| If you don't have an account, Register here |
On this special Fireside Chat podcast, HFS CEO and Chief Analyst, Phil Fersht, caught up with Paul Daugherty, Chief Technology & Innovation Officer at Accenture, on the potential impacts of GenAI on enterprises and the workforce; Paul has been a bastion of AI for well over a decade.
You can listen above or watch this HFS Podcast here:
Read the associated blog, “GenAI isn’t just eating software, it’s dining on the future of work” here.
You can also listen on:
• Apple Podcasts
• Spotify
• Google Podcasts
• Anchor
This transcript was auto-generated from the original recording and lightly edited for readability. We've done our best to catch errors, but names, technical terms, and company references may be misspelled or imperfectly captured. For the definitive version, please refer to the original audio or video. Views expressed are the speakers' own.
So let me start off with some questions that I think could be really relevant. I think the first thing I want to ask you, Paul, is, are you enthusiastic or scared with AI today? What is your overall feeling, especially with 10 months of ChatGPT and ChatGPT Enterprise and the speed of conversation? What is your general mood and feeling about the whole thing?
I’d say overall I’m enthusiastic about the potential of AI, in particular generative AI, both from a personal, human perspective as well as from a business perspective. And it’s important to recognize, as I know you’re aware, that AI itself has been around a long time. We’ve been using AI in all sorts of ways over the last 15-plus years, for different clients and in different parts of our business. So AI is pretty well understood in terms of machine learning, deep learning, and lots of ways we’ve applied that to solve some really interesting problems. Generative AI is an incremental and really a step change, certainly in terms of the capability that we can get through AI, and it’s really this mastery of language and the foundation models that you can train once and use many times. And the ability to create content now through the transformer technology — that’s really the breakthrough. And I think, just like we’ve seen with advances in other technology, it’s going to enable new applications, new opportunities to use the technology, and new ways to increase the capabilities that we as people have. That’s the way I see it impacting what we do and what each of us do and what businesses will do with AI.
Excellent. So how would you describe, as we look at GenAI, the value proposition for enterprises?
Yeah, the value proposition is that you can really tackle these problems of language in the one sense with large language models, which open up a new set of solutions for companies, and then this content-creation capability is very powerful. And so I think the value proposition for enterprises is it really allows you to do some transformative new things. Whereas a lot of technologies that we’ve experienced to date impact the way you use technology — cloud is a different way of building applications and systems — what’s different now about AI is it’s different ways of doing things. It’s changing more the business processes and how you work within an organization, and I think that’s the real value proposition for enterprises. You can rethink and change a lot of the things you do. For example, the way you create content if you’re a media company, or the way you create regulatory filings if you’re a financial institution or a life sciences institution. It dramatically changes what you can do, which is where I think we get to new value propositions for business and where some of the excitement comes in.
Yeah, we’ve run our own LLM on our own research, and the first thing we notice is people are going in and searching for insights who never would have done that before. They would have gone to their teams and downloaded reports and stuff. Now the whole mechanism for how you access information is rapidly moving, and this is very rife on the academic side too. So how and why then do you see GenAI as different from other technological disruptions?
Well, I think it’s a big disruption or a big step-change increase, as I said before. So I think it’s a big increase in capability for artificial intelligence specifically, and then in allowing us to do these new things. Some of the ways we see it impacting what organizations can do: it’s new forms of knowledge management, much easier to access knowledge, and it democratizes the ability of people to access knowledge in their enterprises. New ways of interacting with systems, which creates this tremendous potential in areas like customer care and customer service to bring in new models and new ways of doing things. And really new ways just for people in general to use systems. In some ways, we used to talk about the fact that software is eating the world. I think the new analogy with generative AI is that AI is eating software, which means that we have much more powerful ways of interacting with technology than we had before, which stands to benefit the way that we structure things and the way we work.
AI is eating software. That’s great. So if we look a year out from now — let’s not look too far beyond that — and we look at GenAI, what do you expect to start to retain some scale, and what type of technological advancements do you anticipate are going to drive these changes?
