Drive successful AI through good governance, oversight and accountability

Transcript

Arif Mohamed
The majority of organisations are still in the early stages of building the foundational skills, infrastructure and governance needed to drive AI transformation. Building trust in this new technology is key, but businesses can easily stymie innovation if they don't ensure human oversight is applied selectively and effectively. Before we dive into the discussion, let's find out more about our experts. Welcome Claire, please tell us about yourself.

Claire Reid
Great to be here. I'm PWC UK's Chief Technology and Innovation Officer. My background for years, I'm a technology risk practitioner and I've helped clients navigate cybersecurity regulation, resilience and large compliance risks.

Arif Mohamed
Great to have you, I'm Warren.

Warren Tucker
Yeah, it's great to be here also. excited for the conversation. So, Warren Tucker, I look after our technology industry for the UK. So that's where all of these AI software platform businesses sit. And I have the pleasure of working both with them but also with our big enterprise clients, typically on strategy. How do they find value from these technologies? And then often moving into execution and implementation. So how do we make that real and bring that to life? So it's an exciting time.

Arif Mohamed
Well, it's lovely to speak to you both. So my first question is really that AI is impacting the workforce in profound ways. How do you make sure that you have a broad set of stakeholders and how do you get them to work well together, Claire?

Claire Reid
This is a very real challenge that we are working through at the moment as well as PwC. I think the first and foremost is don't see AI as a technology issue or just a compliance issue. It really is a horizontal that impacts every single job role around our board, around our exco. So making sure you have carefully considered the implications for AI right across each of those functions is critically important.

And where we see clients lean in and treat it just as a technology challenge, that's where they're really struggling to see the value come through through AI.

Arif Mohamed
Do you have anything to add, Warren?

Warren Tucker
Do I mean I you know Claire has this challenge herself directly. I often work with clients on this same issue, and I think there's a lot of press around AI at the moment. some factual, some myth. and I think the main I guess what does success look like? I think success looks like those companies that start to get hands-on at an executive level with these technologies, so it gets brought to life. So rather than moving out of the realms of conceptual, it becomes real. So for example, one of my clients that I'm working with at the moment, a big technology business.

We're build we're building an AI model office. So that is something that they can touch, feel, engage with, whether I'm in operations, in sales, customer service, in the finance organisation, in the supply chain business, bringing that to life. And then I think people can start to see not just the opportunity but the things they're going to need to do personally to engage with it. And then you get people leaning in rather than being pulled in.

Arif Mohamed
This kind of leads into the next question, which is how do you kind of trust in AI when there's also a fear of it amongst people, including the workforce?

Claire Reid
I mean, I would build on what I said before about AI not being seen as a technology issue. I would say 80 % of it is a workforce challenge or opportunity and really getting that transformation right around the technology. There's a lot of hype at the moment about AI is going to be replacing jobs. And what I say to our team members, AI is not replacing your job, but somebody who uses AI will.

So making sure we're giving our teams the space to experiment with AI, try out the tools and build their confidence around it. Because there is fear of the unknown, particularly if you're not a technical individual in an organisation. So it's breaking down those barriers and making it far more accessible. And that's a key point. The other point is the transparency aspect, making sure that as an organisation, as leaders, we're talking about it.

We're bringing it into the organisation. We're encouraging, we're incentivising people to use it as opposed to building that fear up about not wanting to lean in and use it. So having a bit of fun experimenting but having that leadership commitment are all key attributes, I would say, of trying to alleviate some of that fear and uncertainty.

Arif Mohamed
Absolutely Claire and Warren, what's your take?

Warren Tucker
It it's a challenge. and I think again it's about trying to make it real for employees, specifically in the role they're performing. So as opposed to just giving everyone copilot or chat GPT or any of these technologies and saying, here it is, curate the learning experience. That's what our AI return on investment study showed. Those leaders are investing in roles-based training to to myth bust because those organisations that do embrace this type of technology, I think can't get enough of it. And then it becomes a different challenge. back to Claire's skill sets around controls, is how do we control spend? Because if everyone starts using it ad nauseum, then you can imagine the token usage is going to go through the roof as So you get a slightly different challenge. But I think we're still in that adoption curve at the moment.

