2977: AI-Driven Marketing and Personalization at Scale With OgilvyOne

[00:00:01] [SPEAKER_01]: I have never wondered how technology-like, generative AI is reshaping the world of marketing,

[00:00:08] [SPEAKER_01]: specifically in the realm of personalization.

[00:00:12] [SPEAKER_01]: Well today on Tech Talks Daily, I'm joined by Ravi Powell and here's the Global CTO of OgilvyOne,

[00:00:21] [SPEAKER_01]: which is a vertical and WPP company, and we're the career at the forefront of marketing technology

[00:00:27] [SPEAKER_01]: services. Ravi brings deep insights into the opportunities and challenges that gener AI presents

[00:00:34] [SPEAKER_01]: in delivering tailored marketing that only captures attention, but sustains it. So,

[00:00:39] [SPEAKER_01]: big question, how can businesses leverage this technology to create those truly personalised experiences

[00:00:45] [SPEAKER_01]: that resonate with their audiences? We're currently producing something like 30 to 35 episodes,

[00:00:52] [SPEAKER_01]: every single month's reaching around about 130 to 140,000 listeners around the world.

[00:00:58] [SPEAKER_01]: So I wanted to mention the sponsor of Tech Talks Daily and also shine a light on how legacy managed

[00:01:05] [SPEAKER_01]: file transfer tools are dated and lack of security that today's Rho workforce demands.

[00:01:11] [SPEAKER_01]: And companies that continue relying on that outdated technology put their extensive data at risk,

[00:01:16] [SPEAKER_01]: but the good news is you can ensure top tier security with kiteworks MFT suite.

[00:01:21] [SPEAKER_01]: Their software security hardening includes an ongoing bounty programme, regular penetration testing

[00:01:27] [SPEAKER_01]: to minimise vulnerabilities. The convenience of a one-click appliance update

[00:01:33] [SPEAKER_01]: and avoid the hassle of managing multiple unhardened servers.

[00:01:38] [SPEAKER_01]: Kiteworks is also fed ramp moderate authorised. So step into the future,

[00:01:42] [SPEAKER_01]: I'm secure managed file transfer with kiteworks by visiting kiteworks.com to get started.

[00:01:47] [SPEAKER_01]: That's kiteworks.com to get started today, but enough rambling for me.

[00:01:52] [SPEAKER_01]: Let's get today's guest on. Well, buckle up and hold on tight as I beam your ears all the way to

[00:01:59] [SPEAKER_01]: Singapore. Well, at Rafi is waiting a join us today. So a massive while welcome to the show,

[00:02:06] [SPEAKER_01]: Rafi. Can you tell everyone listening a little about who you are and what you do?

[00:02:11] [SPEAKER_00]: Yeah, sure. I think it needs for having me on the show. But there was Ravi and Global City of

[00:02:16] [SPEAKER_00]: I mean, based out of Singapore. Just about my profile. So I'm going to make some product

[00:02:23] [SPEAKER_00]: engineering up both in our products and my life. A little bit of innovation streaks,

[00:02:28] [SPEAKER_00]: I build innovation labs, intrapreneurships, startups. And then, you know, SI and the large

[00:02:34] [SPEAKER_00]: consulting forms, along with the digital flavors. I kind of dial it all in my career. Currently,

[00:02:41] [SPEAKER_00]: in my present role, I do solve the pressing business problems that are client-thersing,

[00:02:48] [SPEAKER_00]: especially with the changing landscape as you know, you and I haven't talked about today.

[00:02:53] [SPEAKER_00]: Work with our technology partners who are always innovating. So how do we create new strategic

[00:02:58] [SPEAKER_00]: capabilities together, offers together to solve those pressing, you know, business problems for our clients?

[00:03:05] [SPEAKER_00]: And I also oversee all the R&D for our company where we create new assets,

[00:03:10] [SPEAKER_00]: accelerator, doesn't sometimes create tools and products for our better productivity.