Yeah, I think one year out we’re going to continue to see a lot of the experimentation that companies are doing playing out. So we’ll see that a lot of companies are in the experimentation stage. They’re picking a use case, they’re picking some models, they’re looking at how to apply it to the business, and I think we’ll continue to see a lot of that over the next year. We’ll see more companies looking to then pick something and scale it to more widespread deployment. Some are doing that today, but I think that’ll become more common as we get a year out. The other big change I think we’ll see one year out is more of the traditional enterprise software companies ingesting generative AI capabilities into their products. So I think it’ll be very common a year from now for a company to be using Einstein GPT from Salesforce, or the generative AI capability from ServiceNow, or other capabilities that’ll just become natural in the way you do things. So the one-year-out view is we’ll move from some of the experimentation into scaling and then more naturally using some of the capability. At the same time, we’re seeing right now a lot of companies — and we’re helping a lot of companies — establish the foundation in terms of the center of excellence, the talent foundation, and responsible AI, and we’ll continue to see that play out so that people get more comfortable deploying generative AI more widely through their organizations.
Do you think anything’s going to disappoint us in the next year?
Yeah, I think we’re going to go through the inevitable backlash. You’ve been around a long time, you’ve seen this happen with every technology. So right now I think there’s this perception that generative AI solves any problem you can imagine, and I’ve talked to lots of companies who come up with all sorts of different generative AI use cases. I think some of the reality is going to set in, and one of the things that’s not being looked at enough right now is the business case around deploying the technology. A lot of people are moving down the path experimenting with use cases, and you might ask, is that use case you’re experimenting with something you can scale in an affordable way when you look at the degree of training you might need to do and how you might need to deploy it? So what we’re encouraging companies to do is start even at the experimentation stage with a business-case-driven view — understand the costs, understand the compute and sustainability impacts and other things you need to factor into your solution, so you know what the next stage is going to be and how you’ll be able to scale it to drive real impact. And I think some are going to be disappointed that they get through some experimentation — the experiment itself worked — but then you look at really applying it at scale, and it either wasn’t as easy, or it was more costly or not as effective as they thought it would be.
Yeah. Definitely. So which roles and functions and even industries do you think are going to be most disrupted as we look at this sort of 12-, maybe a bit further out, 18-month horizon?
It’s a really great question, because it seems like it’s happening in every industry, unlike some other technologies. I can point to examples in every industry, so nobody’s immune from the impacts of generative AI and, conversely, everybody’s got the opportunity. So we see pretty widespread, distributed application by industry. I think what’s going to be a difference maker in our view is that the first movers will have an advantage — those who master the technology, who understand the models, who get their digital core established, because you need a strong digital core and data foundation to benefit from generative AI. Those who are first movers in those ways will have a sustainable advantage going forward. So I think that’s one thing to think about: less so maybe than industry, it’s how do you become a first mover and get all the prerequisites in place that you need to be successful with generative AI. In terms of roles, I think it’s going to be kind of unevenly distributed. From the research we’ve done, we see that 40% of working hours across industries will be impacted in some way by generative AI, which is a lot. But that doesn’t mean 40% of jobs go away, because in most cases generative AI is impacting a part of a task somebody does, and it’s making their overall job more effective and maybe more fulfilling by removing some of the drudgery and some of the detailed work they needed to do. So when you look at roles that are impacted, those that are very information-oriented or knowledge-intensive will be the most impacted. It doesn’t mean those jobs are going to go away, but it means the jobs will change substantially. You’ll see that in areas like customer care and customer service — a natural area. Generally speaking, knowledge-worker types of jobs will be heavily impacted in terms of using the tools in different ways as they look at the way they do their work.
Yeah, it’s very interesting. And then let’s think about the skilling, the reskilling, whatever we want to look at here. How fast does that need to happen, even when you look at your own business? How fast do you think that needs to happen in terms of keeping on top of this and going in the right direction, as you look at the business you’re in? Or do you think people are overemphasizing the need for speed at the moment?