Arif Mohamed
I guess one thing that can help with assurance is having human in the loop effectively. So what's the role of human in the loop and having the right accountability? How can that kind of help organisations have effective AI? Start with you, Warren.

Warren Tucker
Yeah, I think well it's interesting, I think IT organisations are probably the place where I see the most opportunity near term, because the human in the loop is already baked in most IT delivery projects and programs, very systematic processes, not necessarily waterfall, but you still have key documentation and artifacts and workflow systems and processes that are very consistently deployed across design, requirements, build, test, deploy. And of course a lot of that activity is through the parties.

So a lot of clients have third-party outsourced IT operations. So there's a great opportunity to both embrace the roles-based consistency of process and therefore deploy AI into that environment, but also very quickly bank a check that's third party spend to be reinvested back into other sources of AI advantage. And I think the key thing around the IT development process is coherence.

Arif Mohamed
Very good. Claire, what are your thoughts on your advice on human in the loop and best practice accountability? What are your thoughts?

Claire Reid
Yeah, I'll bring in a different perspective to what Warren just said. My number one perspective on this is you cannot delegate accountability to AI. And I think that's the first tip I would give organisations as well. Just like in the olden days or six months ago, if I was writing an internal communication, reviewing a report to go out to a client or a proposal, I wouldn't just assume that AI got it right. And some people are assuming that what comes out must be right. So it's making sure that we're very clear in our training as well, what are the points in that workflow where the human in the loop is critical, be that the human judgment, the critical thinking, and not allowing AI to take the accountability away from the human. And particularly in our organisation where there are aspects of what we do for clients that are regulated. You for our clients expect 99.99 % accuracy every time. There's no way we can assume, I mean the models are far away from that sort of accuracy level now so we absolutely have to have the human in the loop in these processes and that's part of our training isn't it and what we implement it with our clients as well.

Warren Tucker
Yeah, I I think it's spot on around thinking about AI and agents not as technologies over in the car park. They are an embedded part of the team.

Arif Mohamed
This is kind of in your ballpark, how important is governance and how can companies make governance a competitive advantage for them?

Claire Reid
Well, you took the words right out of my mouth there about it being a competitive advantage. If you can get AI trust and governance right from the start, I really believe you can scale at pace the deployment of AI across an organisation. If you don't, there's going to be barriers and roadblocks along the way. So making sure you've properly considered the governance steps. And that includes making sure you've got the right stakeholders involved from the start, such as the risk function, for example, as well as the engineers and the lines of business or I guess the customer facing teams. This is a multi-team effort to get right and governance is no longer just the responsibility of the risk function. It absolutely is now owned by that ex-co and the front of house. And organisations we see are being diligent about governance are able to scale innovation and outpace their competitors because they're able to deploy that AI at scale and bring it into some of those higher risk domains in the organisation as opposed to just being stuck in that experimentation type phase where we see a lot of businesses at the moment.

Arif Mohamed
Absolutely. Well done.

Warren Tucker
I mean, I think that's that last point there, Claire, spot on. I mean, the experimentation phase, I think everyone's dying of death by proof of concept or proof of if you like, distributed activity. I think the governance point is so critical because it's about yes, making things happen and catalyzing action, but it's also about k stopping stuff, killing stuff quickly. So rather than just letting this proof of concept piece just proliferate, it's about speed of governance and decision making. So you make the right decisions early. So by all means experiment. Is there value here? If not, kill it quickly. Don't allow it to persist. And I see many organisations where the bias was very much in response to a a level of board pressure. Say we need to be seen to be taking action. Well that's true, but it needs to be focused, coordinated action. Experiments that are undertaken, is there value? Yes, double down and make coherent end-to-end. If not, kill it quickly, save the money and reinvest elsewhere. That muscle is a new muscle that I think many organisations haven't yet developed. So done step one, push it to the edge. Let's get everyone experimenting. The governance part is is really I think where we are now.

