Brainstorm AI Singapore 2024: Reinventing The Customer Experience With AI
Joon-Seong LEE, Senior Managing Director, Center for Advanced AI, Southeast Asia Lead, Accenture Sami MAHMAL, Data Lead, Zurich Insurance Sinisa Nikolic, Director, HPC-AI, CSP, Lenovo Asia Pacific Moderator: Clay CHANDLER, FORTUNE
Category
🤖
TechTranscript
00:00Okay, welcome, everyone.
00:02I'm Fortune's Executive Editor for Asia, Clay Chandler,
00:05and we're here to discuss
00:07reinventing the customer experience with AI.
00:10Great topic, and we have a fantastic group of speakers.
00:13Before we get started, I just wanna thank
00:15our partner, Accenture, yay, for sponsoring this event.
00:20We really appreciate them hosting
00:22this particular session as well.
00:25Some quick housekeeping, a reminder
00:27that this session is on the record.
00:29Additionally, when we turn to audience questions,
00:32please know that you will need to raise your hand
00:33and someone will come by with a microphone.
00:36With that, let's start the introductions from the panel.
00:38Remember, this is meant to be a group discussion.
00:40We look forward to incorporating
00:41as many voices as we can today.
00:44So on our panel, Jung Sung Lee, Senior Managing Director
00:48for Advanced AI, Southeast Asia Lead, Accenture.
00:53Then we have Sami Mamal, Data Lead for Zurich Insurance,
00:58and Sinisa Nikolic, am I saying that correctly?
01:01Nikolic. Nikolic, all right,
01:03there we go, Director, HPCAI, CSP, Lenovo, Asia Pacific.
01:09So I'm gonna start off with an example
01:12that I shared the other day with Jung Sung
01:16because he works a lot with Changi Airport.
01:20I'm a huge fan of Changi Airport,
01:23and for the story that I wrote for the next issue
01:26of the magazine about Singapore's AI strategy,
01:28I spent some time with the folks at Changi Airport.
01:30And the former CEO at Changi tells this great story
01:33about how when they rolled out Terminal 4,
01:36and they used a lot of AI and they automated it,
01:38and they were very excited about what they were doing
01:42with all their new technology.
01:44And they wanted to test it out
01:46by demoing it with their carriers.
01:48And they said that they went first to the Japanese carriers
01:51because they thought that the Japanese carriers
01:53were especially noted for their great
01:55kind of customer passenger service
01:58and how they use technology,
02:00and they wanted to impress them first and get some feedback.
02:03So they spent half a day walking them
02:06through all these systems, the Japanese carriers,
02:09and they expected them to be completely blown away
02:11by what they'd done.
02:12And he said that once they were done,
02:14he was surprised that the response
02:16was not as positive as he expected.
02:18It was kind of interesting.
02:21He couldn't figure out what their perspective was,
02:24and so they kind of sat him down and had a few drinks.
02:27And it turned out that the Japanese felt that
02:32some parts of the terminal had been too automated.
02:35And they said, you know, in Japanese,
02:37we have this idea of omotenashi,
02:39where we take care of the customer.
02:40We like to show them that we love them,
02:43we fawn over them, and we have to,
02:45so what we do is in the highest value touchpoints
02:49in our operation,
02:50we try to push people forward into those points
02:52so they can interact and face the customer.
02:55And in a way, we worry that taking those people away,
02:59even though it might be more efficient, is rude.
03:02So that's, I think, a central kind of question
03:04to think about for, and Xiaohong said that he,
03:08they made some, they rethought a little bit
03:10some of the things that they were doing
03:11and how they allocated their people.
03:12But it's an interesting perspective
03:15on whether AI can help,
03:19whether it is some cases, in some cases,
03:22can have a lot of risks,
03:24that it gets in between companies and their customers.
03:27And I hope we can sort of dive into that
03:29as one of our opening themes.
03:31Do you wanna wrestle under any of that
03:33as someone who has been,
03:35I mean, I should say Accenture
03:36has been very closely involved.
03:38I wasn't expecting to go dive straight into that,
03:41but anyone from Changi Airport here,
03:43I don't want to steal the time.
03:44We are very fortunate to work with Changi Airport Group
03:48for the last six, seven years.
03:50We had this program that's called DIVA.
03:52You can Google them.
03:53It's all in the public domain.
03:54Whatever I share here is only the public domain.
03:56DIVA stands for Digital Innovation Venture Analytics.
04:00But the context of why I share this with Clay
04:02was against the backdrop of GNI disruption that's happening.
04:07So the question was that, like any evolving technologies,
04:11how do you sustain innovations?
04:14How do you harness the technology
04:15so that it's not a one-off implementation?
04:17You put GNI, install co-pilot, you're done.
04:21That's definitely not the case
04:22because every day there will be new announcement,
04:25new development around large language models, right?
04:27Meta just launched something.
04:29Next week, there will be something else.
04:30So the technology is changing.
04:33But how do you upkeep?
04:34And therefore, how do you then,
04:37how organizations serious about GNI
04:39are harnessing this opportunity, organize themselves?
04:43That's the context of the discussion.
04:44And if I look at Changi, for example,
04:46for many years, they have the DIVA program.
04:49The DIVA program was set up
04:50as like the challenger to the status quo.
04:53Think about you creating this group of people,
04:56highly talented individuals,
04:57a combination of the Changi Airport talents
04:59and leveraging, in this case, Accenture talent
05:03to create this DIVA group of people
05:05and get them really excited about trying these innovations.
05:09And one of those innovations that I remember
05:10was quite interesting.
05:11Think about like the conveyor belts
05:13that we are all used to.
05:13We go to the airport and they send people
05:17to maintain the conveyor belts based on a timetable.
05:21Nothing wrong with that, right?