I think the reskilling is one of the real mandates that does need to happen fast, and it’s along two dimensions. There’s reskilling of workers who need to do the AI — people who can help develop and deploy your generative AI, the data behind it and everything else. And then there’s the reskilling of all the workers and everyone who needs to use the generative AI, because you need to learn different techniques. You might need some prompt engineering injected into a traditional job that you’re doing today. So there are two different types of reskilling, and both are super important. I think you need to get on that path today. That’s what we’re doing within our company, really looking at how we reskill all the people we have in Accenture, and we’re increasingly doing that for other companies and helping them set up the training academies and learning processes so their employees can learn to move forward effectively. So I think it is real and I think it is a mandate, because — back to what I said earlier — you’re not going to be able to be a first mover and out in front if you don’t have the talent in place. You can acquire the talent, and your companies do need to partner and develop their own ecosystems to get access to the talent and capability, but fundamentally you need a good bit of it in your own organization too.
Right, right. And how does that look in terms of workforce training, adoption, understanding the different needs and different skill sets? Are you going to have to bring in training organizations? Are you going to rely on your own technology organizations that you might use for your clients to do this? What are the plans within Accenture?
Well, at Accenture we’re kind of a learning organization, so one of our strengths is really the learning foundation and the abilities we have to equip and skill our people at scale. So we’re applying that now around generative AI and looking at how we train our people. We do it at multiple levels. At the highest level, we have a program called TQ, or Technology Quotient. It’s something all of our people do, and we have that for generative AI because there’s a certain amount that everybody needs to know about it. And then at the other end of the spectrum, we have certification around generative AI model development, which is very specialized — in our case, still large numbers, thousands of people, but not every person in our company needs to understand that. And then there are a lot of skills in the middle: prompt engineering, things like explainability engineering, which is a real skill that’s needed, and many other very specialized skills that are required. And then we’re also accelerating — we’ve made an announcement about the investments we’re making in data and artificial intelligence, a $3 billion investment. Part of that investment is going to doubling our workforce in this area from 40,000 to 80,000 people, leveraging all the kinds of things I’m talking about. So we really do do this at scale. We’re investing at scale, and we’re doing that in a way where what we do is packaged up in learning approaches and training and academy types of approaches that we can take out to the marketplace and to our clients as well.
Yeah, thank you, Paul. I think I read your first AI-aligned book around 2016 or 2017, something like that. So you’ve been covering this space for a very long time. You’ve had a lot of time there. You’ve had a vast experience of the ethical considerations and what people need to take care of over the years with this. Is there anything different now when we look at ethical considerations? People talk about responsible AI, these types of things. Is it any different than it was, or is it just accelerating?
Yeah, it really is different in a few key respects. The foundations of fairness and ethical AI are similar to what they’ve been for years. The first book I wrote on that was in 2017, Human + Machine, which really focused on some of these issues. At the time, though, I described it as kind of a push — we were pushing it to try to educate people on it. What’s different now is that there’s a pull, where companies are looking to access responsible AI capability and ethical AI because they understand the importance of it and the risks. Generative AI has heightened the awareness of it because there are some new concerns. First of all, it’s just so accessible that anybody can use it, so people are getting more firsthand experience with it, which leads them to ask the questions more. There are some new risks around IP and intellectual property, which becomes more of a consideration given how generative AI models are trained. There are new concerns around misinformation at scale, given how you can use generative AI for deepfakes and other types of misinformation at tremendous scale and volume. So there are some new issues that generative AI brings about, and then fundamentally the awareness around these issues is what’s led to this demand for responsible AI. We’ve been focused on this within Accenture for years — since before I wrote the book — and we actually have a compliance program for AI and responsible AI that measures the risk of everything we do with AI on our own or for clients. We have tracking tools and risk-assessment tools and operational tools to monitor and measure this, and that’s what we’re encouraging our clients to do. You can’t just do this by chance. You can’t just train your people a little bit and hope they do the right things. You really do need to put in a systemic responsible AI foundation that’s based on principles, risk and compliance assessment, enabling tools to support it, and then ongoing training along with monitoring to make sure that you’re building the awareness across the organization.
