- il y a 1 jour
AI, Trust & the Future of Customer Experience
Catégorie
🤖
TechnologieTranscription
00:02Hello, hello, thank you.
00:04We're at the end of Viva Tech, so it's very exciting to see everybody.
00:09Welcome.
00:10So, we have a very energetic panel.
00:14I see some photos are already out.
00:16I mean, my goodness, what if we actually talked about some of the data?
00:20We will put the unstructured images into a model.
00:23Okay, so that was an AI joke.
00:25Sorry about that.
00:27It's actually the afternoon.
00:28So, I am here with three amazing leaders.
00:32My name is Lorena Yee.
00:33I'm a senior partner at McKinsey, and I lead our frontier tech research out in San Francisco.
00:38I am here with such an amazing panel.
00:41We have Mel Fruz, who is the global VP of customer science at LinkedIn and has a passion for how
00:48we think not only about data, but behaviors.
00:51I think when we start to mix data and art, we get some really interesting outcomes.
00:57We have Kaylin Voss, who's the EVP for Agent Force, all things AI at Salesforce, which has
01:04been really out on the frontiers of helping companies transform.
01:08She was also recently the CRO of Slack, a technology that many of us are very attached to in our
01:15workday.
01:16And Anka Marola, who is the chief digital officer of Sephora, who backstage, I mean, oh my gosh, what you've
01:25done in the last year, but over the last 10 years has been really leading how we think about data
01:30in refreshing and reimagining customer experiences and delight at LVMH.
01:36So this is a power panel.
01:38I also just wanted to mention that we're talking about technology here, and we have three, and I'll include myself,
01:45we have four fabulous women leaders.
01:47So I think we're going to change the gender curve right here.
01:52Maybe just to set the stage, and then I'm going to pass it over to these phenomenal women, is the
01:58title of this is really thinking about agentic AI.
02:01And so I just want all of you to think about your day so far, and have you had the
02:07emotional urge to have a pit crew, a team of people capturing the notes, capturing the photos, capturing your to
02:15-dos, capturing all the business connections that you've made?
02:19And wouldn't it be nice when you get on an airplane, maybe later today, over the weekend, all of that
02:25is filed, and actually some of those meetings have already been booked, and a synthesis of things that you could
02:32bring to your customers or your products already done.
02:35A couple of you are nodding, why do we care?
02:37This is the power of agentic AI, is not just taking amazing capabilities in terms of knowledge and information that
02:47we see in generative AI, but taking it a step further to take an action, to complete an action, to
02:53be your extended workforce.
02:55And I think each of these leaders are pioneering how we make that happen.
03:02But, we also have a challenge, which is, that promise sounds amazing, but how many of you have that in
03:09your hands today?
03:10And if you've been working with these technologies, with these capabilities for your businesses, you may even be experiencing something
03:18I like to call pilot purgatory, where you've got lots of ideas out there, but how many of them are
03:23actually reaching a level of measurable productivity?
03:27So, as a kick-off, I wanted to ask each of you, how are you leveraging agentic AI, not just
03:35to make things faster, better, but to unlock innovation?
03:39Mel, let's start with you.
03:41Well, it's great to be here. Thanks so much.
03:43So, at LinkedIn, we think about AI and agentic AI really as a tool, right?
03:48And for us, it's about something that can turbocharge the way that we create value for our members and our
03:55customers.
03:55And so, the way that we think about AI is we put it directly into our products so that we
04:00can put it directly into the hands of our members and our customers.
04:03So, for those of you who are members, thank you.
04:06You will know that we use AI to help you connect to opportunities, to showcase your skills and expertise, and
04:13also to gain access to knowledge so that you can do your job, right?
04:17And so, you will know some of our AI-driven products are things like AI-driven job matching, AI writing
04:24assistance, and also personalised AI coaching.
04:28And for those companies, we have our very first agent and AI was around the hiring assistant, and really this
04:36was designed to help take on a lot of the work that recruiters were doing that was repetitive tasks so
04:42that they could focus on their higher value activities, like consulting with hiring managers and creating these amazing candidate experiences.