05:22Every now and then, you go and check
05:23and make sure everything's okay.
05:24But this team came up with a smart idea.
05:27What about a smart luggage?
05:30Think about you creating this luggage
05:32and putting all the sensors into the bags
05:36and then let the bags just travel through the conveyor belts
05:39and picked up all the acoustic signals
05:41from the conveyors and therefore change
05:43the way you do predictive maintenance.
05:46That's the result of the DIVA program.
05:49So I think the context is that as we look at GEN-AI,
05:52look at all these disruptive technologies is evolving
05:55and looking at how you then harness this technology
05:58on a sustained basis,
06:00and I look to DIVA for inspiration.
06:02That's what I thought is worth sharing with this group.
06:04No, look, I wanna be very clear.
06:07Changi is my favorite airport
06:08and it's not for nothing that it is-
06:09It has to be the case for now on.
06:12Year after year, this gets rated
06:13the best airport in the world.
06:16And they have a lot of challenges
06:18that they're grappling with,
06:19one of which is that they can't have processes
06:24that require too many people
06:25because they've got limits in the labor supply.
06:28And so it's really about getting that combination right
06:32and figuring it out.
06:33And the things that they're doing both in baggage handling
06:36and everything from that to the customer experience
06:40when you're shopping, the retail experience,
06:42it's clear that Accenture has given them
06:45a lot of great support and helped them
06:47to do a fantastic job with that.
06:50They're the envy of kind of other airports
06:52around the world.
06:55Let me ask Sammy if I could about banks,
06:58financial services, insurance companies.
07:01I deal with a range of different banks,
07:04some local, some global, personally.
07:06And sometimes I find that banks' efforts
07:10to embrace AI are helpful,
07:12and sometimes they're not.
07:16And sometimes, we talked a little bit
07:19about productivity earlier in the afternoon.
07:21I feel like the question trees
07:27and these kind of automated voice,
07:28automated answer systems maybe save the banks some money
07:33by getting rid of a call center somewhere
07:35or a customer service representative somewhere.
07:37But it's costing me money because it inevitably turns out
07:40to take a lot of time, be less efficient.
07:43And I am sure that this is particularly
07:45in the kind of private wealth space.
07:47This is a thing that has to be thought through
07:49very carefully to keep from alienating
07:53important customers.
07:55How do all these issues look to you
07:57as data lead at Zurich?
07:59So if I can, from a lens of the insurance perspective,
08:05so we see Gen AI, so we as insurers,
08:09we've been using AI for many years already.
08:12We're not discovering AI.
08:13I think that's what's really changed for the last years
08:17is that now AI is more accessible.
08:19So there's a lot of technology that is easy to implement.
08:22Like for example, when we all talk about GPT,
08:24two years, four years ago, no one knew about GPT.
08:29No one knew about the previous models of GPT.
08:31GPT or GPT models were not released
08:33for the first time two years ago.
08:35Actually, the first earlier version of GPT
08:37was already out there.
08:38So I think that there was one thing
08:40that really changes the access to the technology.
08:43And this democratization of the technologies,
08:45also there is a bit of impact to the culture.
08:50Everyone now knows that, hey, you can generate
08:52or you can interact with natural language.
08:54So what we see from an insurance perspective
08:58is that there are multiple opportunities,
08:59but these opportunities are transformative
09:02to the way that we do our business.
09:03So coming back to your example
09:05on the Japanese customer experience,
09:06we don't want it to have automation everywhere.
09:08So usually when you have to deal with your insurer,
09:11you have to deal with the moment of the claim.
09:12So it's not necessarily a very good moment.
09:15So when we know that,
09:16so we call it the moment of truth for us.
09:18So how can we make sure,
09:21if I talk about this specific moment,
09:23how can we make sure that this is going very well
09:25with the customer, the customer gets out of it
09:27with a very good experience.
09:28So it's not only about purely automating.
09:32We cannot just say that,
09:32hey, we want to fast track this
09:33and the customer does not even have to deal with this.
09:36It's how can we make this experience
09:38very good for the customer,
09:40reducing the time so we don't have to chase us
09:42for the next one week, two weeks
09:44before you can get your claim reimbursed.
09:46So how can we make it,
09:47I can give you a very clear and very fast answer.
09:49Actually, we see that the fast answer is better
09:51than a complete claim reimbursement,
09:55but it happening in four or five months time.
09:58So this is something that we see from a data perspective.
09:59So the question on how we are taking it is,
10:02how can we utilize these different technologies
10:05to improve specific,
10:08and we can touch base more specifically
10:10on some concrete example
10:11on how we are using GenAI specifically.
10:14But really, we are looking at how we can,
10:17how the business, how the claims teams,
10:19how the underwriting team can utilize this technology
10:22to not only accelerate the processing,
10:25but really make the...
10:27Give us an example with maybe a specific type of...
10:30Yeah, I didn't want to jump into the example,
10:31but I can give you some concrete example.
10:33Like for instance,
10:35one use of the search for GenAI
10:36start with AI deep learning.
10:39Just last year, we deployed for Indonesia market.
10:43Before that customers, before issuing the policy,
10:46the car motor policy,
10:48they had to wait for the insurer to come to their home,
10:52take pictures, assess the car,
10:54before they can issue a policy.
10:57So can you imagine, you just bought your car,
11:00it's a second hand,
11:02you have to wait one week before Zurich
11:05come into your place.
11:06And this is actually a regulatory requirement.
11:07It's not Zurich imposing it,
11:08it's actually Indonesian market imposing it.
11:10You need to go and assess the vehicle condition.
11:14So you had to wait a couple of days,
11:16can go up to two weeks in some specific regions.
