The State of Generative AI

Where we are, where we’re going, and where the value lies

Digital experiences that build customer confidence

Generative AI has a potential $2.6 to $4.4 trillion economic impact, but how do leaders unlock it? Join Tom Godden, Director, AWS Enterprise Strategy, and Aamer Baig, Senior Partner at McKinsey and Co., as they discuss the current state of generative AI, and where it can take us in the future.

Where we are and where we are going

Learn how generative AI can address long-standing challenges in enterprise technology, strategies for introducing your workforce to gen AI, and why a strong data foundation is an essential part of deploying the technology responsibly. See the details of the conversation below:

Transcript of the conversation

Featuring Tom Godden, Director, AWS Enterprise Strategy, and Aamer Baig, Senior Partner at McKinsey and Co.

What is the value of generative AI?

Tom Godden (00:10):
Can you share from your perspective what generative AI means? What's the business value? What are we looking at when we're looking at generative AI?

Aamer Baig:
Generative AI has the promise of being what we would call a generational technology, and allow businesses and functions to really reimagine themselves. What we believe is it's an evolution across many years of development of AI, and this is the next step, where you have a field of AI that actually generates content, synthesizes content, and can do a lot of other very powerful things. The value for enterprises is mainly in productivity, but also in creativity. At McKinsey, we looked at 63 use cases across a number of functions. When you ask about value, think of this number. The value is about $2.6 to $4.4 trillion in economic impact. Among 63 use cases, about 75% of this is coming from four functions: sales and marketing, R&D, software engineering, and customer operations. These are the four that we believe will be most fundamentally transformed.

Tom Godden:
Why is that the case for those? Why those four?

Aamer Baig:
Well, you have to think about where the actual craft and the activities in these functions can actually be re-imagined or transformed using a new technology.

Tom Godden (3:10):
Can we talk about that for just a second? We see this a lot with organizations that rush into that new technology and they use that new technology to just redo what they used to do. It's when you use the new technology to completely reimagine the process that you find value.

Aamer Baig:
Let’s take some of the examples. If you are a customer service rep, you get all these calls. You have to process a lot of information and react to it in real time. How wonderful would it be that you get suggestions on what the potential answer could be to which you can actually apply your own human judgment and then help the customer? In R&D, coming up with new designs, new formulations, new blueprints has always been learned over time. How great would it be that you have a technology that can assist with new combinations of that?

Boosting workforce confidence in generative AI

Tom Godden (4:47):
This technology is going to really help propel people to new levels of creativity and efficiency It's going to be an incredibly positive thing, but people are still worried. How how are you seeing organizations help people get comfortable with this technology?

Aamer Baig:
There are three important strategies that will help with that. One is your posture and frame of reference as far as the potential for this opportunity. If it's a human assisted technology, human empowered technology, I think it opens the aperture and opens minds and hearts. That's one. Second is it has to also be followed up with an emphasis on scaling up the right people to be able to deal with it.

Tom Godden:
Train your people, train your people, train your people.

Aamer Baig:
Exactly. Then, the third is to adopt a set of policies that will also give comfort but protect people as well. You want to make sure that people are using information safely. You also want to follow through, especially if it is a customer facing application, that you're checking for things like toxicity. You are putting the right guardrails around mistakes that sometimes this technology still makes. That all will be collectively confidence boosting.

Identifying a generative AI proof-of-concept

Tom Godden (7:55):
How are organizations approaching, "Where do I put those bets on the right proof of concept, the right ideas, so that they can become something and then so I can scale them?" They're doing cool things, but not cool things that necessarily deliver business value, which candidly should be what we're doing.

Aamer Baig:
I think it's an important management challenge right now, to both allow for some experimentation and learning, but also really place your bets where it makes a difference. We're suggesting a two by two approach, which is pick two where you can make an impact very quickly, so you actually have some learning that you can actually go forward with, and two where you believe it could be game changing for your business.

Tom Godden:
Yeah. What's your favorite example? Where you went "That's pretty clever"?

Aamer Baig:
I go back to what I was trained out of school to do, which is in software engineering. I actually think the opportunity closest at hand is how we can really drive developer productivity using this.

Tom Godden:
I love that one. There's very few things in the CIO that have made me just go, "Wow." When I look at the productivity gains with things like Amazon CodeWhisperer, it blows me away. 57% increase in productivity, 27% more likely to be successful. How can you not be adopting that technology? It's unbelievable.

The data you feed into generative AI models

Tom Godden (11:23):
Can we talk about the foundation here a little bit? In particular, you got to get your data right.

Aamer Baig:
It seems like it's a perennial problem, which is data quality and data management and availability of the right data. It was important before, but it's even more important now, and it's coming into sharp relief. It's not an exaggeration to say that whatever you want to do with AI, not just generative AI, is gated by the data you feed into the models. The effort to build the right data ecosystem is an important down payment for getting the benefit out of generative AI. For some situations where your proprietary data is not used, maybe one large language model is appropriate. When something is closer to your crown jewels, your crown jewels being your data, your institutional knowledge, people are being quite careful about what to use. A lot more testing and analysis is happening there.

How generative AI will impact CIOs

Aamer Baig (14:45):
How do you see this changing the IT organization in the future and the role of the CIO?

Tom Godden:
What this is doing is it's making us realize that technology should be embedded in almost every single thing an organization does. As part of that, it's a massive decentralization, I think, of IT.

Aamer Baig:
I would start by saying first, anytime there was a generational technology, one of the secondary effects of that has been on the impact of an IT function in companies. You have mainframe computing that led to the rise of an IT department. Then you had the internet, which allowed us to access offshore ocations because you have bandwidth and you have availability and the rise of various different providers. Then you had cloud and mobile. You could argue that propelled the adoption of product and platform operating models and how IT organizations are structured. Then you have the proliferation or the mass adoption of AI coming in this decade. I land in a similar place that you do, which is technology will no longer be a function, it'll be a capability that's embedded in every aspect of how you deliver value to the customer.

What’s concerning and exciting about the future of gen AI

Aamer Baig (17:49):
As you sat in your chair as a CIO what would give you concern and what would give you excitement here?

Tom Godden:
One of the things that gives me concern coming from life sciences is the perceived randomness of the potential answers you could get from a generative AI. You sit there and go, "Gosh, I love all the potential and everything, but I just can't have it generating answers." Now with that said, I think there are techniques that can solve that, RAG being one of them, retrieval augmented generation. You can query a known repository of answers, but still get the rich contextual dialogue information of that answer, but make sure that the answer is always blue, because that's the approved answer, not green. Blue's the answer. What excites me on it, we talked about the developer one. That certainly excites me on it, but I think just the ability to democratize IT. When you put power in the hands of the people, it's one of the best things that you could do.

Aamer Baig:
I think there have been some intractable problems in enterprise technology. I feel like this has given us a chance to start attacking them at a cost that might be manageable. The three that come to mind, one is tech debt. We seem to accumulate more and more every year. Second is talent. We've always had a dearth of talent, so I'm actually not worried about jobs going away. I'm actually excited about more work getting done with the talent we have while driving the experience of the developers.

Tom Godden:
 I think it's going to great jobs we haven't even contemplated.

Aamer Baig:
That's a great point. Then the third, we've always had an issue delivering large tech enabled projects on time. I'm hopeful that there's some interesting applications of generative AI that can actually solve some of these intractable problems.

Tom Godden:
I think finally the technology's ready to meet the moment.