[00:03:16] [SPEAKER_00]: Especially in, again, in the areas of emerging tech so that we can advise out planes

[00:03:20] [SPEAKER_00]: where these solutions might fit for solving their problems. So that'll be the summary of what I do.

[00:03:28] [SPEAKER_01]: And one of the reasons I invite you on the podcast today is every day we talk about

[00:03:32] [SPEAKER_01]: how technology is transforming industry. In ways you don't imagine there's a lot of hype around

[00:03:37] [SPEAKER_01]: AI right now, everyone's talking about it. But also I want to try and get away from the AI itself

[00:03:43] [SPEAKER_01]: and understand what impact is it having here? What problems are we solving now? At Ogre V1,

[00:03:49] [SPEAKER_01]: you design the brand, you turn the brand into an experience and help your customers communicate

[00:03:56] [SPEAKER_01]: the brand's story and make brands matter. But can you elaborate on exactly how Gen AI is transforming

[00:04:02] [SPEAKER_01]: the landscape of personalized marketing and also what kind of benefits can brands expect from

[00:04:08] [SPEAKER_00]: effective implementation of this technology? I think that's very beautifully put,

[00:04:14] [SPEAKER_00]: very different from how usually I'm asking this question. So let me try and, you know,

[00:04:19] [SPEAKER_00]: I think for the listener since it's audio only, can I put it in some sort of a visual perspective?

[00:04:26] [SPEAKER_00]: Imagine on x-axis that as a market here, you are doing your day-to-day job on the left-hand side

[00:04:32] [SPEAKER_00]: of the x-axis and you're trying to see if you had a new campaign idea, how do you evaluate it?

[00:04:39] [SPEAKER_00]: You could go to a Gen AI engine today and ask for some ideas around what do you want to achieve?

[00:04:43] [SPEAKER_00]: I want to increase my customer acquisition. So one of the ideas I have, right?

[00:04:47] [SPEAKER_00]: You could do that. So I call that incremental day-to-day improvement in how you communicate with

[00:04:54] [SPEAKER_00]: your prospective customers. On the right-hand side of that x-axis, if you go to the extreme

[00:05:00] [SPEAKER_00]: like you're talking about customer facing deployment of this technology. So you would hear a lot

[00:05:06] [SPEAKER_00]: stuff around, you know, the service boards or service agents, we're now taking advantage

[00:05:12] [SPEAKER_00]: of natural language. You can have more natural conversations, more post-life conversations.

[00:05:18] [SPEAKER_00]: But I think the true beauty of scale is that you can actually personalize every aspect

[00:05:25] [SPEAKER_00]: property of that communication that you're sending out to your customer or your consumer depending

[00:05:32] [SPEAKER_00]: on your V2C or V2B client. Think about it, the voice of the communication, the tonality of

[00:05:40] [SPEAKER_00]: that text, the advertising banner, the creative on the page or the webpage, the product description

[00:05:49] [SPEAKER_00]: in the way NIGI would like to see it, including the experience, interaction,

[00:05:56] [SPEAKER_00]: itself. So we've reached a point where Genie I could actually also become Genie UI so you could

[00:06:01] [SPEAKER_00]: kind of compose your experience in a way that, you know, NIGI and Ravi can experience the same page

[00:06:08] [SPEAKER_00]: slightly differently in the way it would be presented to us. So that's that's on that axis of

[00:06:14] [SPEAKER_00]: personalization on both sides. Now if you take, let's say a vertical axis and you said, hey,

[00:06:21] [SPEAKER_00]: what about productivity use cases and transfer me to use cases? So in productivity think about,

[00:06:28] [SPEAKER_00]: you know, we all have business work flows, some market here, you and I, you know,

[00:06:32] [SPEAKER_00]: you're recording this point because you may go through steps multiple steps. What if you

[00:06:36] [SPEAKER_00]: deployed some agents and those agents take each of these tasks and do those tasks very effectively