11:19And you have to, well, I'm talking about the carbon waste,
11:22because someone has to drive up to your home,
11:24align with you and ensure
11:26that he can assess the vehicle condition.
11:28So how we utilize AI more largely
11:31and GenAI to a certain sense to impact this process
11:34is that now customers takes pictures of the cars.
11:37There is an automated process to assess the damage
11:39and automatically accept or refer the policy
11:43for further assessments.
11:44So now we switch from a process where we had to wait days
11:48and have a manual assessment
11:49to something that's happening in a couple of minutes.
11:52This is kind of multimodal.
11:54They're sending images, right?
11:56Yeah, it's taking pictures.
11:57And actually, it's not beyond sending images,
11:58it's also taking the location
12:01so that we avoid all of the fraud.
12:03So it's taking a lot of information,
12:05and it's even going up to the pictures and ultimately.
12:08So, but reimbursing an insurance claim
12:09is an interesting transaction
12:12because there are steps that go beyond
12:15just the accident and the assessment itself,
12:18because then you've got to figure out, okay,
12:20how much should the reimbursement be?
12:23And that involves, okay,
12:26we figure out how serious the accident is,
12:28what parts are involved here?
12:29What's the cost of those parts in that particular area?
12:34How do we decide what the service charge would likely be
12:38in the proximity of the claimant?
12:41Do you use AI for all those kinds of calculations as well?
12:45We are doing a lot of tests on this area.
12:50So I can talk about other countries,
12:51but for example, I talked about the policy issuance
12:53for motors in this case.
12:55So we are assessing to accept or whether or not
12:56we want to accept the policy issuance.
12:58Now we are doing the same for a claims perspective.
13:01So I just went into a damage.
13:04I take pictures, and now we are using the same technology
13:06to identify what other damage,
13:08to which associated panel, the damage,
13:10what is the intensity of the damages,
13:11and then it's not an AI exercise anymore,
13:14just looking at the database of your cost.
13:16So this is actually one of the initiatives
13:18that we are currently doing.
13:19So interesting.
13:20And are you building your own technology,
13:22your own models to do this,
13:24or are you using ChatGPT or some other large language model?
13:29So what I was saying earlier is that
13:31what's really changed for us is the access
13:33to the technology.
13:34So we don't have to build everything.
13:36So we are using a mixture of many models.
13:39We're using GPTs to get some of the insights
13:41on the pictures because now,
13:43I mean, you're all playing with ChatGPT 4.0.
13:46When you take a picture, you can now do,
13:49you can do the exercise, take a picture of a car
13:50and say, oh, can you assess the condition?
13:53Can you see any damage?
13:54We are playing with this a lot.
13:55So this is only one part feeding to our models.
13:58We are using other third parties as well
14:00that have deep learning models,
14:02specific models that can identify damages.
14:04And then we are adding layers, in-house layers
14:07with our own functions that are dependent
14:09to the countries or specific on the market.
14:11Sinisa, let's come to you and how does,
14:16tell us a little bit about your role
14:17and how you're grappling with these kind of issues.
14:20It's an interesting one.
14:20So I have responsibility for high-performance computing.
14:24So when you talk about large language models,
14:26you're doing billions of points of data analysis.
14:32That's our systems at the back end to do that.
14:36I also have responsibility for AI,
14:38which covers enterprise AI.
14:41So just outside of the large language model,
14:43the training aspects, the building of the systems,
14:46these are the, maybe I'd call it
14:48the inferencing end of the world.
14:50How do you then take that data and do something with it?
14:54And then I also have responsibility for our cloud business,
14:59where a lot of the LLMs are running today.
15:01So do I grapple and do our clients?
15:04Better off to talk about how our clients are doing
15:07and how our clients are grappling with these things.
15:09We ran a global CIO questionnaire
15:15and we turned it into what we call a playbook.
15:17So a CIO playbook for AI.
15:2196% of the respondents, the CIO respondents,
15:27are going to make a decision on AI.
15:30Half of them, however, said we have no idea
15:35what that decision, 54% to be accurate,
15:38had no idea what that next step was going to be.
15:43So what are we doing and how is Lenovo
15:47sort of tackling these things?
15:48A little bit like Accenture, I guess.
15:50Our consulting business, which is,
15:52we call in Lenovo SSG, our SSG business,
15:56we have set up multiple centers of competence for AI.
15:59We call them AI discovery centers,
16:01populated by data scientists, computer scientists,
16:05and consulting folks.
16:09And so we will then go out with the client,
16:12we'll go through what is the KPR,
16:14what is it that you want to achieve,
16:16and not once talk about hardware,
16:18not once talk about a SKU, not once talk about storage.
16:21Is it what is the end result?
16:23What is it you want to achieve?
16:25Is it efficiency?
16:26Is it less downtime on the manufacturing floor?
16:31Is it an increase in NPS scores for client satisfaction?
16:35What is it that you want to do?
16:39And then we've invested over a billion dollars
16:41across AI in Lenovo.
16:43And some of that billion dollars went into ISVs, software.
16:49We don't have software solutions from that perspective,
16:52but we partner.
16:53So we've driven into the ISV space
16:55so that we can optimize, tune, and understand what they do
16:58so that we can position well with our clients
17:01to either modify those codes and that application suite,
17:05modify those use cases,
17:07so we can deliver a decent outcome for a client.
17:11The hardware aspects and all of those sorts of things,
17:13we do this every day, this is what we wake up to do.
17:16That stuff is easy.
17:19That stuff's easy for us.
17:22And it should be easy enough for a client.
17:24We try to simplify the process,
17:27we try to simplify that 54% of folks
17:30that really don't know where to go next
17:32and how to implement it.
17:34Really interesting.
17:36I want to open this up to a conversation from everybody.
17:39But before I do that,
17:40I'm going to ask one more question to Jun Seong,
17:46which sort of came up in Sammy's comment.