04:49And for those marketers and brands in the audience, certainly agentic AI is foundational to helping marketers and sellers and
04:58buyers connect faster.
05:00And so, we know, for example, that creating campaigns can be really challenging, particularly if you've got small teams that
05:07can take a long time and when resources are tied.
05:09And so, we have a product called Accelerate, and this is an amazing product that lets marketers create campaigns in
05:16under five minutes, right?
05:18And it does things like provide recommendations for end-to-end campaign development, automatic optimisation so you can reach the
05:26right audience with the right creative.
05:30But ultimately, for us, it's about helping marketers become more productive, more efficient, for sure.
05:36But it's really about helping them become more creative with AI as well, right?
05:41And we really believe that that is inherently a human skill, right?
05:45And we think that AI can help marketeers really, like, tell their brand stories in entirely new ways as well
05:51as drive performance.
05:52So, those are a few ways that we're thinking about infusing our products with AI to help our members and
05:57customers.
05:58I mean, if you can develop a campaign in five minutes, I think you definitely leave more time for creativity,
06:03which I love.
06:04Kaylin, tell us a little bit about how you guys are thinking about this at Salesforce.
06:09Well, first, I have to say, we thank you for having us.
06:12This is a privilege.
06:13We have relied heavily with partnering with McKinsey while at Slack, now over at Agent Force, on just rethinking this
06:19whole transformation we've been through.
06:21So, thank you to McKinsey.
06:22And I'm on LinkedIn all the time, just as a consumer.
06:25I love the platform.
06:27And then Sephora, I mean, lifelong customer.
06:28How we are unlocking innovation, I would say, really starts with, I think, of being accordion, where you think big,
06:35but start small, like, bring it out, start simple, and then advance in maturity.
06:40And I'll give some examples of that.
06:42We want to be customer zero of driving our own transformation just so we can understand the same challenges that
06:48all of you are facing with the letdown of some of the AI not delivering on the ROI we all
06:54expect.
06:54So, we want to go through those pains and learnings together.
06:58And so, we want to be customer zero.
06:59We are using our own internal sales agents.
07:02We are using external sales agents.
07:04But I'll share some of these learnings along the way of driving innovation.
07:07You brought up this metric.
07:09I mean, it's 41% of our work.
07:10You'll probably all nod at this, like, on the drudgery of work, on repetitive tasks, on trying to just find
07:16things.
07:16And so, we use agents to centralize, to automate a lot of the pre-meeting work that is done, and
07:23then the post-meeting follow-ups, right, this pre-during, post-loop.
07:28And by doing that with Agent Force, you talk about what's the unlock in innovation.
07:32We hear a lot around AI efficiency, the efficiency it drives.
07:37So, we've seen, on average, a 36% improvement in our sales cycle, which is powerful, right?
07:43Like, hey, here's how we move faster.
07:45You don't have to find all of this information.
07:47We've pulled together this brief.
07:49That helps you go faster through a sales cycle, which is great for efficiency AI.
07:53But where the next point I have is be an optimist, because it also drives growth AI.
07:58And what do I mean by growth AI?
08:00It's also helping us improve our win rates.
08:03Hmm, that's interesting.
08:04How?
08:04Our win rates, since using Agent Force, and this is for our sales teams, is up 11%.
08:11And you think, why is that?
08:12Well, it's really tapping into you.
08:14Think of the long-term memory that companies have.
08:18You know when you can cross-pollinate what the best teams are doing, where you're losing,
08:23what's not working, what are the trends of your industry.
08:26And when the agent platform knows that and can surface that up, that delivers growth AI.
08:32And I think that's really powerful, and that's something we've seen.
08:35So, we are using agents to unlock innovation in that way, not just for efficiency AI, but growth AI.
08:39Tracking the numbers, because we don't have agents wandering around here, is four minutes for a marketing campaign, 11%
08:47win rate increases.
08:48That is sounding really good.
08:50And, Anka, your last year has been unbelievable speed.
08:54So, tell us about what you're doing on the innovation front, because we can all think of an experience at
09:01a store or online.
09:05Just to add, because obviously, I think everyone started from the employees, because that's where we were forced to do
09:10so, to make sure that our data doesn't leak.
