Vivek LUTHRA, Senior Managing Director, Growth Markets and ANZ Data and AI Lead, Accenture Debanjan SAHA, CEO, DataRobot Moderator: Ellie AUSTIN, FORTUNE
Category
🤖
TechTranscript
00:00You both work with companies across a range of sectors.
00:03Can you each highlight one or two industries
00:05that you think are really leading
00:07the way in terms of turning AI integration into revenue?
00:12Vivek, let's start with you.
00:13Sure.
00:14Thanks, Ellie.
00:14Good evening, everyone.
00:15My name is Vivek.
00:16I'm based out of Melbourne, Australia.
00:17And I run our Accenture's regional business
00:20for data and AI.
00:22I think the key sectors where we're seeing business leaders
00:26are what we call strategic bets, include areas
00:30like banking, insurance, health.
00:32Those three would be the top three sectors
00:34where we see a lot of action.
00:35OK, Dipanjan.
00:36You're kind of similar.
00:37There are four industries where we see a lot of traction.
00:40One is financial services, including
00:42banking and insurance, health care, retail,
00:45and manufacturing.
00:46There is a lot of actual work in manufacturing
00:48where they're using both generative and predictive AI.
00:50And are there any industries that
00:52are encountering specific challenges, do you think?
00:57Look, I think, surprisingly, the industries
01:00which are more regulated are actually using it a lot,
01:03contrary to what I think you would feel.
01:06Actually, if you see whether it's health, it's regulated.
01:09Financial services are regulated.
01:10I think it's mostly manufacturing industries,
01:13which are, if you think about the spectrum of AI,
01:17classical AI, which is in terms of, let's say,
01:20predictive, prescriptive modeling,
01:21has been around for a while.
01:23I think as we start to join generative AI to it,
01:27we see, for example, manufacturing companies
01:29taking a little bit different approach,
01:31taking time to be able to get it.
01:33But we are already seeing some initial leaders
01:35in that sector as well.
01:36And when you say taking a different approach,
01:38what are they doing that's different?
01:39Sure.
01:41The context is a little bit different.
01:43Asset-heavy industries have different types
01:45of business conditions in which they need to work.
01:48If you think about, to get maximum value out of AI,
01:51you've got to get your data sets ready.
01:53And for asset-intensive industries,
01:55the data sets could be very specific to the type
01:58of instrumentation they use, which is not the same as,
02:00let's say, financial services, for example.
02:02So I think those would be the kind of thought process
02:04which some of the clients we work with are going through.
02:08Zubanjan, if you're looking at your portfolio of clients,
02:12can you outline some of the specific obstacles
02:15that you think exist to unlocking the bottom line,
02:19as this session is called?
02:22I actually see three gaps, which I
02:25think our customers are kind of dealing with.
02:28One I call the value gap.
02:31There are a lot of excitement about AI,
02:33but translating that into business outcome
02:36is not easy for various different reasons.
02:38Sometimes there are not enough tools and platforms
02:42that people can use.
02:42Sometimes they don't have the data.
02:44Sometimes mapping the business problem into an AI problem
02:48is not easy for people.
02:50The second thing that I see, I call it confidence gap.
02:52There are a lot of people who are kicking the tires.
02:55They're doing prototyping, but they're not
02:57confident enough to take those applications, AI applications,
03:01or models to production, because they're not sure
03:03about the accuracy of it.
03:05They're not sure about the safety and other risks
03:08that are associated with those models.
03:10And third, of course, is the expertise gap.
03:14This is kind of a huge area of focus for a lot of people.
03:18There are not too many people who actually
03:20know how to build, for example, foundational models.
03:23But luckily, you don't need that many people
03:25to build foundational model.
03:26What you really need, people who can use these models
03:29to actually solve business problems.
03:31Even that, I think, require a lot of expertise and training.
03:35Other thing that I think is also going to happen,
03:38people are worried about, it's the dual of this,
03:40that AI is going to disrupt the skill mix in an enterprise.
03:45And people are worried about what
03:47impact it's going to have, for example,
03:50various different jobs that people do today.
03:53What is going to happen to those jobs?
03:54And there are new jobs which are going
03:56to be created by AI, which requires also
03:58skilling up the workforce.
04:00So what would be your advice to someone, a young person who's
04:03thinking about what to study at university?
04:05They want to thrive in this new AI landscape.
04:07What should they be doing?
04:08What should they be studying?
04:09I have one of those at home.
04:10My son just graduated high school
04:13and is a computer science major at UC,
04:16University of California.
04:18And he's a little bit worried what
04:20is going to happen to the coding jobs,
04:22because there are a lot of coding jobs which I think
04:25these AI co-pilots are going to do.
04:28But I do think it's also going to create new types of jobs.