17:51There's that kind of amazing breakthrough moment
17:53in, what is it, November 22,
17:55when suddenly the number of ChatGPT users
17:58went from zero to 200 million in two months.
18:03And it was really unprecedented
18:06in the sense that without much,
18:11it now appears that even without much knowledge
18:13or understanding from the OpenAI board,
18:16a decision was made to just toss this thing out
18:18into the wild where consumers could get it
18:21and play with it and see what it could do.
18:23And people were amazed.
18:25And so now we've, for almost two years now,
18:29this thing has continued to grow in the number of users.
18:32People are getting the hang of how to use it.
18:35They're getting very savvy about prompting.
18:37And they're seeing that it can do
18:38really powerful, amazing things.
18:41So what does this mean to companies?
18:44Because you talk about the companies that are saying,
18:46oh gosh, I don't really know what to do next.
18:49But consumer expectations are rising quite rapidly
18:54because this technology has been thrown out there
18:55where everybody can play with it.
18:57Can you talk a little bit about that
18:59and what pressure it puts on boards, CEOs,
19:02to kind of reform the way they interact with customers?
19:05Clay, this is an extremely important topic,
19:07which is why I guess we are here.
19:10And some people call it the iPhone moment
19:12when GeneAI arrived for the sheer disruptive nature
19:16of this technology.
19:17Because for the first time,
19:18the world has the ability to harness
19:2180% of the world's unstructured data.
19:24Unprecedented, democratized.
19:26Now the one dimension of this GeneAI phenomenal
19:29is really about how powerful this technology is.
19:32But there's one other element that we always overlook.
19:35You think about when iPhone arrived,
19:37it changed the way we use phones.
19:40The same thing that Chechnya has done to us
19:43and changed the way consumers expect from companies
19:46and how we interact with information.
19:49I have to share a personal example.
19:52I was navigating in one of the bank's website the other day,
19:55trying to figure out how to solve a problem
19:56to link one of my account to another trading account.
19:59I couldn't find an answer through the website.
20:03I went to Gemini.
20:05Gemini told me one, two, three.
20:08I solved it immediately.
20:09I went to the bank chatbot.
20:11The chatbot couldn't answer me
20:12because the chatbot couldn't understand my question.
20:16So think about it for a moment.
20:19In Accenture, we use a way to frame this.
20:21Think about the traditional search and librarian model.
20:25Librarian means that you go to the website,
20:27you navigate, you look for stuff.
20:30To an advisor model, you get the answer right away.
20:35You go to a place, you're not searching for answer.
20:38You want the answer.
20:40Now think about now how,
20:42if you are a consumer-facing business,
20:44whatever businesses you own that has a customer element,
20:47how will they then expect to interact with you?
20:51That is the big challenge.
20:52That's the big aha moment.
20:53That's the iPhone moment,
20:55not in the context of how disruptive this technology is,
20:57how it will fundamentally change
21:01what consumer expects from companies.
21:04From anything, any information,
21:07how you consume to whatever, how you produce stuff.
21:09And that's how destructive this is, right?
21:11I think, if I can add, if I can,
21:15everything you say is wonderful, perfectly accurate.
21:19I think the other part from the consumer side,
21:22when you start talking to the board,
21:23and one of the biggest things
21:24that you need to talk to the board about
21:27into the C-suite is around governance.
21:31How is and how will AI be governed in the organization?
21:36And that comes down to a number of facets, right?
21:39The biggest one is truthfulness
21:41and the truthfulness of the output.
21:43How do you actually govern
21:45against these hallucinations that we have?
21:49Otherwise, we get into trouble, right?
21:51The user gets into trouble,
21:52the organization could get into trouble,
21:54dependent upon the use case, of course.
21:56So I think that's one of the largest considerations,
21:59I think, for any board,
22:01is to really look at that piece, right?
22:03Yeah, I mean, that's clearly
22:05gotta be top of mind for boards,
22:06but I do think that this countervailing pressure
22:09of consumers learning to expect more.
22:12I mean, if you think about a,
22:14I mean, a lot of you probably use tribal sites
22:16to book trips, and I've tried a variety of different ones,
22:22but it's a little bit frustrating
22:24because I can now get on ChatGPT,
22:25and I can say, hey, I'm going to Singapore for a week,
22:28design an itinerary for me.
22:30And you might say, I'm going on a family business trip,
22:33or on a family trip,
22:35and it would design a different itinerary for me
22:38than if I were going on a business trip
22:40where the company was paying, right?
22:43I'm interested, you know, here's my situation,
22:45I have two or three kids, you know,
22:47we want to be able to go to Universal, or whatever it is.
22:51You just drop those parameters in,
22:53and the average travel site,
22:55if I go to like, you know, Agoda.com or something,
22:59cannot do that.
23:00It's the search function,
23:01it's where you have to really hunt around,
23:04is that the one I want?
23:05So you and the next generation of consumers
23:07all learn to expect that now, right?
23:09So which is why I think,
23:10going back to the question about board,
23:12and I have the good fortune of interacting with many boards
23:15over the last one year.
23:16And one of it is really to help them
23:19understand the opportunities
23:20so that they can invest right,
23:22and also the risk.
23:23Now let's talk about the opportunities.
23:25The key thing that I wanted to bring across
23:27to this audience is also that
23:29there has to be a clear demarcation
23:31about what this gen AI technology can do
23:34in terms of helping you to improve
23:35your business performance.
23:37That's okay, that's given.
23:39But think about this is the re-invention opportunity.
23:41This is the opportunity to completely change
23:43the way you do things,
23:44you completely change the way you,
23:46information is being shared.
