the marketing department trying to drive more growth, or an operations group focused on throughput and output efficiency. AI is successful everywhere. But my second point is, it's not always successful. Not every company is doing everything that they'd like to do around AI, not every company is seeing the ROI that they could achieve. So while we have seen it successful all over the marketplace, there are many opportunities for it to have a bigger impact and hopefully we can talk about that today.

Hobbs  

Absolutely, absolutely, that in a sense is the purpose of the podcast. But before we go any further I do want to just get a vision of some of these AI winners that are around us at the moment. Stela what would you pick out?

Solar  

You know, when we think of organisations with AI who are adopting AI, we see three distinct profiles. They're not very creative names, but we see beginners, intermediate, and leaders in AI. And what we're seeing is actually anyone could be a winner when it comes to AI. And if we look at the learnings from the leading companies adopting AI today, it's a data-driven culture. It is data-driven decision making. And it is a huge focus on skilling in the area of AI and data. And so I think these three areas of learning are super important to understand how anyone could become a winner when it comes to AI.

Hobbs  

Stela, we're obviously going to talk about Microsoft in particular moving forward. And the service that your mega cloud Azure offers to multinationals is obviously extremely important. We'll talk about all the big stuff. But first of all, can I ask you a silly question, because I'm fascinated how your state-of-the-art AI manifests in your workplace. I mean, do you work in a futuristic environment, voice commanding the tech around you? Or are we really talking more here about software subroutines that are kind of, I don't know, crawling inconspicuously through big datasets.

Solar  

You know, I speak with a lot of customers and those who are new to AI, quite often Skynet will come up in the first couple of minutes. So it's interesting to understand what AI has in terms of a foothold in an individual's mind. Now, when I think about what the reality of what AI looks like, in the Microsoft workplace, you know, right now you and I connecting through a video interface, and you can see the background behind me that is AI, understanding what the person in front of the background is, and what is the background, and then substituting the background into this beautiful sailing experience when I was in Hawaii just two weeks ago, but also, I'm getting continual tips on how I can do things better. I'm a terrible PowerPoint designer. So my slides are never as nice as I would like them. And so I get the designer popping up in PowerPoint, recommending layout to me. I've also presented just on Friday, last week, to a Spanish-speaking audience that did not speak English. And so I was doing live translation as I was presenting. And obviously, we have AI and machine learning that is embedded in the core of how we operate our cloud, you know, maintaining governance, resiliency, load balancing, and most importantly, monitoring for any security threats. If I think about probably the most highlight or exciting example of how AI came up for me in the workplace, once back in the physical world, I walked into a building, and there was a robot there to greet me at the elevator. And you tell the robot who you're looking for. So let's say I asked the robot, where is Simon Hobbs's office, and then the robot will actually tell me where in this complex building I needed to go to actually find you, Simon. So I really enjoyed that example of AI and how it shows up.

Hobbs  

That's really very cool. Um, you said something in the middle of that, that I just want to pick up on? Because oftentimes, and Hugh does the same thing, from the conversations I've had when you mentioned AI, you often mentioned machine learning alongside it. What is the distinction for those of us that are basically lay people here?

Solar  

Yeah, at the core of AI is machine learning. So if we think of machine learning, it is the ability to digest large volume of data, and build models to predict or do something with this data. Now, it is absolutely foundational, this machine learning is absolutely foundational to AI. Because when we think of AI, it is really a specialised machine learning model. For example, object recognition is a cognitive service in AI. Really what it is, under the covers is a highly specialised machine learning model that is very good at object recognition. And so machine learning is that core foundation of what really builds richer and richer AI models that become more specialised on top.

Hobbs  

And Hugh, would you frame the understanding in the same way?

Burgin  

I would absolutely. I think it's important to put some context around how we're talking about AI and thinking about AI. There's a few different types. Almost all of the examples we'll talk about today are what we call narrow artificial intelligence, meaning that a specific business problem we're trying to solve for a specific value for a business or an individual. And there's other types of AI, there's AI that tries to be as intelligent as the human brain, there's AI that tries to be more intelligent than the human brain. The examples that Stela just shared are great examples of a very specific type of AI to solve a very specific and valuable business problem. And there's many examples of that. We think of machine learning as being one part of the AI building blocks, if you will. So you can think of it as a combination of big data, machine learning, sometimes a functional capability, like the ability to read or the ability to listen. And then the ability to take an action or drive a decision. And when you put all of these things together into one process in an automated way, that's really a very specific AI implementation, solving problems for our clients and for people.