Arif Mohamed
Yeah, really good. I know that governance is essential, but I guess on the granular level also, what's the role of engineering and when it comes to kind of your approach, standardisation, role-based design. Let's stay with you, Warren.

Warren Tucker
Well, this is something Claire and I work very closely on together. So having spent some time actually building our engineering teams in the past, and this is now a big function under Claire's responsibility, is around how do you make engineers not something which is if you're like in the basement, but how do you embed it into the business, but into the the practitioners that are out there in our business working with our clients, whether that's strategists, so it's it's strategy with proof of value very, very quickly by having engineers deployed at the front line in with our clients to demonstrate the true practicalities of the value and mythbust often around the blockers and the things that would hold it back, which again, lots of opinions will come out of the woodwork. We couldn't possibly do this for XYZ reason. It's hard to myth bust that unless you can see it, touch it, and bring in some of the skills that Claire described, whether it's control, cyber, observability, which is a key, how do the model work? How do we know the output that it's generating is accurate? All of those things become visible the front end but it's the same everywhere finance HR supply chain get those engineering teams out of if you like the basement and get them out there problem solving not trying to become advisors or strategists but problem solving because that's what engineers love is solving a problem it's about giving them the right problem to solve so I think that's how engineering can play not just a a if you like a subservient enabling role it's a catalyzing role that can show the art of the possible so great time to be an engineer I would say.

Arif Mohamed
What's your take on that?

Claire Reid
Well, we're just launching our forward deployed engineer capability, which is very much bringing that depth of engineering capability into the forefront to solve our client problems. So it's not just engineers experimenting in science innovation projects. It's genuinely market back and it's relevant for clients. And they bring the art of the possible. And that's where we're seeing the unlock. But the trick for us and for our clients as well as the pace at which you can prototype at a rapid pace because what we cannot have are months and months months long R &D type processes that we might have seen in the past. We need to go into days and weeks of innovation rapid prototyping. That's the unlock and that's where we're seeing the opportunity to really get the value from the AI deployment.

It's where you're bringing that engineering and business capability together to work side by side.

Arif Mohamed
That's a really good point about speeding up deployment cycles and proofs concept, getting them. So what other things can help with acceleration of projects?

Claire Reid
I think it's listening to both parties, it's sharing the problem solving together, not seeing the engineers as just a support function, it's bringing that capability together, it's empowering those teams, it's giving them access to the right tools as well. Not losing the guardrails though and the architectural standards that are really important that again our clients expect us to be delivering for them, but it's making sure that you're bringing all of that capability in, into that team, into that squad that you're innovating around.

Arif Mohamed
you add anything to that.

Warren Tucker
I mean the other theme I think we've seen working together on this topic is that it's also trying to get organisations off of this notion that AI is for efficiency and cost. It's all about cost and efficiency. It's about growth. You know, our studies are showing that those organisations that are leaning into AI as a source of new business model, new opportunity, new growth, and doing things that you otherwise couldn't do is a brilliant way of of making it real and I think capturing a different, I guess, level of creativity from inside the workforce.

And we work with a big global publishing house who sells books and titles to lots and lots of listen brands all over and different locations, different sizes. So the notion of being able to tailor and curate marketing campaign messages around specific publications to thousands of locations simultaneously, all personalized, was just meant it couldn't be done. Like it wasn't humanly possible to do that within the realms of what was sensible for those teams, but suddenly it was an unlock. But it wasn't looking at it through how do we make this more efficient, it was looking at how do we build new capability to give ourselves a better fighting chance for growth on a wider set of customers that otherwise we couldn't serve. So I think the more that you can bake that mindset in, that how do we see this as something to to help power our growth, and we all know growth's a big thing at the moment in the in the global economy, I think that's the route to success as well.

Arif Mohamed
Great customer example that you shared there, Warren. I'd love to take you back to an earlier point about coherence and you talked about how it's important or you advise customers to have coherence in their AI strategy. Can you unpack that a bit more for us?