23:48You think about years ago,
23:49we were working with a bank in the US
23:51and they were trying to change the way
23:53customer transact by using voice to text and text,
23:57using the text to chatbot to say,
23:59can you transfer my money from this account to this account
24:02without going through the app?
24:03Think about it for the moment.
24:05This has, this can become a reality now
24:07with the power of large language model.
24:09So we are trying to get them board to understand
24:11and CEOs to understand that,
24:13that don't take gen AI as any of the technology evolution.
24:16This is different, this is not incremental.
24:20And therefore we have been going around
24:21and obviously encouraging companies
24:24to think about re-invention, enterprise re-invention.
24:27In fact, don't look at one function only
24:30as piloting in a contact center
24:32or doing some pilot,
24:34of course you go through the pilot phase,
24:36but scaling it in the context of re-invention.
24:38That's what we are.
24:40Also hearing from the clients
24:41who with the ambition to leapfrog
24:44are seizing this opportunity to do that.
24:48Questions, comments from the floor?
24:50There must be some.
24:51Yes, can you stand up and,
24:53we got a mic back there.
24:57We got one, quick, quick.
25:00There we go.
25:01And tell us your name and affiliation if you can.
25:05So my name is Sian, I'm a venture capitalist.
25:08And I think both of you or all of you
25:12have enough experience, right?
25:14GPT has been around for a couple of years now.
25:17And I'd like to try to bring some,
25:21for lack of a better word, substance, right?
25:23And numbers, right?
25:24To the whole equation for the audience here.
25:26You've participated in a large number of projects.
25:29So let me make a hypothetical problem for you easy.
25:34So that we can try to get some numbers out of it.
25:38If I'm in a, we talked about banks, right?
25:40So let's say I'm in a medium sized bank right now, okay?
25:44I've got a lot of structured data, right?
25:47Now, I want exactly like you to have my customers
25:52operate, right?
25:53With a chat interface that actually works, right?
25:57Where most chat interfaces,
25:58at least in Southeast Asia don't work at all, right?
26:01But I need to restructure all my data.
26:04I got to invest inside vector databases.
26:06I probably need to restructure
26:08all my internal operations, right?
26:11Probably have to fire my CTO as well
26:13because I need a 30 something year old.
26:15I'm not being facetious.
26:17I'm being honest, right?
26:18And I will then need to dictate what the parameters are,
26:23right?
26:24Through this whole exercise,
26:25before I even start using LLMs, right?
26:29To take all the retrieval augmented data and all.
26:33Can you give me, or us,
26:36a rough number of a return on investment
26:40for this approximately year and a half long process?
26:43Ooh, tough question, tough question.
26:45That is a tough question.
26:47It's tough because some of those parameters are missing
26:51for us up here to give you that.
26:54What is the net, as an example,
26:56what is the net saving
26:58or the net that you would gain
27:01from an increase in client satisfaction
27:04or an increase in efficiency?
27:07If there's a dollar amount and all of these things
27:10on any client value proposition or value proposition,
27:12which turns into a return on investment,
27:14will drive a result, right?
27:16You need a starting point.
27:17Pardon me, no customer satisfaction.
27:20Just tell me in efficiencies what I will save.
27:25I can tell you from the manufacturing floor.
27:29I'll give you a Lenovo example
27:31and then I'll pass it on because I'm talking too much.
27:34Lenovo eats its own cooking, right?
27:37We've implemented AI across our supply chain
27:42from beginning to end,
27:44and we've been doing this now
27:46for the last three, four plus years,
27:49part of this $1 billion investment that we're making.
27:53We've increased efficiency
27:57in certain aspects of our supply chain
27:59to the better part of 80 plus percent.
28:02Now, when you talk about supply chain,
28:04go back to our horrible COVID days,
28:08that painful, painful experience for everybody.
28:11That increase, we sustained.
28:14So this is going back
28:15to the other thing about sustainability.
28:17We've been able to sustain that
28:19and increase that slowly over time
28:21to the point that on,
28:23I think it was the latest Gartner survey,
28:25Lenovo's supply chain is top five in the world
28:30and I think one of the top two or three in Asia Pacific.
28:34What does that mean?
28:35We reduce our cost.
28:37We can pass that cost onto our client,
28:39better gross profit margins for us as a business.
28:43I can't give you the amount in millions
28:46or tens of millions on what that return,
28:49but you can infer what that would mean
28:52to us as a business.
28:53Efficiency, less stock on hand,
28:55less stock on hand, better for the P&L.
28:57We move it out, et cetera.
29:00And particularly now,
29:01if we start talking about technology
29:04and what's happening,
29:06AMD, Intel, NVIDIA,
29:09these changes are happening every six months, right?
29:11Moore's law no longer exists.
29:14Moore's law.
29:15Can I press you a little bit on this point
29:17because in some ways,
29:18the situation for Lenovo or manufacturer
29:21is somewhat different than a medium-sized bank,
29:23which is kind of the hypothetical
29:25we're thinking about here.
29:26Because you can make cost savings
29:27by making the supply chain more efficient,
29:30figuring out where you store things,
29:31just in time,
29:32all these kind of very physical
29:34manufacturing and logistic kind of things.
29:35But if you're a bank,
29:38the savings are likely to be
29:40in lowering the labor costs, right?
29:42Yeah.
29:43So, one more point over here.
29:45This is more about predictive AI,
29:46which is four years back, right?
29:49So you analyze the past data,
29:51you put AI,
29:52you make them all more efficient.
29:54What about savings through generative AI?
29:55So if you can cover both.
29:56Predictive as well as generative,
29:58the panel,
29:59if you can cover both savings.
30:00So I'll provide a comment.
30:02Generative, more specific.
30:02Yeah, I'm trying to answer
30:04by not directly answering
30:05what consultants do.