Hobbs  

And you know Hugh we should point out that you guys at EY, are a first mover here, you're the first of the Big Four to set up teams dedicated to helping their clients develop the capabilities that they need to harness the AI that Stela is offering over at Microsoft. I mean, fundamentally, how do you see the business imperative here, I saw a survey in which three out of four executives believe that if they don't scale AI in the next five years, they risk going out of business entirely.

Burgin  

AI is a competitive differentiator. You can think of it as a way to optimise business performance, whether you're optimising your inventory levels, or you're optimising your sales, to get the most ROI from your sales effort. It's really a way to compete more effectively. And Microsoft is a leader in cloud for our clients, and a leader in AI for our clients. And so our clients are asking us, hey, I need somebody who specialised on not just AI, and not just data, but AI and data with Microsoft. And so they're looking for a partner who can bring that combination of broad AI and data skills combined with a specialised focus on Microsoft technology.

Hobbs  

And you're very clear that organisations that are struggling with AI, probably need to think about tackling it differently. And that I think, is against a backdrop of many believing that we're at an inflection point, but the initial hype surrounding AI, has now given away to a much stronger focus on executing, Hugh.

Burgin  

Well, AI is not about insights. It's not about a better dashboard, or a better report, it's about driving an action or driving a decision. And so in order for companies to get the most value out of AI, they have to figure out how are we going to actually operate differently, how are we going to actually make decisions differently and embrace the insights that it creates to actually manage our business in a better way.

Hobbs  

And Stela I think you make the observation that in the past, cool stuff was too often developed by people way down an organisation's totem pole, almost like pet IT projects in basements. And those employees, often they lack the political patronage, they need to get those innovations adopted by decision makers at the top of the organisation, and that was a systemic problem, as far as you were concerned.

Solar  

We definitely see that organisations leading the AI adoption, they are implementing it across the core business functions, they're rolling out AI into full production. And this has been enabled by very strong leadership, sponsorship, leadership engagement. And what this has resulted in is that these AI projects, because leadership is involved, they become much more tightly aligned to business goals, business priorities. And so that then tends to deliver the kind of results that start to create this spiralling effect where the AI projects are aligned to business outcomes and goals, they end up achieving those goals, because that was the objective from the start. And then the business case to continue, rollout becomes even more powerful. In fact, we have done a lot of work with organisation to understand what really differentiates those who are winning with AI. And, you know, I spoke about data culture, I spoke about data capabilities, skills, decision making, the results are very real. The results are happier customers who do, by and large, see that organisations implementing AI do have happier customers higher customer satisfaction. They're also having higher rates of innovation in terms of new product, new services that are being launched, and much higher employee engagement. In fact, buyout survey from you know, speaking with that 10,000, over 10,000 individuals across organisations, 92% of employees want to know more about AI. And so if an organisation is focused on AI, making it available for everyone, and ensuring that it is a key critical strategic effort driven from leadership that also has flow on retention, and employee satisfaction implications as well.

Hobbs  

Yeah, yeah, and let's talk about the absolute need that both of you express for the C suite to lead on AI – full sponsorship and internal promotion – so AI gets the attention that they need to succeed, you know. Stela as you say, that starts with the resources, it's the people it's the technology, it's the love, if you like,

Solar  

The love, the people, technology. I love how you structured that, and the pain. I do want to bring up the pain that in the current environment in particular, you know, we always, at any age have competitive pain, business performance pain. But over the last year or so, you know, as we have been in the context of the pandemic, the pain has been even greater. The data is showing signals that the organisation hasn't seen before, the forecasting models have suddenly had to be rebuilt. And so it is it is super critical to again, come to the core of it, of aligning the AI work to the core business objectives. And being really agile with it. sponsorship from leadership allows for that agility, allows for the investment. And we're seeing many organisations actually use AI today to respond to this pain in the market and in the world, by being able to build new services, being able to understand models and what the data is saying in a much higher way. But all of this really is paved by a strong leader at the front or leadership group at the front, that is lighting and spotlighting what technology can do to help an organisation be agile and thrive in the current environment.