[00:06:42] [SPEAKER_00]: for you? So we call that like a, you know, efficiency by automating certain things. You use these

[00:06:49] [SPEAKER_00]: agents with Genie I skills. But when you go to transform it where you're thinking about,

[00:06:55] [SPEAKER_00]: I will actually transform my business for future readiness using Genie I. In that construct,

[00:07:02] [SPEAKER_00]: you have to think about, you know, reengineering your business process, you have to think about reengineering

[00:07:08] [SPEAKER_00]: your organizational construct and for those things, you have to think about how the AI can be applied

[00:07:15] [SPEAKER_00]: at scale. There are very few who do that. Okay, we won as part of WPP, WPP has a platform

[00:07:24] [SPEAKER_00]: we call it WPP open. It's your listeners search for it then find some literature on it. What we've

[00:07:31] [SPEAKER_00]: been able to do is take, you know, the bits and pieces that I described in the other three pillars

[00:07:37] [SPEAKER_00]: and then turn it into this operating system of sorts where if as a market here, you had invested

[00:07:44] [SPEAKER_00]: in different technologies for experience management, for analytics management, for commerce management,

[00:07:51] [SPEAKER_00]: for communication management. But you could then surface all of them up tie them together with the

[00:07:57] [SPEAKER_00]: power of data in AI and then create this one uniform orchestration tool that just has the power

[00:08:04] [SPEAKER_00]: of transforming your business in a way or change the post-flication communication with customers

[00:08:10] [SPEAKER_00]: in a way that you can't imagine. So if you, if I'm able to probably visualize it for our listeners,

[00:08:18] [SPEAKER_00]: I think that's the true, you know, implement or true impact of the implementation of Genie I

[00:08:24] [SPEAKER_00]: then each one of us including Rock Eons can see. I love that example, especially how we could

[00:08:30] [SPEAKER_01]: be looking at something that's very similar, but we have a personalised look at that. And

[00:08:36] [SPEAKER_01]: I suppose these some ways Genie I is like a human, he does overuse certain words though and by that

[00:08:41] [SPEAKER_01]: I think delve tapestry, vibrant and landscape often gets thrown up. So with Genie I being widely

[00:08:47] [SPEAKER_01]: adopted, would you think of some of the big challenges that market has faced when it comes to

[00:08:53] [SPEAKER_01]: integrating this technology into their existing workflows? So I think the, we can look at

[00:09:01] [SPEAKER_00]: challenges again in various lenses. So I'll try and keep it simple but also something that you

[00:09:07] [SPEAKER_00]: know listeners can sort of visualize and keep it with them. So there, there are again let's

[00:09:12] [SPEAKER_00]: look at it in the people process data and engineering sort of lens. If you think about it,

[00:09:18] [SPEAKER_00]: data is the sole for any AI Genie I or classical AI conversations. So you need to have like

[00:09:31] [SPEAKER_00]: necessarily not reach any compliance privacy etc. But if you think about it, data conversation is

[00:09:38] [SPEAKER_00]: not new right? I mean in 2000, the first data mining company that I worked with and we

[00:09:47] [SPEAKER_00]: were sitting in 2024 and we are still talking about data challenges and there are various reasons,

[00:09:52] [SPEAKER_00]: some legit some not so legit right? So as an example a client will implement a certain technology

[00:09:59] [SPEAKER_00]: for a different reason with their social, political or organization structure. They might have

[00:10:05] [SPEAKER_00]: the ownership of that technology with someone else. So you not by default created in data island

[00:10:10] [SPEAKER_00]: then you have another technology then you might have centralized data analytics team. So you have

[00:10:15] [SPEAKER_00]: you know by nature of our complex organizations whether we are on the service side or on the

[00:10:28] [SPEAKER_00]: data available for you to create magic through emerging tech like AI. The second, second part of

[00:10:36] [SPEAKER_00]: this challenge is you know when we look at chatjp gpt and as a user when we interact with it

[00:10:44] [SPEAKER_00]: we find it really simple. So sometimes we take that to be that that's that's exactly how

[00:10:49] [SPEAKER_00]: simple probably the implementation is going to be. But actually you need to think about lot more

[00:10:55] [SPEAKER_00]: things than that every model has a version, every model has a certain performance associated with it.