30:08Anyway, I think I just gonna quote
30:10the research that we did, right?
30:11We look at banks specifically,
30:13154 banks globally.
30:16But the question was more about
30:17asking the executives to project
30:20what the return is going to be.
30:21So it's not looking back
30:23because no one can tell us
30:24that they have done generative AI
30:25for the last 10 years,
30:26therefore they could prove this is the results.
30:28So every numbers are wrong.
30:29But directionally,
30:30hopefully you can see that
30:32what they are projecting
30:33average 20 to 25%
30:35in terms of uplift
30:37from a profit before tax standpoint.
30:39That is what they're looking at.
30:40Now, when you start breaking down,
30:41there will be like marketing, finance,
30:43not happy to share the research.
30:45I think what is however important
30:48is to have proof points.
30:49So the proof points are true
30:51all this experimentation
30:52that people are doing.
30:53We work with one of the financial services client.
30:55We put in a knowledge management platform
30:57to support the sales functions.
30:59Immediately the sales uplift went up significantly,
31:01productivity went up.
31:03So that is validating the business case.
31:06You see where I'm going with this.
31:07You cannot prove,
31:08nobody can prove to you,
31:09no consultant can tell you for sure
31:11it's going to be 25%.
31:13But directionally,
31:14it looks like it's heading that way
31:16because from the experiments
31:17that have been done
31:18and some of the scaling that happened,
31:20that's pointing towards the same direction.
31:23I hope this is helpful for you to think about.
31:26So I want to come to Seth here,
31:27but just quickly,
31:29the Wall Street Journal
31:29had this really interesting editorial
31:31a couple of months back
31:33where they sort of made the distinction
31:35between using AI as a painkiller
31:40to reduce costs
31:41versus as a vitamin
31:43to improve the quality of the product.
31:46And I guess there's a bit,
31:48I mean, with the bank example,
31:50we've stipulated,
31:51I only want to look at the hard things,
31:53the cost reduction part of it.
31:54But you can also use this technology
31:56to improve the quality
31:57of the customer experience
31:59and the product.
32:00And I think that there is
32:02lots of complicated debates
32:03about how to project those values.
32:06Yeah, massive.
32:07I'll just give you one example.
32:08Look at digital marketing.
32:09We're working with a client
32:10on digital marketing
32:10and hyper-personalization space.
32:12In the past,
32:13they can only run 400 to 500 campaigns
32:16in a month.
32:18Now they can do massive
32:19because they use GDI to generate contents,
32:22to do hyper-personalization.
32:23They can do thousands of campaigns.
32:26So you think about the productivity gain,
32:28you can tell from that
32:29that directionally,
32:30they're heading towards that direction.
32:31But of course,
32:32when you can do that,
32:33your competition can also do that.
32:34So it gets to move the,
32:36you know, you raise the bar, right?
32:37You know, eventually I think then,
32:39you know, it's who stay ahead,
32:41right?
32:41Of the game.
32:42Yeah.
32:43When I was at IBM,
32:44four years ago,
32:45I commissioned a study
32:46from Forrester
32:47to understand how much
32:49you could get the cost.
32:50Would you say that Seth
32:50is the former chief AI officer at IBM.
32:54To see how much you could get the cost
32:56of a call down,
32:58you know, customer service call center down to
33:02by implementing chatbots.
33:04And this is pre-generative AI,
33:05so four years ago.
33:06You could get,
33:07if you got it to a one-time resolution,
33:09every call you got down
33:10to one-time resolution,
33:11you reduce your cost of a call center
33:13down to $4 per call.
33:16Whatever your cost is now,
33:17you can get it under $4,
33:18and that was in the US.
33:20So you can calculate your cost that way.
33:23And I want to continue
33:24on this same topic.
33:26So from an insurance perspective,
33:27you know that our business
33:28is really operational cost heavy.
33:31So we manage emails,
33:33we manage message on WhatsApp,
33:34omni-channels,
33:35and we manage phone calls.
33:37So the estimate,
33:38I'll give some numbers
33:40to give you an estimate
33:40on how we're looking at it
33:41from an insurance perspective.
33:42So from an operational cost perspective,
33:44we are imagining an impact
33:46between 30 up to 50%
33:48on operational costs
33:49by using chatbots,
33:51but not only chatbots,
33:52but even proactive chatbots.
33:53You know that today,
33:54there are technologies
33:55where you can monitor the call real-time,
33:58and it's proactively listening
34:00to the conversation
34:01and reaching out for content.
34:02So we have today interfaces
34:03that we are testing,
34:04whereby the customer is speaking,
34:06and real-time you have a sidebar,
34:07which is giving all of the proactive clicks
34:10of all two actions.
34:11Oh, customers will be maybe interested
34:13by this specific product,
34:14you want to trigger this.
34:15Oh, by the way,
34:16you have this relative information,
34:18we already picked up the date of birth
34:19and the benefits
34:21that the customer will be interested in,
34:23and we already pre-coded the benefits
34:25or the policy
34:27that he would be interested with.
34:28So this is where we are able
34:30to reduce the TAT time
34:31of the conversation.
34:32So very simple,
34:33how can we reduce,
34:34if I'm able to manage the conversation
34:37on the one interaction,
34:38I'm able to do it proactively
34:40and close the conversation
34:41instead of spending 15 minutes
34:43to do it in 10 minutes,
34:44then you get translation
34:45in your operational costs.
34:46Well, this,
34:48I will not say it's only Gen AI,
34:49as you said,
34:50it's a complete transformation
34:51on the whole way
34:53on how you're managing your operations,
34:54your IT,
34:55because your data needs to be structured,
34:56everything needs to be there.
34:57So this is how we are looking at it
34:58from an operational perspective.