Hobbs  

We will come on and talk about this pain aspect Stela, I think it's really important. But I just want to, I just want to emphasise the need for this C suite buy-in to maintain the traction. Hugh you know, projecting enough political capital through the organisation that there's a patience for the time that it inevitably takes for a significant return on investment. You are explicit, there will be speed bumps Hugh, AI development isn't easy.

Burgin  

Yes. And that's why the C suite leadership is so important. Number one, I think there's a few things to think about. Number one, AI is proven, as we talked about earlier, companies are being successful implementing AI, they are creating value through AI. Multi hundreds of millions of dollars, sometimes even billions of dollars of value. And so number one, it's proven. Number two, it's not easy, to Stela's point, there is a lot of pain, it is not a straight path from beginning an AI project to achieving value, it's about the journey that you're going to go on. And each step along the way will not always be successful. It's about having that leadership commitment to the journey, so that you can, number three, drive towards those actions I talked about earlier. How do you drive to better decisions? How do you drive to better actions as a result of the AI. And so the C suite sponsorship is so important, because number one, you have to be competitive. Number two, it's not always going to be easy. So you need that sponsorship to keep the momentum. And number three, in order to really embrace the value of AI, you need to enable it to drive those decisions and actions. And that requires sponsorship from the top, otherwise companies tend to kind of revert back to the way they've always done things. And that's not always going to generate the value that you're looking for.

Hobbs  

Yeah, I mean, Stela it's not like I guess you install software and it runs for two years, then you upgrade it, AI requires attention, it requires regular optimization, and investment to eventually get it to full effectiveness.

Solar  

Regular operationalization, management, maintenance inspection of AI systems is super critical. You know, there's a whole topic of machine learning operations or AI ops just how to stay on top of the ever-changing data, because you can imagine how data is at the core of any AI systems, the data is always changing. So, how are you staying on top of the changing data, then as a result, how are you evolving the models that are built on that data? And then also thinking about the ethics and responsibility. You know, as data is changing, as models are changing, as society is changing. How do you continue building a system that is responsible that is trusted at the core? This requires regular maintenance, monitoring, management, it is core to not only the strategy. AI is also core to the operations of the company and ongoing operations,

Hobbs  

You know, Stela, Hugh spoke about the relationship between Microsoft and EY, how important is it for you, for partners like EY to assign specific teams to build capabilities amongst their clients so that they in turn, can harness the services that you're offering. Talk to me about this relationship you have with Hugh.

Solar  

We have seen that every company wants to benefit from AI. However, most organisations do not have the skills to fully benefit from AI and machine learning in the market today. And so partnering with EY who has such a deep skill set and such deep expertise to help organisations get up and running with AI, implement AI, expand AI and start that skill and change management aspect. That is required, that is so important for the value to really be seen by organisations and also more broadly across the world. It's a real tension point in the world right now, the desire to benefit from AI and the ability to capture this benefit. And so we are very happy about the strong partnership we have with EY.

Hobbs  

Hugh, you mentioned earlier how difficult it can be to overcome resistance to change in an organisation, whether that's emotional, or just because processes have been embedded since the dawn of time. And as you say, they there's a tendency to revert back to them, you're saying something very different in the need to go big. To generate that decent ROI, you need to target a critical big problem you need to build at scale is a critical conversation for both of you.

Burgin  

Absolutely, AI can be effective at solving so many business problems. But the greatest opportunity is to optimise performance, where there's the biggest opportunity for bad decisions, or non-optimal decisions, or inefficiency, or a lack of throughput that can impact the overall business performance. And so companies really should focus on thinking about the decisions they can make differently the actions they can take differently, and then looking backwards at how could an AI solution optimise that process. And then as they look at that business process, figuring out different points along the value stream, where there can be a really, really large-scale opportunity for better decision making and better throughput. And I think that that end-to-end process understanding which EY obviously has a has a lot of throughout our organisation, combining that with Microsoft's scalable AI technology really is creating a lot of value in the market.