Yeah, it’s well put and well thought out, Paul. So I have to ask this question — I don’t like it — but do you believe white-collar jobs are under threat with GenAI any more than they were in the past, and how do you see this starting to play itself out?
Yeah, we had a concept in that first book I wrote called no-collar jobs, because we really saw AI bringing about a fusion of people with different sorts of skills. It wasn’t about the blue collar or the white collar; it was kind of the no-collar jobs, as a lot of skills merged together. And I think that’s even more true today. So, specifically to your question on knowledge workers or white-collar workers, I don’t think it’s a threat at all to white-collar workers. It’s a threat to people who don’t learn how to use the new tools and new approaches and technologies in the way they work. We’re going to need people who understand a wide variety of domains to do the kinds of jobs they do today, but they’ll be jobs that are informed and infused with generative AI assistants that are helping them do their job more effectively. For example, a large multinational bank we’re working with — in some of their back-office processing they still have the people doing it, but a lot of the work is being automated and prepared for them so it can be done more efficiently by generative AI. Or an energy company, where workers from a safety perspective are getting better information on where the safety conditions and risks are, and the regulatory and environmental safety precautions they need to take are clearly laid out so they can operate in a safer way going forward. These are people doing important work, but doing it informed by tools and AI systems in different ways. And I think that’s really what the future is about. Do you need fewer people to do a body of work than you do today? Probably, but that’s the way technology’s been forever. That’s not a new thing. Technology — not just information technology, but for centuries — has been about enabling people to do things more effectively. So that’ll continue, and I think there’ll continue to be a need for people to bring all sorts of expertise to bear, fused with technology, to do their work in new ways.
Yeah. So I’m going to ask you one final question — we’ve been going for quite some time here. How would you sum up what the world of enterprise tech will be like in another two years? So think about two years. How fast are things moving based on your new experiences?
You know, it’s really early days, but I’m fascinated by this question. I’m thinking about it a lot, because I think this is going to bring about one of the biggest changes in enterprise tech that we’ve seen since before the internet — in a long time, let’s just say it that way. And maybe one of the biggest changes ever, because when you think about enterprise architecture and enterprise technology, I think intelligence is a new feature. We used to talk about data and processes and infrastructure and different layers of architecture. Intelligence is now a core component of the architecture. And what are the implications of that? How do you build not just a single model but multiple models into your architecture? How do you know that they’re up to date and consistent, being fed by the same data in the right way? How do you know they’re responsible and operating in the right ways from both a responsible and a regulatory perspective? So I think there are tremendous implications, which is really exciting, because it’s going to create new opportunities for organizations and new ways to do things as we go forward. So it’s a really interesting period that’s going to create a lot of complexity for CIOs. For my CIO friends who are operating in this world, I think it’s going to create a need for new, structured approaches to deal with this. We call this the digital core — the modern digital core — of how you bring this together and integrate these new capabilities and how you operate going forward.
Awesome. Well, this has been fantastic, Paul. I’ve really enjoyed hearing your insights, and hopefully we can meet up soon and talk a bit more about some of this stuff. It’s been great hearing from you again — we haven’t spoken for a while, I think it was the pandemic time.
It’s great to catch up, Phil, and it’s changing so fast. That’s what’s interesting about this period of time.
OK, great.
Thanks, Phil. Good luck.
Your account has been created. You can continue exploring free AI insights while you verify your email. Please check your inbox for the verification link to activate full access.
With the exception of our Horizons reports, most of our research is available for free on our website. Sign up for a free account and start realizing the power of insights now.
Our premium subscription gives enterprise clients access to our complete library of proprietary research, direct access to our industry analysts, and other benefits.
Contact us at [email protected] for more information on premium access.