35:00From a customer experience perspective
35:02and the uplift to the top line,
35:03it's a much more tricky one,
35:04how can you correlate the TNPS uplift
35:07to the top line increment?
35:09We did some tests actually,
35:12but coming back to the example
35:14of the personalized marketing,
35:15today we're sending personalized notification
35:18to the agents,
35:19looking at the portfolio,
35:20giving them insight,
35:21and we are actually able to do it
35:22even on a daily, weekly basis
35:24to motivate the sales.
35:25And we are seeing some really good impacts.
35:29So instead of spending our times
35:30with the managers
35:31and calling in all the financial advisors
35:33and re-motivating them
35:35against their objectives,
35:36how they were doing for the previous week,
35:38what are the policies
35:39that are more likely to lapse
35:40and how to retain them?
35:41It was a very heavy effort.
35:44Now with Gen AI,
35:45you can feed in the data
35:47and put some narrative around it
35:48and sending this smart, personalized nudge.
35:51And we are seeing some impact
35:53to the top line.
35:53So this one,
35:54we are seeing some impact
35:55on small cohorts
35:56because it's not working on everyone,
35:58but up to plus 10%.
35:59But I hope that this can give you some insight
36:01on how you can nudge
36:03or move your top line and bottom line,
36:05which is a very big topic for insurance
36:06because we are still asking,
36:08and I think that you all are here
36:09wondering the same,
36:11what can Gen AI add to my top line
36:13and bottom line?
36:14Which really, it's not only about Gen AI,
36:16it's only around IT and process change.
36:18What were the CSAT scores pre and post?
36:21What were the customer satisfaction scores
36:24pre and post the exercise?
36:25So that's a very good question.
36:27And we are seeing that the TNPS is improving
36:31in the areas where we are testing this.
36:32But what's interesting
36:34is that the TAT time is not reducing.
36:36So sometimes the TAT time
36:38is actually increasing,
36:40but we are able to close the case
36:42in one conversation
36:43instead of having to say,
36:44well, you know what?
36:45We'll come back to you.
36:46So the TAT time, actually,
36:47it's a bad KPI for us
36:50to look at the ROI of our investment.
36:54Thanks, everyone, for the panel today.
36:56And I'd like to echo the gentleman's point.
36:59Tell us who you are and your affiliation.
37:01Okay, hi.
37:02This is Wan Wei.
37:02I'm an angel investor
37:03in generative AI and web-free startups.
37:06And I also run my own community
37:07helping solopreneurs and SME
37:09defined as less than 10 employees currently.
37:12So I echo with the gentleman over there.
37:15He's doing VC.
37:16I'm like helping people
37:17with their cashflow businesses.
37:18So top line is really important to us,
37:20especially with solopreneurs
37:21for under 10 employees.
37:23So my question to you is,
37:24what do you personally think
37:26that gen AI will help
37:28these small businesses
37:29or solopreneurs even in the next,
37:32like, say, five years?
37:33Really keen to hear your thoughts.
37:35Anyone want to tackle that?
37:37Small businesses and solopreneurs.
37:40I think what Jung-Wang was talking about before
37:43around this massive reach
37:46and individual EDM work, right?
37:49This marketing, this massive reach,
37:50this ability to hit your consumer
37:53or the market increases dramatically
37:57with generative AI.
37:58There's no question about it.
38:01How you position your product
38:02or their product directly,
38:04this personalization,
38:06which is this next,
38:07the next generation
38:10or the next step
38:11is this ultimate personalization of material,
38:14whatever that may well be.
38:16Whether you have an AI PC in front of you
38:19or your telephone,
38:20how does that data get personalized
38:22in the case of these small businesses?
38:23How do you actually market broader?
38:26More customers, more potential,
38:29more pipe, these types of things.
38:31I think that is,
38:32I think that's the biggest hit rate here, right?
38:35How do you do that?
38:36Guys?
38:37I can give my perspective
38:38for small businesses
38:40that do not have the capacity
38:41to invest in transformation and IT
38:44or get in a head of transformation
38:46or head of AI.
38:47For me, it's the integration
38:50of the Gen AI services
38:51across the different solution
38:52that you're already using.
38:54Copilot Microsoft is a good example.
38:55Today, everyone is using Gen AI with Excel,
38:58with the one that have the license of Copilot,
39:00but you already have a gain of productivity
39:02by using this for your day-to-day.
39:04And we are seeing today
39:05when you look at the offerings
39:06of the solution that you're already using,
39:08I think there are really rare cases
39:10where your SaaS vendors
39:11are not integrating a Gen AI component.
39:13So I think that Gen AI
39:15will start being integrated seamlessly
39:18into the solution that you're already using
39:20for SMEs and they will not have to invest.
39:22And actually, even for the bigger part,
39:24it's very tough for us to find
39:26the sweet spot of usability,
39:27the use case and where to invest Gen AI.
39:29And I think that's a lot of specialized SaaS
39:32are actually doing it very well.
39:34And it just appears out of nowhere.
39:35It's like, oh, actually this is a cool feature
39:37and it's actually Gen AI in the back working on it.
39:39So this is where I think that small businesses
39:41will have, will get the most value.
39:44I think we've probably got time
39:44for how many, one or two more questions?
39:47How are we doing?
39:48Okay, anyone else?
39:51Yes.
39:58Hi, I'm Mike Walker from Auth Gen AI.
40:01We're seeing more and more about talk
40:03about sort of like neural chips and devices.
40:07Like Chrome and having like Gemini models built in
40:10that you can use from the console.
40:13I'm interested in what your take of sort of what to expect
40:16in terms of like a timeline of things moving
40:18from sort of like LLFs running in clouds
40:22and they just end up moving more to kind of running
40:25on devices and what your expectations are
40:28of sort of like people wanting to be able
40:30to run things locally or latency
40:33if I just want to be able to like delegate stuff
40:35off to like bigger stuff in the cloud.