09:13So, everyone started building internal chat GPT.
09:15So, obviously, we've seen the value as well, the same statistics of usage.
09:21However, then, we focused on our customer experience omni-channel, and we looked at areas where we can optimize with
09:27AI.
09:28And we immediately went after the stores, first of all.
09:31So, we do have now also chatbots for our BAs in the stores.
09:36We also have AI skin diagnostics that help you scan and recommend routines.
09:42And that's actually something that customers are very excited about, because it addresses a real concern and a real need.
09:49And I think that's the learning, is, you know, don't solve for problems that don't exist, when you have plenty
09:56of problems that do need solving today.
09:59And we've also went after the online experience.
10:02We have chatbots.
10:04We're experimenting with these for two years now.
10:07They are assisting you with your shopping advice, finding the right product for you.
10:13We also have, for example, a summarization of reviews.
10:18As a retailer, we have so much wealth of information from our communities, because we are lucky to have very
10:25passionate customers.
10:26I'm sure there's some in the audience, some Sephora passionate shoppers.
10:30And they leave lots of reviews, but it's hard to give back that intelligence of the crowd to an individual
10:37customer.
10:38So, summarizing the insights around the product, we've seen that that's really appreciated for our customers.
10:46I think, really, a big unlock that we're now focused in terms of efficiency is content creation.
10:53And I think everyone has seen the power of social media now and trends.
10:59You know, by the time you figured out that the trend is happening, it's gone in the next week.
11:04So, if you don't create content within that instant, then it's already passing.
11:10So, how do you keep that relevancy in a world that is so fast-paced?
11:14And I think this is where Gen.AI has been really useful for us.
11:18And honestly, we were very skeptical in the beginning, but we do have our Sephora collection, a content studio in
11:26-house.
11:27And we gave the creative team this challenge to test and learn.
11:31And so, what we've seen is we have generated 20 times faster content.
11:3820 times faster.
11:40So, something that would take you a month, now you can do it, you know, in a day sometimes.
11:44And so, that's really magical.
11:46The other insight from content has been that it was much more engaging on social media.
11:52So, content that we've posted on social media that is generated with AI has been more engaging.
11:59And so, we've asked ourselves, is it because the AI is just more creative than us?
12:03And actually, it turns out that our creative teams have great ideas, but they're very costly to produce sometimes.
12:10So, before they weren't able to put this, all of their creativity into practice.
12:16And now, with Gen.AI, they've really unlocked this.
12:18And honestly, no one expected this in the beginning.
12:20And that's really an area where we feel it really brings a lot of value for us right now.
12:25And we're doing many other things, but it would take half an hour.
12:28My goodness.
12:30So, I mean, that is an astounding number that you can do 20 times faster content creation.
12:37You have 11% better win rates.
12:40You've got four-minute campaigns.
12:42You've got completely different things.
12:43I feel a little bit exhausted.
12:45I also have to say, I feel like an underperformer.
12:48Like, I can't possibly do this.
12:51And that, I think, is my prompt to actually, for Mel, for you, because at the end of the day,
12:56we create this technology for humans.
12:59Feeling tired or maybe feeling at threat on my ability to be a marketer, to, you know, sometimes people hire
13:08McKinsey to do a segmentation.
13:10Sounds like you may not need us anymore.
13:12I'm feeling really insecure here.
13:14So, tell me, Mel, about in your job marketplace data, tell us a little bit about the journey for people,
13:20and particularly something we've talked about, which is moving from jobs gains, jobs loss, to the skills we need to
13:28succeed.
13:29Yeah.
13:30I mean, certainly, it's an incredible time of job transformation right now.
13:34In fact, about half or 44% of this year's jobs on the rise didn't even exist 25 years ago.
13:42All right, so that includes things like artificial intelligence engineer, which you would guess.
13:47It also includes things like workforce development manager, right, or the emerging role around the chief growth officer, right?
13:55But we really think that AI's influence on jobs is actually best understood through the lens of skills, right?
14:02And so, we really think that there's a lot that we can learn here.
14:07And interestingly, as AI ramps up, what we're seeing is that it's actually the human skills, the people skills, that
14:14are becoming increasingly important in the workplace.