[00:11:02] [SPEAKER_00]: KPI is associated with it. You need to make sure that you have an engineering team that kind of

[00:11:09] [SPEAKER_00]: understands the pipeline that brings data into the model, the pipeline that tunes that model,

[00:11:15] [SPEAKER_00]: the pipeline that evolves that model can be implemented in your technology investment because you're

[00:11:21] [SPEAKER_00]: going to suddenly say I'm going to approve all my existing investment and you know bring a new

[00:11:26] [SPEAKER_00]: technology. So you need to actually evaluate all of that and then you start to get into the

[00:11:32] [SPEAKER_00]: the people process lens. The process would be it can can the existing approaches or delivery

[00:11:40] [SPEAKER_00]: methodologies sufficient for AI because we don't know through ROI for most of our business processes

[00:11:46] [SPEAKER_00]: if we were to you know let's say augment them with Gen AI right the cost to better filtration. So

[00:11:51] [SPEAKER_00]: go to experiments. So how do you create the value stream based you know the delivery models?

[00:11:58] [SPEAKER_00]: But if you think from an IT standpoint it can be very scary right I mean you're you're

[00:12:04] [SPEAKER_00]: bringing in a completely new technology and you need you still want some sort of control like

[00:12:09] [SPEAKER_00]: like for anything if you have a cloud platform today I won't control over what's happening, who's

[00:12:14] [SPEAKER_00]: logging in what sort of data has been collected what's there you know transparency that's available

[00:12:18] [SPEAKER_00]: to me to make sure that I have a assurance that you know nothing is being breached. Now we have

[00:12:24] [SPEAKER_00]: a tool or a technology that can actually be a decision maker so get about you know being just one

[00:12:31] [SPEAKER_00]: of the tools that a user uses right. So having those controls the assurance the telemetry

[00:12:38] [SPEAKER_00]: that allows you to peek into what's happening at any given point in time is remains a huge challenge

[00:12:44] [SPEAKER_00]: so we need to work with you know different clients or market years have to work with their IT teams

[00:12:50] [SPEAKER_00]: to make sure that the two teams can figure the answer out and then finally I think the governance

[00:12:56] [SPEAKER_00]: challenge is well understood right the ethics the privacy the responsible use of AI the bias

[00:13:03] [SPEAKER_00]: the grounding that is required to make sure anything that AI does both sides the customer as far

[00:13:10] [SPEAKER_00]: as the market you can trust it because it's it's not you know I am using it but when I explain it

[00:13:17] [SPEAKER_00]: to my teams I always explain it like a persona because basically this technology is more than just

[00:13:23] [SPEAKER_00]: being a slave right it has its own personality it can sort of dream that dream and give you the

[00:13:29] [SPEAKER_00]: answer right so it's a little more than the it so therefore I think the trust is very important

[00:13:35] [SPEAKER_00]: because it's going to do things that the previous technologies haven't done before.

[00:13:39] [SPEAKER_00]: Value longer chance but you know I've tried to marry the challenges with you know so the

[00:13:45] [SPEAKER_00]: practices or the solutions we need to look for between business and IT.