04:32And where I think it's going to be probably
04:35most lucrative and productive in terms of skilling up
04:39is that when you can marry the AI skill with,
04:42for example, a specific domain-specific skill.
04:46For example, if you are an expert in biology,
04:49how to apply AI to do new medicines
04:52or personalized medicine would be a huge thing, I think.
04:55Even some of the other areas, for example, education,
04:58I do think personalized education will be a big thing.
05:01So how do you use AI to create those new industries,
05:05new use cases, new businesses?
05:07That's where I think going to be the most action,
05:09and that's what people need to learn.
05:11So that's what people should be thinking about
05:13as they map out their careers.
05:15I'm interested in the confidence gap
05:17that you described and the expertise gap.
05:19Vivek, is that something that you've witnessed as well?
05:23And if so, are there any demographic specifics to that?
05:26Is the confidence gap more prevalent in some parts
05:30of the world than others?
05:32Is it women rather than men?
05:33Are there any specifics around those factors?
05:37Go ahead.
05:39Thanks.
05:40I think, look, we are going through a large pivotal moment.
05:46I think not having the right talent in the right places
05:49is a challenge which every business is facing right now.
05:51We just can't find enough talent to go about doing
05:57what we call reinventions, thinking end-to-end.
05:59How do you really not apply this incrementally,
06:02but actually reinvent?
06:04Examples, we worked maybe a little bit sidetracked,
06:07but we worked with a food and beverages company
06:10in globally.
06:13And there, by using an AI-driven solution
06:15for marketing content generation,
06:17we could actually build one year equal worth of content
06:21in about eight days.
06:23And that is phenomenal in terms of productivity enhancements.
06:26But at the center of it is,
06:28how do you actually transform your workforce?
06:31To be able to really scale this,
06:33one needs to think beyond use cases to business capability.
06:38How do you think about a product launch?
06:40Versus a photo shoot in which you can apply something
06:43and so on and so forth.
06:44So I think in terms of the question you asked,
06:46I think talent is difficult to find.
06:49Companies have to think about what does it mean
06:51from a workforce transformation perspective.
06:53At Accenture, we specifically have worked a lot
06:56on what we call what the future of work is gonna be,
06:58starting from there, to be able to think through,
07:01so if that is the future of work,
07:03what is the type of workforce you need?
07:05And then be able to define which type of workers
07:07with what skill sets are needed
07:09and have structured curriculums
07:11on how do you actually transform the workforce.
07:13That kind of becomes quite key.
07:14And I can assure you today,
07:17I mean, we are trying to get talent
07:19in every market we operate in.
07:20Talent is limited.
07:21So it's a combination of getting the right set
07:24of basic capabilities and training them internally
07:26in your firm, actually moving people.
07:28I mean, conventional data versus modern data
07:30is a big area, for example.
07:31Because if you don't get your data right,
07:32how do you kind of really unlock value
07:34from AI and so on and so forth.
07:35So hopefully that gives some ideas.
07:38So in terms of confidence gap,
07:39I think it depends on the risk level
07:42of the use cases that you are addressing.
07:44I think the EU AI Act actually has done a pretty good job
07:48in defining those risk level.
07:50There are use cases where you probably should never use AI.
07:53For example, social justice and things like that.
07:56And there are use cases which are high risk.
07:58There are use cases that are moderate risk.
08:00There are use cases which are low risk.
08:02And it depends on the regulation that people have.
08:04And one of the things I tell people
08:06that the CIOs, when you talk to them,
08:11they are concerned about three things.
08:13That can AI create top line growth?
08:15Can AI save cost?
08:18And how do I stay out of jail?
08:19And stay out of jail is where the confidence come into play.
08:23And if you talk to the CIOs of regulated industries,
08:26that's where they are most worried about.
08:28Because it can not only create business risk,
08:30it can create reputational risk.
08:32Sometimes it can create many other type of risks.
08:35They can be on the wrong side of the law, for example.
08:38So what's your advice to the leaders in this room
08:41on bridging that confidence gap,
08:42getting to the other side of it?
08:44Yeah, I think the first thing to do
08:46is to define the risk profile of the use cases
08:48that you're working on.
08:50And depending on the risk profile,
08:51there are different steps that you take.
08:53For example, what kind of testing you need to do?
08:56What kind of disclosures you need to have?
08:58What kind of moderation you need to apply
09:00to the results produced by AI, right?
09:03So if you think through that in a systematic way,
09:05I think you can take care of the risk
09:08and figure out where you can take it to production
09:11and what actions you need to take if things go wrong.
09:14I'm going to open it up to the floor
09:15for questions in a minute.
09:17I've just got a couple more for you both.
09:20Dibanjan, you've been at DataRobot as CEO
09:23for about 18 months now, I believe.
09:25Is that about it?