14:16We actually see this in our data.
14:18We have data on over a billion members and 69 million companies.
14:23So, we really have a lot of data about skills, and every year we publish a report about skills.
14:29And what we found this year is that people's skills like communication, like adaptability, like strategic thinking, are actually some
14:38of the skills that are emerging as the top most in-demand skills across the globe.
14:44So, it's good news for us, the humans in the room right now, but certainly it's interesting to think that
14:49these are likely the same skills that are set to be the most resilient skills in the future, right, as
14:56AI continues to change our work.
14:59And what's really interesting, Kaylin, is, I mean, your product is more B2B, but it actually starts with skills.
15:06And so, maybe just geek out with us a little bit on how you guys built the Atlas skills engine
15:12and how, as a user, you know, I can relate to employing the type of agents I need.
15:18Yeah.
15:19So, when we look at agent force, we know we've got customers of all types, right?
15:24And so, when we looked at what we built, it was really two things.
15:27We've got out-of-the-box agents when you think about the skills needed for function, right?
15:31So, you can hire digital labor to put these agents in your sales team, your service team, your marketing function,
15:38right?
15:39And they are trained based on the most common use cases, templates that we see by function and industry.
15:44But we also wanted to be a platform to allow our customers, our partners on the agent exchange to build,
15:50and then using this reasoning engine from the millions and millions of transactions and volumes we've seen from around the
15:57world.
15:58When I think around, then, the human skills we need to work with these agents, right?
16:02Like, we've built the skills that are needed by function or by industry or our customers to build with the
16:07agents.
16:07But how are we re-skilling the people who are using them, who are building them?
16:12Because these are all new technologies.
16:14We have to be really intentional.
16:15It's close to my heart on re-skilling our employees, our companies, our partners, our customers around this.
16:22And if you look at, say, our internal transformation, I talked around our sales use cases, we put service agents
16:29on our website.
16:30And actually, around 85% now of our use cases are customers globally using agents for service, for case deflection,
16:38for handling common questions.
16:40That's majority of the use cases, as you can imagine.
16:43But then what happens to that capacity that was previously people handling those questions?
16:48And when we look, I think this is a fascinating insight that we saw.
16:52We used to have humans that were handling our website, the 72 million hits that we were getting a year.
16:58We put AgentForce right on that.
17:00Now, 86% of those interactions are being handled by agents.
17:03But if you look at the employees of Salesforce, they're actually going up.
17:07And what we did was we repurposed and re-skilled that capacity.
17:10And then to where you say, where are the skills needed?
17:13We're re-skilling people who maybe used to sit there and say, here's how you reset your password.
17:18Or here's how you set up AgentForce.
17:20To now, let's bring them with us on this journey, training them to be forward-deployed engineers,
17:26to have the AI and agent engineering skills that are required.
17:30So what we have done is re-shifted that capacity.
17:32And we then see it, which is critical, of the agents we have deployed kind of supercharging these humans.
17:39And that it's really the two coming together for success, but not a reduction in our headcount.
17:44And so I'm an AI optimist.
17:46That's my other, I've got be an accordion, be an AI optimist.
17:50It's exciting to see the two come together.
17:52So I'm feeling a little less nervous.
17:54Hopefully all of you are as well.
17:57And all of that sounds great because it feels like it can play to all of our strengths.
18:03Let's take a specific job in mind.
18:05Anka, you mentioned the BAs, and those are the beauty advisors, right?
18:09Those are the folks who are in the stores.
18:12They are my shepherd.
18:13When I need, you know, concealer or rouge, they figure out, you know, what I need.
18:18How, like, tell us, flash forward, you know, two years, three years.
18:23Tell us the edge of innovation that you would imagine with the beauty advisors
18:28and kind of a preview of what our experience would be like.
18:32And honestly, I think the future is really the seamless experience omni-channel.
18:39And it's about making sure that you innovate, but you connect the dots to remove any sort of friction.
18:45And really what the customer wants is an exciting experience and advice to find the right product for them.