Hobbs  

Yeah, I mean, unless point out the obvious thing, Hugh. You know, the basics on scale, acknowledging that maybe 10 years ago, server capacity within an IT department was almost based on the size of the of the physical server room. And today, vendors like Microsoft Cloud, can sell at almost infinite scalability on demand around the world. But it's a different environment now, correct?

Burgin  

That's right. I mean, you can increase your server capacity by sliding the dial from left to right on your computer screen, you don't need to go install a bunch of hardware anymore to do that. And so combining the scale of the cloud with Microsoft's native services, which honestly make it so much easier to implement things and get things up and running, once you have that technical foundation in place, the biggest opportunity then is leadership and sponsorship to drive the adoption. And I think that combining all these things together is creating so much value for a lot of companies when they can get the right mix for their clients for their for their business.

Hobbs  

Stela, just on the scalability points, if you would, I've been reading about what some call the AI productivity paradox that it's easy to build a proof of concept to prove your point, if you like in an experiment, but scaling that through an organisation is infinitely more problematic. And that's when it often fails. Correct?

Solar  

A common question is the productivity question. I was just recently asked by a customer, does it make sense for them to build a bot for their particular use case. And the verdict we got to was the particular use case they were talking about didn't make sense, they had few big wholesalers. And these wholesalers would then be taking their product to market at scale, but they weren't really getting these high volumes of customer requests that were coming in. And so the productivity paradox also comes down to being brave to qualify out projects, and making sure that you really double down on what will add most value to the organisation. And quite often the answer is somewhere around considering the volume, the velocity, the variety of the data, the types of requests, these are the factors you want to consider to ensure that the project that you select, and you know, we have very limited resources, you know, time, budget quality. So what do you choose? What is the project that you choose, you want to make sure you're choosing the right one, to bring that value to the organisation. So it is a hot topic. And unfortunately, because AI has quite often been driven by someone who's very passionate about AI, who might be in the IT department or a pet project. Sometimes it gets set up from the start for failure because it isn't necessarily aligned to this bigger overarching business goal. And so really thinking about the productivity, the maximum impact and how to align a project to where it matters most, that is super critical to this challenge.

Hobbs  

Hugh, where are you on the debate about abandoning proof of concept entirely. The Harvard Business Review says too many organisations never get past proving a point or experimenting, wasting months, during which they don't even think about how they'll implement it in practice. Kind of to what Stela was talking about just now?

Burgin  

Well, I think companies do want to prove value, and they want to demonstrate value. So I think pursuing a proof of value is good. But AI is proven overall, it does drive benefits, does drive actions and decisions. So the focus should not be on, can we prove that if AI works, the question should be how can we adopt AI and make sure that the insights and the value that it generates actually gets integrated into our business in a sustainable way. And that ongoing operating model, and Stela mentioned it earlier around AI ops and data ops and machine learning ops, these operating models are really key to ensuring that AI is not a point-in-time project, it's a journey, and over time, it will build upon itself and there will be ups and there will be downs. But over time, you will see a lot of value by committing to it and committing to the journey that it presents as it creates value and drives better business performance. 

Hobbs  

Stela, I want to pick up on something that you said earlier. And in preparation for this podcast, you said it in answer to a question about what's working now, to get C suite to better engage with AI, this startling observation that a lot of the world's or a lot of the clients that you're talking to are dealing with pain, you're really clear, you're really explicit, talk me through that, what is the pain that C suites are working through? 