09:26And when you started, the company was,
09:28it was in a bit of a rough spot
09:29that I think there'd just been quite a few layoffs.
09:32But you've really executed a turnaround.
09:34There's a new platform, 9.0.
09:36You're rolling out various integrations.
09:39Can you talk to us a bit about the role
09:40that AI has played in that turnaround internally?
09:45Well, I mean, you know, when I joined DataRobot,
09:49generative AI kind of was not the thing, right?
09:52So generative AI happened.
09:53I remember I was actually in a Morgan Stanley conference
09:57in San Francisco, and everybody asked me about generative AI.
10:00And the way back, the Uber driver quizzed me
10:03for half an hour on generative AI.
10:04I knew that generative AI has arrived.
10:07And so generative AI has played really a big,
10:11you know, part in turning around the company
10:14because there is so much opportunity
10:15that I see with generative AI.
10:17And what we have also done,
10:19we have been in AI business for a long time.
10:21We have been a pure AI company for about, you know, 12 years.
10:24But some of the things that we developed for predictive AI
10:27applies quite well to generative AI,
10:29especially when it comes to governance and observability
10:32and things like that.
10:33And what I have seen is that if the combination
10:36of predictive and generative AI, the intersection of that,
10:39you find the most value-added use cases for businesses, right?
10:44And I think of this in terms of the left brain
10:47and the right brain. The left brain is your analytic part.
10:49The predictive AI and the right brain
10:51is your communicative part.
10:52And when you put the two together,
10:54that's when you get the iPhone moment of AI,
10:56which is what has happened, right?
10:59Go ahead.
11:01Vivek, a question for you.
11:03So Accenture has committed, I think, $3 billion to AI.
11:10How are you going to ensure that that investment
11:13changes the way that you interact with your clients
11:16in a positive way, maybe accelerates the process,
11:19but you also maintain that high-touch human element
11:23that Accenture is so well-known for?
11:24What's the balance there?
11:25Sure, and thanks for going there.
11:27I think it's a very committed investment from Accenture.
11:30Our CEO kind of announced that in 2023.
11:33I think the key parts to that was,
11:35how do we kind of really work with industry
11:37and shape the industry's future?
11:39Part of that was, how do we build
11:41industry-specific solutions, functional-specific solutions,
11:45build a network of what we call Gen AI studios across the globe,
11:50be able to really help our clients
11:53see what's the latest across the globe
11:56when they come to such a center?
11:58The other part was training of our own workforce,
12:02trying to kind of do that.
12:04So largely, how do we keep humans at the center of this?
12:10If you look at most of the research,
12:11Accenture has published quite a bit.
12:13About half of the benefits come from automation,
12:15to the Bunjian's point.
12:17As you go from classical AI, which
12:19was all about data, to generative AI, which
12:21is so much about language.
12:23When you put it together, what you
12:24can do in terms of automation, like autonomous processes
12:27at scale, is amazing.
12:30However, as you kind of go towards putting this together,
12:36the rest half of the benefits are
12:37what we call augmentation-based.
12:39So you've got to start thinking not incrementally
12:41around processes, but you need to start thinking reinvention.
12:44How do you actually start to think in a new way
12:46how you would do customer care?
12:47And it's very, very different than looking
12:49at how you're doing it today versus how would
12:51you want to do in the future.
12:53For doing augmentation, you've got
12:54to start with putting humans at the center.
12:56Thinking about, in this new world,
12:59what is the role which machine plays
13:02to kind of really support humans and still get to benefits?
13:05And the benefits are not just automation, productivity
13:07benefits.
13:07The real benefits are around outcomes.
13:10Having different types of work, which is more interesting.
13:12Getting to what we call a 10x impact in client's business,
13:15and so on and so forth.
13:16So I would think that's how we're looking at how do we
13:18put humans at the center.
13:19At the center.
13:19But one quick question before I open it up to Bunjian.
13:22You mentioned before the iPhone moment for AI.
13:24I'd love to know from both of you
13:26what would be your personal iPhone moment for AI.
13:29Maybe it's happened, but maybe it hasn't.
13:31What would that look like for you?
13:32That light bulb, massive light bulb moment
13:34that changes everything?
13:36Well, I mean, it is when AI becomes pervasive in everything
13:40that we do.
13:43And I don't know if I'll call it iPhone moment,
13:45but this is an example use case which kind of fascinated me
13:48for a while.
13:50So we have customers who use classical AI
13:53to do various different kinds of forecasting.
13:55We have a grocery store customer.
13:57They own a lot of grocery stores in the US.
14:00And they use our models to predict, for example,
14:02how many mangoes they need to order in a particular store
14:05in California.
14:07And they have been doing that for every product
14:09in every store.
14:10Sometimes various different days have different models.