18:52And I think, for me, the opportunity that we have is to empower our BAs with more advanced and sophisticated
19:00tools
19:00so that, A, they can do their job faster, but, B, to also have a more accurate read
19:07and diagnose more accurately the concerns that the customer might have
19:12and then bring that intelligence also online so that when you're shopping, for example,
19:16I know your skin shade because you came in the store, I scanned you, I have your skin shade.
19:22When you go online, you should already have predefined on your product page.
19:26Here's your selected shade for this product.
19:29That experience has to come together omni-channel.
19:31Otherwise, you know, it's much poorer, I would say.
19:35And to me, I think, for example, on the e-commerce front is where I see a lot of disruption
19:41coming up next
19:42because, frankly, for the past decade, the web and e-commerce has not really changed.
19:48And it has been very hard to create that magical experience.
19:53And I think with Gen.AI, we're going to see a shift from product catalog to real experiences.
19:59And so what we want to do is truly create this flagship experience also online.
20:07And so what does that mean is when you go to the store, you're welcomed by a BA who talks
20:13to you
20:14and then you go and explore the products that are relevant to you.
20:18And I think that's the future.
20:19It's really the reinvention of the UX online that allows you to have, similar to discussing with a BA
20:27and then exploring that experience that is tailored to you in a much more fun, seamless, and magical way.
20:35And that's really, to me, the future.
20:37And it's true that today what we're doing is we're putting a chatbot on the side.
20:43But that's not where it ends.
20:45The whole UX will change.
20:47And it will be similar to how we move from the BlackBerry to the iPhone.
20:52We still do the same things.
20:53It's still a phone.
20:54So it has the same functions.
20:55But it's much more enjoyable.
20:57It's much more fluid.
20:58And it's really just easier.
21:00And it's a real disruption.
21:04I could not agree more with what you're saying, where it's going from just what were kind of previously
21:09these very deterministic bots to now totally changed UX experience, multimodal, omni-channel.
21:16And what's insightful about this, when we look at our customers who are doing this
21:19and making it that you can have a conversation, which we've all gotten used to as consumers
21:24with each other, but with brands, we actually see the customer lifetime value go up from
21:29anywhere between 10% to 30%.
21:31Once you really can deliver exactly what Sephora is, removing the friction, being self-service,
21:37being personalized, giving those recommendations, if you can do it successfully, then you see
21:42an increase in the customer's lifetime value.
21:44They want that with you.
21:45They just want it frictionless.
21:46And I think it's a really powerful point.
21:48It's exciting to see what you're doing.
21:50And, I mean, you know, interesting piece of data is, you know, I always had this conception
21:55that it's maybe a leadership or it's the employees are holding us back, you know, this concept
21:59that maybe they don't want to move forward.
22:01But in fact, actually, in a lot of research that we've run, we've seen that employees are
22:06most eager to get the training and the skills and the access to be able to use these tools.
22:11We also see that they are most likely amongst all people to already imagine that 30% of
22:18their work will be different and replaced by AI over the next two to three years.
22:23So what is your, like, pro advice to become an expert?
22:28Like, how can employees empower?
22:31Like, what skills?
22:32Mel, maybe starting with you.
22:33What skills?
22:34Like, how can people in the audience listening think about being able to be at the top of
22:39their game as these things change?
22:41Yeah, I mean, I think what's interesting that you shared is this bifurcation between leaders
22:47and individual contributors across workplaces.
22:51And actually, what we're seeing that's quite promising is that even though business leaders
22:55have been slower to adopt, they are upskilling in AI literacy.
23:00And I think that that whole space of AI fluency and literacy is certainly the table stakes for
23:07organisations, but when we look at our data in the last two years, for example, we see
23:12that the C-suite is three times more likely to add AI literacy skills to their profile than
23:18they did a couple of years ago.
23:19And actually, they're 1.2 times more likely to be adding AI literacy skills than the rest
23:26of the workforce, which is super interesting.
23:29And so I think that certainly leaders leading the charge is one part.
23:33I think fluency is super important for individual contributors across the teams.
23:39But certainly, we're seeing so much strength in groundswell applications of AI from teams
23:45across multiple functions, across different regions.
23:50And really, we encourage that we surface those examples and use cases get spread more broadly.