Solar  

Now suddenly, in this context, leaders are being asked to make very important business decisions in the absence of fully understandable data. And in the absence of really even having a pure view of what might happen next, I think we have all been very shaken by the last year or so in the pandemic. And none of us could have predicted this. And so there's just a lot of ambiguity, a lot of unknown, and the trust in some of the models and data that leaders potentially used to have, you know, all those models turned upside down over the last year or so. And so, the ability of an organisation to respond quickly and to respond in a you know, in a way that allows them some movement ahead to re-establish and to thrive in this environment is super critical. And what we have seen is that couple of core factors have made a difference in organisations here. One is the availability of data and AI tools and skills across the organisation. We have seen that more agile organisations where this skill and knowledge is available across the board, whether that be you know, data scientists, developer, business leader, or regular day to day employee who may not be working on anything technical at all, but might be in the areas of sales or marketing, having the right tools for them to make quick decisions, leverage data, leverage AI, we have seen these organisations thrive in a much more effective way. And I don't think anyone wants to use the word thrive in the current environment, but respond much better to the unknown scenario. So this idea of AI and data for everyone, it includes skilling, it includes tooling choices, so really investment in underlying platforms and guiding principles. And it also includes an organisation empowering every individual to think data first, to think AI first, so that it is foundational to the culture, and foundational to the way individuals talk, make decisions and steer the business. And then as an additional element is around, trying to navigate this at a rapid rate, but in a responsible way. And obviously it's not just the pandemic context that we are we're in and hopefully exiting, but it's not just that it has brought up many other topics and themes of equity diversity, and we all want to do our best and grow and improve as organisations as individuals as the world. And so the responsible use of AI and data really becomes another fundamental factor and how do organisations ensure that they are doing the right thing here? This is another super-hot topic that C suite is trying to respond because with fast action, and empowering the organisation that also comes questions of how do you do it responsibly when you have such agility?

Hobbs  

Are you there? I mean, have you got it? Can you in a sense, formalise that mathematically or statistically? Can you use the cloud? I mean, where does that take you?

Solar  

It was 2016, I believe, when Satya first created our six responsible AI principles.

Hobbs  

Let me interrupt, this is Satya Nadella, who's the CEO of Microsoft, who's focused you guys on the growth area of the cloud?

Solar  

That is right, you know, no, we know him by a single name - Satya. It's a bit of a celebrity name for us. Well, Satya created the six principles of responsible AI. And, you know, I think they will be self-explanatory if I list them, fairness, reliability, and safety, inclusivity, privacy and security, transparency and accountability. And so this was our early way that we were starting to create a principle or a framework around it. We found though, that principles were not enough. We needed to roll this out across the organisation. And so every single Microsoft employee is trained in these principles, whether they're an engineer, whether they're a seller, whether they're a marketer, whether they're a developer, every single individual needs to be trained. But even this skilling rollout was not enough. So we moved to add a set of practices. And so we invested heavily in guidelines for human AI interaction, for conversational AI, what is ethical, what is not? And we're integrating these from the ground up into the decisions that our engineers make and the choices that each individual makes every day in how we're building things. And what are the best practices for that. We also have a regular AI in Research and Engineering committee. So the ISA committee that is governing the decisions that we make on our platform, and then again, documenting these practices, having these guidance documents was still not enough. And so the current area that we're in, that we're investing super heavily is in the area of tools, what are the tools that are available to organisations in the world and ourselves to implement AI responsibly? And so a lot of work around interpreting machine learning, how do you exactly know what your machine learning model is telling you, and why it's telling you – so interpretability is super important. And we've built an open-source toolkit for that. I also think about privacy, how do we protect individual data and so things like smart noise and differential privacy toolkits, that actually insert white noise to mask any personal data. So you cannot track the personal, personally identifiable information all the way through, and things like you know, homomorphic encryption, I love this word. And what's interesting about homomorphic encryption is it allows the data scientists to create models and built them and get accurate results without ever getting view of what the underlying data is. So these kinds of tools are really important to ensure that responsible AI becomes a reality.

Hobbs

Hugh let me turn the conversation to you and in a sense, to mirror the type of things that Stela is saying. But from your point of view, once an organisation has C suite buy-in, you say it then becomes about data, talent, tech, and trust. And I'd like to take each of those in turn. You wrote recently that CCEOs need to elevate data to a business-critical asset. What do you mean by that?

Burgin  

Well, it is a business-critical asset. And whether it's an AI or it's an ERP implementation, you know, data is essential to allow these initiatives to be successful. But what I would say is a data will never be perfect. Also, I've never been to a client organisation where they had all of their data figured out. It was all highly structured, it was all high quality, there were no issues, it's never going to be perfect. So while the target is important for an organisation around data, organisations can have significant success and significant value along that journey. So I would not let the perfection of data be a blocker for getting value out of AI or other initiatives. Instead, make it a key part of your roadmap, a key goal and maximise the return you can get along the way.