14:12And they have thousands of those models.
14:14And what they have done with generative AI,
14:17which is quite fascinating, now they
14:19figure out how many excess mangoes they are going
14:21to have in a particular store.
14:23And they are creating personalized coupons
14:25that they are sending out to people that they know.
14:28They're also creating, for example,
14:30very, very localized radio ads that they
14:33are putting in various different radio stations.
14:36For example, if they know that Indians like ripe mangoes
14:40and Mexicans like green mangoes, they
14:42will put a Hindi language ad in a local news station.
14:46And they are doing all of these things automatically.
14:48So pretty much everything that we do probably
14:51can be made more intelligent with the use of AI.
14:54And it's when it becomes pervasive and invisible,
14:56that's when you know that iPhone moment will happen.
14:59So no leftover mangoes.
15:01Vivek?
15:01That's a very interesting analogy.
15:03It's super interesting.
15:04Look, I would want to say that maybe last year, it
15:08was mostly about experimentation when we worked with businesses.
15:11But this year, we are seeing businesses really
15:13put their heads together to be able to scale.
15:15To be able to scale, you need to think about the competencies
15:18across.
15:19Having a large language model is a small part
15:22of that, which you use for doing whatever.
15:25There are normally seven things.
15:26We see leaders who do this well bring a lot of things
15:29together to be able to scale this up.
15:31So think about being on cloud.
15:33That could be one thing.
15:34Second is getting your data ready, which we spoke about.
15:37Third, think about the enterprise-grade AI solution.
15:40That's not just about the model.
15:42On the model, you have adaptation.
15:43On top of adaptation, you have consumption.
15:45So you've got to put that thing together,
15:47which becomes important.
15:48You also want to integrate that with your digital platforms,
15:52like your enterprise, the ERP platforms, the marketing
15:55platforms, the sales platform.
15:57So there are a lot of these things which come together
16:00to truly be able to unlock the value and be able to scale up.
16:03And I think that, to me, is where
16:05we're seeing a big difference from last year to this year.
16:08Any questions for Vivek and Abhijan?
16:12If so, raise your hand.
16:14If not, I will continue with one last question.
16:17I want to finish on the subject of timelines.
16:20So you've just laid out multiple steps, Vivek.
16:25If a client comes to you and says,
16:27hey, we're about to invest a ton of money now in AI.
16:30This is our priority going forward.
16:32What's a realistic timeline for them to see meaningful returns?
16:37Sure.
16:38And let me take examples.
16:40I think in the press, Accenture and National Australia Bank
16:44recently announced the journey we've
16:45been on with them for about one and a half years now.
16:48Through this process, they focused
16:49on how to reinvent using AI, generative AI,
16:53specifically in customer service and operational efficiencies.
16:57Think about 20-odd use cases, with about eight of them
17:00being enterprise-grade level.
17:01So that's about one and a half years.
17:03What I want to say is, I think the key thing is,
17:06it is a multi-year journey.
17:07Multi-year, OK.
17:08If you're thinking about reinvention,
17:10my encouragement to leaders I work with
17:13is to think beyond use cases.
17:15To be able to go at scale, you've
17:17got to move from use cases to business capability.
17:20And that takes time.
17:21That takes a lot of thinking.
17:22I would highly encourage all of us
17:24not to really get bogged down by incrementally
17:27changing the process.
17:28Investing in your own workforce is going to be very important.
17:31And then thinking through all these seven components
17:33we spoke about.
17:34But really, it is a multi-year journey.
17:37You've got to be thinking, how do you not
17:40do incremental changes to what you're doing today?
17:42And one final tip for the people in this room,
17:44if they're in companies that they want to really unlock,
17:49as we said, the bottom line going forward,
17:50what's the one thing they should go into their offices
17:53tomorrow and start thinking about?
17:55Well, I do think use cases and returning value to business
17:59is important.
17:59I mean, for GNI, we are in the honeymoon period.
18:02And that's not going to last very long.
18:04So it is important to pick the use cases which
18:07show near-term value, but also have a long-term plan
18:10so that you can show the value at scale, which I think
18:12is going to happen.
18:13But showing near-term value and showing
18:15some return on investment, showing some early successes,
18:17I think is very, very important.
18:19One final bit of advice, Vivek?
18:21Pretty similar.
18:21I think lead with value.
18:22That's what we do with all clients.
18:24You've got to be really knowing what you're solving for
18:26from a business perspective.
18:27This is not a tech for tech.
18:29After decades of technology lifecycle,
18:32you have something where the technology is
18:34adapting to how humans work.
18:35I think it's an exciting era.
18:37If approached correctly, there is so much
18:40to be done in terms of unlocking value.
18:42Vivek and Devanjan, thank you so much for joining us.
18:45Thank you for inviting me.