40:38I can kick it off, I guess, just because I'm a hardware guy.
40:43So we've announced these AI PCs, our ThinkPad AI PCs.
40:48And that is really generative AI at the cold face,
40:53at your point in the world that you're talking about.
40:57And that really is this personalization.
40:59It will learn as you use your laptop, your system,
41:04your device, whatever you want to call it.
41:06It learns your daily habitual routine
41:10and all of us have one.
41:12You open your laptop at nine o'clock or eight o'clock
41:14or 6.30 or 5.30 in the morning for me,
41:16what is the first thing you do?
41:18I'm checking my email, I look at my calendar,
41:20I do all of these sorts of things.
41:22Having that on flip open, it goes, right,
41:25this is what you will do, this definite personalization.
41:30And then as that next step is,
41:35what do you do throughout the day?
41:36So, Sammy, you're talking about co-pilot,
41:39you're doing those sorts of things.
41:41It's then setting that up for the day,
41:43looking at your calendar, doing that interaction.
41:46And that's sort of now, somewhat infancy, just announced,
41:51but we'll see that progress over time.
41:53And you'll see that device work or that device level
41:58get smaller and smaller.
42:00So laptop, I mean, we're starting super computes,
42:02then workstations, now down, it'll shrink and it'll shrink.
42:08And you're starting to see a little bit of that
42:10with some future announcements
42:11on our next generation of think phones.
42:14And that's where it's gonna go.
42:16And AI for me is not artificial intelligence,
42:21it's augmented.
42:23And I've been saying that for 10 years, right?
42:26When I worked with Watson, it's augmented intelligence.
42:30It's there to help, not replace, ultimately, right?
42:36And so that's what you're gonna see,
42:38I think, from the device perspective.
42:42Got a question here.
42:44Hi, I'm Shishesh from Spark.
42:47So the question is,
42:48the AI is impacting each and every department, right?
42:51So marketing, like the creatives you are seeing,
42:54personalized creatives, it's impacting finance,
42:58it's impacting other department.
42:59And then there's a huge IT infrastructure chain
43:02that may come through with the progress in AI,
43:05in terms of the hardware and all that.
43:08So, and for our friend from Accenture and other guys,
43:13how are you bringing the whole team together,
43:16the heads, the finance, the marketing, the IT,
43:20to implement the whole thing organization-wide?
43:22So how's been your experience?
43:25What were the challenges you faced?
43:27I think that's a brilliant question,
43:29because this is no longer about,
43:32as you look at companies now, I think in 2023,
43:35a lot of companies are looking at experimenting
43:39with technology, and now it's looking at scaling.
43:41And when it comes to scaling,
43:42I think it's so critical to get right,
43:45in terms of putting the multidisciplinary team together.
43:47So if I can label this like a Gen-AI COE of sort, right,
43:51in that context, that's what we see a lot of clients doing,
43:54is that very quickly form the team,
43:56whatever you call it, a COE, a task force,
43:59but to have top-level sponsorship
44:02to bring this multidisciplinary team,
44:04from finance, from marketing, to come together,
44:07and aided by, of course, the data scientists
44:10and data engineer that you have,
44:12to look at those strategic bets
44:14that, as a company, you need to make with Gen-AI.
44:17I called out strategic bets because you want to make sure
44:20that those use cases you pick
44:21are the ones that move a needle.
44:23You can install Copilot for your meeting notes.
44:26That's not going to move a needle for your company, right?
44:28Then you can say you have Gen-AI,
44:30but it's really looking at the core of the business, right?
44:32If you are into insurance and banking, it's underwriting.
44:36How you look at underwriting,
44:37that will transform how you do underwriting.
44:39If you're in a telco business, for example,
44:40different use cases and all that.
44:42So this team is in a way to answer your question.
44:46What we see clients doing,
44:47and where we are fortunate to have our clients,
44:49is to set up this task force.
44:51And I know some of these task force
44:52are actually championed by CEO themselves.
44:56I was with a bank CEO just last week,
44:59and we spent six hours thinking about
45:00how to orchestrate this reinvention.
45:03He himself sponsoring this.
45:05He spent six hours with the team,
45:07bringing together a multidisciplinary team.
45:09The technology guys is there,
45:10the head of functions is there,
45:12to look at every single opportunity to reinvent.
45:14I mean, that's how you orchestrate something like this.
45:16Yeah, so-
45:17Journey is a long journey.
45:18Correct.
45:19Because AI will keep on changing.
45:20Correct.
45:21And what about change management?
45:22Correct.
45:22Because a lot of people change management.
45:23Yeah.
45:24Okay, you almost finished and completed my sentence,
45:26which is the last, the second part,
45:28which is that, in fact,
45:29the way I want to leave you with a framework
45:31to think about if I may, right?
45:33You start with leading by value,
45:35which is where you put your strategic bets.
45:37You look at your digital core.
45:39It's where you put all the infrastructure,
45:41technology together.
45:42We call it a digital core with security.
45:44You look at the way of working,
45:46which is your change management.
45:48You look at responsible AI
45:49so that you can address risk, governance,
45:51and finally, look and think about Diva.
45:55Then we started the Changi,
45:56how you sustain the innovation.
45:57You need all these five things to come together
46:01to sustain and to be able to sustain these innovations
46:05and to drive reinvention,
46:06and that's what you need.
46:08It's a great long-term framework,
46:10and I think it's a good point to end on.
46:12We're right out of time.
46:14We've now got 15 minutes to break
46:18to get to our main stage events,
46:21which will resume at what time?
46:24Exactly 1.30.
46:25So thanks everyone very much,
46:27and please join me in thanking our panelists.