Solar  

I love this point of the perfection of data. It is a really huge topic because you know if we think about it, as it relates to responsible AI that we also just spoke about. We know that data has historically not been equitable. It hasn't represented the diverse populations, diverse abilities. And so perfection of data is something that we absolutely do not have. And so, again, another toolkit that we've made available is Fair Learn. And I know who this is one where EY Microsoft work very closely together. Fair Learn looks at the biases in data. And the project that EY Microsoft worked together looked at the loan approval process. So there's a loan approval solution that EY provides to organisations. And the loan approval algorithm is only as good as the underlying data. And this data is historical. And obviously, it's never perfect. But it was uncovered through the Fair Learn toolkit, that there was a low a disparity in loan approvals between men and women. So there was a bias where men were getting more loans approved than women. And this was due to the underlying data. And there was actually a 7% loan disparity where men were getting preferential loan approval to women. And so we worked with a Y with the Fair Learn toolkit to eliminate this bias in the underlying data. And now the discrepancy has moved from 7% to half a percent. And we continue to be reducing this. One more thing to really be considerate of is the data desert. We really talk about the data desert a lot in contexts of representing individuals have different abilities, different cultures as well, different contexts and situations, it becomes super critical for us to build robust systems to ensure that they are representative of all of everyone and of the holistic world.

Hobbs  

Okay, let me let me come back to these four areas that Hugh was laying out. Obviously we just spoke about data. Your second concern is talent, Hugh, the desire to embed data scientists, in these business units, or at least at a shared service centre comes up repeatedly in EY podcasts. But it's not necessarily a walk in the park. Is it to attract those graduates? What are you learning about how you pull this whole thing together?

Burgin  

Well, there's several different ways companies are organising themselves around data and AI services. And whether that's a hub and spoke model or a centralised team, there's, there's lots of different ways companies are thinking about how they organise around AI. And companies also don't want to outsource AI completely, you know, that they want to really control and own the talent within their organisation. They know it's something that's critical, long term for their success, and for their competitive differentiation. And so they're looking for partners who can help them grow their own talent, and also fill in the resource gaps they have, how can they partner together. So that's why many of our EY projects around AI are not EY delivering an AI project, it's EY helping the client deliver the project to be successful, and in really maintaining that intellectual capital and maintaining that knowledge and growth within the within the company. And we can help accelerate that for them to drive more success.

Hobbs  

Okay, your third point was tech. Your fourth point is trust. As far as technology is concerned, obviously, the scalability of the integrated cloud at Microsoft is absolutely huge. But I think we should mention the speed of delivery here that there's no longer a need to spend, I don't know, one or two years building a foundation or migrating assets before you harvest the value. In a sense, it's turnkey, correct?

Burgin  

That's right. Yeah. AI and really, technology in general used to be, it used to be a black box. So AI used to be a thing where you would shove a bunch of data into the box and see what comes out. It also used to be the type of initiative where you would need to spend a year or two years building the servers, building the foundation. But with the cloud capabilities today and the native services available from Microsoft, you can turn on that capacity so quickly. And you can activate those AI services so quickly, that it really allows you to shift from how do I build this big foundation to a shift towards what are the use cases that I want to go implement? And how can I go implement those use cases quickly? How can I implement them next month, or in three months? And so I don't need to wait three years to get that foundation, I can start solving business problems much more quickly than I used to be able to?

Hobbs  

And what about this fourth question of trust. I mean, this as I say, this comes back I think, to what Stela was saying about managing the risk to the business when you come up with an AI solution, right.

Burgin  

EY did a survey in the last year and it was interesting and found that nearly half of respondents cited a lack of confidence in the quality and trustworthiness of data as a challenge for enterprise a wide AI programmes. And so, you know, fundamentally, whether it's the data or the machine learning models running on top of that data, you know, trust is super important. And so our clients are really trying to put in the guardrails and the frameworks that allow their AI solutions to be trustworthy. And I think that starts with purposeful design. I mentioned earlier in the conversation that that AI for us is a narrow a narrow AI, it's a specific business problem we're trying to solve. And so purposefully designing a solution that balances the you know, whether it's robotics, or machine learning, or digital solutions, finding a purposeful design is a great way to build trust in that solution. And then when you provide the right governance and the right supervision over that solution over time, it really builds that trust that our clients need to be able to really scale AI and adopted in their organisation.