23:56But if I think about the role that leaders can play here, I think it's more than just improving
24:01their own AI fluency, they really need to prioritise this across the entire enterprise, across
24:07the entire company.
24:08Because at LinkedIn, what we're seeing is that AI really is redefining work.
24:13And that means you need new playbooks.
24:16And Anka, I mean, given your role, people will look to you as kind of the source of truth.
24:21How are you doing that at Sephora and broadly at LVMH?
24:25Since I arrived, it's really the question that I get asked all the time inside Sephora as
24:31well.
24:32I remember my mom worked in a hospital.
24:36And this was in post-communism Romania.
24:39It was actually, we didn't really have computers very soon.
24:43So I actually remember distinctively when they were all panicked because they had to learn
24:48how to move from paper to computers.
24:51But these people didn't really know how to work with computers.
24:55And so to me, I really remember distinctively that angst and that fear, you know, even though
25:01today it seems so natural for all of them to know how to use a computer, an email, and
25:05so on.
25:06And to me, it's really the same parallel that we are faced with now.
25:11It's, we have the paper, which is the way of working, and we have to force ourselves to
25:16learn these new tools because it's going to save us massive amount of time afterwards.
25:21And so it's not enough.
25:22I think one thing that we have done at LVMH is we've created the LVMH Data Academy that
25:28is upscaling with different trainings, all of the mesons and executives and so on.
25:35But that's not enough.
25:36You really have to appropriate yourself and task yourself with developing a new reflex.
25:42And so that's what I would encourage everyone is test it, use it.
25:47You know, a CEO, if you are a CEO, force yourself to use ChatGPT or the likes every day as
25:53your
25:53first reflex.
25:54And it's really a new muscle that we have to develop.
25:58And once we do, then we start to understand.
26:00If you don't do it, you know, someone who's using paper, it would be very hard to understand
26:04what can we, you know, project yourself, how computers may change your life.
26:08So it's the same thing for me.
26:10I also think people could start trying to be coders.
26:13I mean, actually, the barriers now have gone down so much.
26:16And Kaylin, you mentioned something at the beginning.
26:18You said customer zero.
26:20And I mean, you have an enormous workforce at Salesforce.
26:24How are you bringing this forward?
26:26Oh, my goodness.
26:27So we talked about being an accordion, right?
26:29We've talked about how to be an optimist.
26:32And this one is where we look at how do we be an agent boss?
26:35And what is an agent boss?
26:37I mean, that sounds good.
26:38Is that something I can find on LinkedIn?
26:40I'm going to be like profile.
26:42Agent boss.
26:42Be an agent boss.
26:43And so as customer zero, we've all spoken to this point, but top down and then bottoms
26:49up.
26:50And so there have been situations when we've been helping our customers roll out their agents
26:55if one requires fine tuning or it's not delivering the accuracy that's needed, somebody will
27:01say, hey, we need a forward deployed engineer.
27:02I'm like, put me on that call.
27:03Oh, we got a forward deployed engineer.
27:05And I'm like, no, I want to know how to tweak.
27:06Hang on.
27:07Slow down.
27:07I'm going to ask something.
27:08And then, you know, you can call Dial Mel for help on this.
27:12You've mentioned forward deploy engineer a lot.
27:14That is not a role that I have seen college kids applying for.
27:18Can you explain what that is just in case others may not know what that is?
27:22This is when we talk about the AI, the engineering skills.
27:24These are the hands-on skills required for the future of AI.
27:28When we look at how do we tweak little variables that can really change the output of these
27:34agents, these are our forward deployed engineers.
27:36And what's fascinating is you see them almost as an extension of product.
27:40You put them into your customers to help understand what are their needs, and you come back and
27:44you build what your customers need.
27:47And so I share this with my team, which we talk about top down, bottoms up.
27:51We're all forward deployed engineers because we're all learning so much together.
27:55I want to sit on those calls.
27:56I want to see how we're tweaking the context variables.
27:59I want to show that it is important for all of us from the top down and then bottoms up.
28:04And what I also, we talked about this internally.
28:06If I see people using external tools, I love it.
28:10Because I want to see what's working.