Warren Tucker
Yeah, I think we're working with a big organisation at the moment and we talk a lot about value streams. So what is the end-to-end value stream for like lead to cash is a good example. So finding a customer, selling to that customer, converting that sales process into a set of orders, ordering, delivering the goods and services and devices in some cases, collecting and running the financials around it and then billing for it. So that's a coherent process. Doing a few interventions, looking at it from a perspective, I can deploy some tools here, tools over there. It might make each of it a little bit more effective. But value to the enterprise comes from that process coherently using AI to improve quality throughput and in some cases profitability. So, how might it optimize the contract on behalf of a customer? So looking at it end to end is a different way. So when I say coherence, it's really that building that end to end.

Process understanding, which some people call foundations, but then looking at the workflows and the data that underpins that. And then looking across that for the AI opportunities to be able to drive an effective series of changes in unison, that's the unlock. I think that's why you're also seeing some of the small fast growth businesses growing very fast because they're designing with that mindset from the get-go. Big enterprises haven't historically had to do that, but now's the opportunity for them to move quite.

Arif Mohamed
on that point of fast growing organisations and projects. As we come to a close, what are some of the best organisations doing well in terms of their AI growth? And that could be human in the loop, it could be the technical skills they're embedding, it could be having good coherence. Just some more examples, I'd love to hear that. Claire.

Claire Reid
I think something we touched on earlier around training and AI sponsorship and adoption. think those organisations that are genuinely have the boards behind AI and that training. We talk about learning that can't be delegated. think for the you know, prior to AI, you could probably get away with not being that up to speed with all the latest technical tools because that's what your teams would be doing, would be using their time for. AI is a game changer for me now. They can't delegate that learning to anyone else, they have to lean in. So we've got our whole board in the UK personally getting bespoke training around, and we're using Co-Pilot as our house, kind of core LLM that we're using to build our own agents and start to bring that into the boardroom as well, because that will have a really powerful cascading effect across the organisation.

So that would be my number one tip I would say around getting that upscaling across the whole organisation.

Arif Mohamed
Thank you Claire, Warren.

Warren Tucker
Covered a lot of them actually. I do think Claire's spot on around executive hands-on sponsorship. And I would say that the part that I would emphasize is trust. You've got to find a way of building organisational trust that these technologies are a force for opportunity and growth, not just another mechanism to cut heads. And I think there's been quite a lot of myths around that. I think a lot of organisations probably overinvested and over recruited.

During the heady days post lockdown. And there's been a correction of that. I think a lot of that got attributed to AI. But just some of the studies that we're seeing, there's not necessarily consistency there. I mean we're seeing companies, it's only, while 81% of CEOs are backing big AI investments, only 12% are seeing a combination of both efficiency and revenue growth at this point in time. So the two things can't be true. So it can't be a massive source of cost efficiency, but yet the real stats don't back that up.

And so I think it's about myth busting the role of AI to engender trust and build, if you like, at the edge, an employee level willingness to participate and actually get hands-on and see the opportunities there. So that trust can only be built top down. So it requires all of the executives and leaders to get hands on with the technology. I was just sharing with Claire a funny story that was listening to someone the other week, and they were saying they've just hired a former chair who I think is something like 70% four years old or something like that. Don't quote me on that. And this person has gone all in on AI personally and is now going around to boards and saying, I use AI and here's how I use it. And actually then if you like training, go back to Claire's point, training the boards of companies to get hands on with it. So if they're giving an opinion or asking to pass judgment on an investment plan and strategy, they've got some hands-on experience. So

Again, there's lots of new techniques and methods coming out there to try to break down this trust barrier, but I think that is gonna be a big unlock for every organisation. It's something I know that we as P W C are investing heavily in.

Arif Mohamed
Brilliant, fantastic. Thank you so much. It's been a fascinating and useful conversation. So thank you so much, Claire and Warren, for being with me today. Well, that brings us to the end of this CIO video with PwC on the key topic of AI and human oversight. My thanks to Claire and Warren for sharing their expert insights, as well as everyone who watched.

Claire Reid
Thank you.