Solar  

And here, you might have seen this also, we have seen that this trust really gets built when there is cross organisational collaboration. So individuals who have different skill sets with different perspectives with different views of the data of the customer experience of the organisation experience. When all of these subject matter experts come together, that's when these trusted systems are able to be built with a diverse perspective and contribution.

Hobbs  

We're coming to the end of our time sadly. Stela, can you give us some colour on where you hope Microsoft R&D will take us soon? In a functional sense, or are all the basic AI tools now effectively on the table as far as you're concerned?

Solar  

I'm never good at reading tea leaves. What I do know is we will continue focusing on AI becoming available for all so that means democratising access to AI. That means creating low code, and no code experiences as well as fully rich customisable experiences. also thinking about skilling. How do we continue investing in skilling for all? You know, we have done a lot of work in creating the AI business school to make AI available for leaders, decision makers to understand the full potential and change management aspects. How do we do even more in the area of skilling, we also will continue our investment in the industry clouds, we are seeing a lot of opportunity in US maintaining our leadership position in being the most certified and accredited cloud, and also delivering industry specific, industry relevant scenarios to our customers. So this will be continued investment, and then leading and influencing to our best ability as we can the responsible use of technology, the responsible use of AI. We're living and breathing this with our own investments in AI for good initiatives in creation of the open available toolkits for implementing responsible AI systems. And I imagine we'll be seeing a lot more to come in this space.

Hobbs  

I was going to ask you if there's anything that AI or machine learning will not be able to do but listening to you there it seems like you're putting trust and responsibility at the heart of what you do in order to answer. There is nothing that we can't do. Correct?

Solar  

It's interesting to philosophise on this I have read many things where AI potentially may be limited in some creative topics or in some more high-dexterity contexts or even lean lateral thinking. And then, you know, models emerge where we are proven wrong. And so it's an interesting area to keep observing. I am reminded of a question that I was once asked on a panel, I was asked, would you let AI choose your life partner? And I think about this because when I buy something online, the ad for that item still follows me even though I have already bought it. And so I think about the maturity of AI and if it's still following me asking me to buy something I've already own. I don't think I would let it choose my life partner.

Hobbs  

But listen, we're coming right to the end of our time here. Hugh, peer into the future for us. If you would take us five, if you dare, 10 years down the line. How will AI have shaped business from now to then do you think?

Burgin  

Well, it's a great question. I think AI will have much more of an impact than it has already even in two or three years from now. I mean, whether it's the mundane tasks that we do throughout our life that it can automate for us and make easier things like scheduling meetings or taking notes or, or emailing people, I mean, a lot of a lot of the regular repeatable things we do in our life are going to get a lot easier because of AI, the other area will where it will have a big impact is in the big ROI areas and big investment areas of our businesses. So when we think about inventory, and capital costs, or we think about sales investments, these investments will still be there, but the yield that we get, the efficiency that we get will be so much more significant. I think companies who embrace AI are going to differentially perform in the market. So whether it's your daily task of taking notes or your big decisions around inventory, AI is going to have a big impact.

Hobbs  

Stela, where do you think we'll be in five years’ time?

Solar  

We will get to a stage, and we may already be there today, where we notice when AI is not there, we notice when AI is not part of a process or a system, we noticed that it becomes more manual when AI is not there. Maybe the experience is not as personalised and so my prediction is it will reach this integration point where it becomes almost invisible when AI is there, but when it is not, it's going to stick out like a sore thumb.

Hobbs  

Stela Solar, Microsoft's Global Head of artificial intelligence solutions, and Hugh Burgin, Data and AI leader for the Americas for EY’s Microsoft Services Group. Thank you both. It's been an absolute pleasure.

Burgin

Thank you so much for having me, Simon. This was a really enjoyable conversation. 

Solar

This was a fun discussion. Thank you.

Hobbs  

For more information, visit ey.com/Microsoft. A quick note from the attorneys. The views of third-parties set out in this podcast are not necessarily the views of the global EY organisation nor its member firms. Moreover, they should be seen in the context of the time in which they were made. I'm Simon Hobbs I hope you'll join me again for the next edition of the EY podcast, EY and Microsoft, your digital world realised.