28:12I want to see what's bringing in value, whether it's Notebook LM or deep research, you name it.
28:18I want to encourage that and not gate employees to very specific tools right now.
28:24Because you want to incentivize creativity and exploration.
28:29So it sounds like you have a program.
28:31We run our whole Agent Blazer program and want to take that externally.
28:35but re-skilling everybody.
28:37And we want to be customer zero of all being forward deployed engineers.
28:42So, I mean, we are now at a pretty optimistic place.
28:46You all have basically showed me that it's going to go faster.
28:48There's all these applications and you're already implementing it.
28:52But it doesn't foot with the macro data in the following sense.
28:56Which is, we know that 92, 93% of companies all say they're investing more in AI over the next
29:03two or three years.
29:04So that's good.
29:04Everybody's budget goes up.
29:06But the challenge is, is those same companies self-describe, you know, my statistic is 1%.
29:12I've seen more generous statistics at 10%.
29:14But these are very low statistics of actually having it at scale.
29:18And without scale, this isn't actually in our business.
29:23It's not something I can get at every Sephora store or, you know, every interaction with Salesforce.
29:29So, I mean, give us the real real in terms of how, as leaders, are you traversing to reach scale?
29:38What's the advice for that?
29:41Maybe I'll just address one point that we see is,
29:44whilst a lot of companies aren't reaching the point of maturation that you're talking about,
29:49one thing we do see in our data is that some of those AI experiments are paying off.
29:55And so we released our work change report recently,
29:58which showed that 51% of businesses who adopted generative AI
30:03are reporting a 10% plus increase in revenue.
30:06So the business case for change is very strong.
30:10But you're right.
30:10The workforce adoption remains an area of opportunity.
30:15And I certainly believe that we need to go beyond efficiency plays
30:20for broader adoption of our employee bases.
30:24Certainly, our employees need to see that this drives meaningful business results.
30:28Sharing those case studies becomes really important to driving that broader adoption.
30:33And certainly, that's a space that a lot of companies are trying to figure out right now.
30:36But on your point of scale, certainly, this is something that we think about at LinkedIn.
30:41And we believe that responsible AI can't be something that is a filter at the end of development.
30:48Rather, it needs to be foundationally part of the entire process.
30:53And so for us to scale our AI product efforts, we actually take a principles-based approach.
30:58We have five principles for how we govern all of our AI development efforts.
31:03That's around ensuring we're advancing economic opportunities so that we only build things that are actually there to help our
31:09members achieve their goals.
31:11It's about upholding trust, ensuring that we're proactively addressing security, privacy and safety in everything that we do.
31:18It's about promoting fairness and inclusion.
31:21And that means that we need to get all of our cross-functional and diverse teams building together and designing
31:27not just the solutions,
31:28but actually also the guardrails.
31:30That's the only way to ensure that our AI tools are proactively addressing discrimination and bias.
31:37We also need to promote transparency in how our AI models work.
31:41And then finally, in order to really scale these efforts at an organisation our size with a member-based and
31:49customer-based our size,
31:50we actually need to embrace accountability.
31:53And for us, that means employing pretty robust AI governance frameworks to ensure that our products remain trustworthy.
32:01And for us, the principle-based approach means that we can continue to build products responsibly,
32:08but that we can do so at scale.
32:11And importantly, across it all, trust is the cornerstone.
32:16I couldn't agree more with trust when you think of what are the barriers that we're most frequently seeing to
32:22driving success at scale,
32:24or why is it that AI hasn't delivered on the ROI that people have been looking for.
32:29I think there's some common themes that we see.
32:31One of those is clearly going to be trusted agentics.
32:34Are you following the permissions of what that person is able to see of the right data, right?
32:40The moment you have a breach in trust, you do not want to interact with that brand anymore, or with
32:44AI in general.
32:45And so trust is foundational.
32:47I'd say some of the other patterns we've seen that are preventing to your question around scale,
32:51and then how do you address it so you ultimately can, is the right data foundation, right?
32:56The right canonical model, semantic layer, that if you ask a question of data, or of AI,
33:01and you don't have your data house in order, it's only going to be an output of that data that
33:07you have.
33:07And so you think of, I'll give some examples, if we have customers saying, hey, I'm in Paris,
33:13how many opportunities do I have in Paris that could be available to me?
33:18I'm just giving an example.
33:19If your canonical model isn't aligned, and you have that data mapped up in your metadata layer as leads instead
33:25of opportunities, right,
33:26there's a common vernacular that our customers are really having to pause and get their data house in order ready
33:33first,
33:33so that AI can deliver on the value because you've built the right foundation.
33:37That's a big piece of scale, right?
33:39And then how do you properly do the ingestion of that data?
33:44And then the other thing I would say for scale is we really learned at first,
33:47and I'm sure you've seen this with Sephora and LinkedIn as well,
33:51we threw probabilistic technology, right, into deterministic scenarios.
33:58And deterministic scenarios, especially with use cases, say, in healthcare or flight patterns,
34:04there's not a lot of creativity on those.
34:07Like, there's some black and white answers.
34:09So we've also learned to go into scale how to match the product with the right guardrails based on use
34:14case,
34:15whereas creative campaign in real time where you catch the opportunity,
34:19that you can have a little more probabilistic what's happening in the moment.
34:22And so we're also having to tweak, we talked about the data foundation, right,
34:26but the product itself to match the right guardrails for the use case to help drive scale as well.
34:33Love it. Okay.
34:34For me, it's first thing is really focus.
34:37If you need to focus on few priorities that really mean a lot to you, to your customers,
34:44and are sizable enough to pursue them relentlessly because it is a process of test and learn.
34:49And then second, obviously, to complement what you said is, it's not, I'll give you the example of our chat
34:56bot.
34:57It's not hard to put a chat bot online, but what's hard is to make it have your personality,
35:02have your brand, tone of voice, have the curation that you want to embody.
35:08And so that's really the time that it takes to make it truly a state of the art.
35:14And I think for me, that's the risk that we're facing now in terms of brands is how do we,
35:21when we all use more or less the same foundational models,
35:24how do we preserve that uniqueness of our brands, of our company going forward?
35:30And that's what's really taking a long time in this process of scale.
35:35I love it.
35:36So we are over, but I'm going to force a quick lightning round of personalization,
35:40which is we are all learning every day.
35:43I mean, I think that is the most amazing thing.
35:45So what's your one thing that you would encourage people to try on AI?
35:51I'm going to start with you, Kaylin, because you mentioned a couple of things earlier.
35:54One.
35:55Like one, yeah, one feature.
35:57Like you mentioned Notebook LM.
35:58Like what would you put after that?
36:00If someone were going to play with AI.
36:02Well, obviously I lead Agent Force and Data Cloud.
36:04I love Agent Force.
36:04But I would say one of the tools I use most, I'd say there's a couple, is Slack Search.
36:11You can't name your own product.
36:13Sorry, I did not define the rules properly.
36:15I love Notebook LM.
36:16Okay.
36:16Non-product.
36:17I was trying to go outside of Agent Force.
36:19Slack Search is powerful for all your conversations, your context.
36:21But I love, I do like Notebook LM a lot.
36:24I do.
36:24And deep research.
36:25Gemini.
36:27Mel, what do you like?
36:27I think, without naming products, use cases, I think like get in and generate some images.
36:35It's like such a great way to like test your own prompt engineering skills because some products are better than
36:43others.
36:43And I think it's just a great fun space for experimentation.
36:47Lowers the bar.
36:49Anka?
36:50For me, it's really, I think it's really the chat GPTs and the like because that's how you really develop
36:56the new muscle.
36:56And that's how you can really project what the future of commerce can really be once you see the power
37:02and the difference of how we use it versus other channels.
37:06So, to me, it's really developing that daily muscle with all sorts of problems so you can understand the breadth
37:12of it.
37:12And I find it to be the most educational, frankly.
37:15Wow.
37:15Well, with that, thank you so much, ladies, for this.
37:19And hopefully that made everybody feel a little bit more optimistic about how we deploy agents in our workforce and
37:27lives.
37:28Thank you.
37:28Thank you.
37:29Thank you.
37:31Thank you.
37:32Thank you.
Commentaires