[00:13:51] [SPEAKER_01]: I'm glad you did it really helps bring everything to life and there's also an argument we're all

[00:13:56] [SPEAKER_01]: talking about AI on a daily basis but not necessarily enough about data because AI is useless

[00:14:03] [SPEAKER_01]: without it so how important is the quality the availability of the accuracy of data for the

[00:14:09] [SPEAKER_01]: general I driven marketing and what steps can any market still listen take to ensure that

[00:14:15] [SPEAKER_00]: that data is up to standard. Yeah again brilliant question right so like you said availability

[00:14:22] [SPEAKER_00]: and accuracy are important so so Jenny as we understand so I'll try and simplify for so that I can respond

[00:14:30] [SPEAKER_00]: to that yes you right so it's a combination of whatever data has been said and it's been encoded

[00:14:35] [SPEAKER_00]: so think of it as a world view. Jenny I model has got a world view based on all the data that

[00:14:41] [SPEAKER_00]: has been fed into it and now whatever you want it to do you want it to answer your question

[00:14:48] [SPEAKER_00]: or you know create some sort of classification or segmentation of what have you it is going to

[00:14:54] [SPEAKER_00]: use that data set to respond to your questions right so now if that's the basis of what Jenny

[00:15:01] [SPEAKER_00]: is then of course like you said the data becomes the solved for this yeah this engine AI engine now

[00:15:09] [SPEAKER_00]: how and why this is important so have the other thing that let's say I'm going to use Jenny

[00:15:15] [SPEAKER_00]: and it's an expensive technology right so let's not let's be clear about it. If I'm going to

[00:15:21] [SPEAKER_00]: use it I'm going to use it for things that are going to give me the right ROI so as an as an

[00:15:26] [SPEAKER_00]: example if I want to use it for making some business decisions for me or augment making business

[00:15:32] [SPEAKER_00]: decisions for me. If I need to do that then I need to make sure that the data I provided to train

[00:15:39] [SPEAKER_00]: the context, the parameters, the weights, the biases all the tuning and the attributes the emphasis

[00:15:47] [SPEAKER_00]: we call them features in technology term but the features that I emphasize that how do you make

[00:15:53] [SPEAKER_00]: a decision when I want you to you know helping make the decision based on the situation I'll give

[00:15:58] [SPEAKER_00]: you or the data set that I'll give you. You can't do that if you hadn't provided it the right

[00:16:03] [SPEAKER_00]: data set from your company, your organization, your department lens right similarly if you were

[00:16:10] [SPEAKER_00]: looking for better performance and the performance in this case is not the page performance that

[00:16:16] [SPEAKER_00]: did happen in one second or two second or three second the performance in this case is more

[00:16:22] [SPEAKER_00]: about accuracy and you know over a period of time whether that accuracy degrades or improves

[00:16:28] [SPEAKER_00]: right so that's performance in a very layman term in this case so all the studies if you look

[00:16:35] [SPEAKER_00]: at the benchmarks have proven that quality of data gives you better performance on any I model

[00:16:41] [SPEAKER_00]: than the quantity of data so you don't want to just give it all sort of data you want to make sure

[00:16:46] [SPEAKER_00]: eat it with the quality data so that you can get the performance output and then finally I think

[00:16:52] [SPEAKER_00]: the competitive advantage right so I'm kind of marrying your your the other part of the question

[00:16:59] [SPEAKER_00]: as part of the response as well. When I when I'm you know going to be doing the

[00:17:05] [SPEAKER_00]: desired data skeet toward whether it is my that's a campaign performance data that I have as

[00:17:11] [SPEAKER_00]: an organization or my first party data on any other business with action data that I have then

[00:17:18] [SPEAKER_00]: I'm creating an AI model that is my own which understands my business it understands my you know

[00:17:25] [SPEAKER_00]: the past performance and therefore it can help me predict the future and it creates that you

[00:17:31] [SPEAKER_00]: need sort of an IP for you and therefore a competitive advantage instead of just using you know

[00:17:37] [SPEAKER_00]: off the share of the box something that you're everybody else is even saying so

[00:17:42] [SPEAKER_00]: data is really the key I think it's it's obvious but I will add so that it's not neglected

[00:17:52] [SPEAKER_00]: the compliance the security and and they you know the the responsible use of data is equally

[00:18:01] [SPEAKER_00]: important so which means that you need to make sure where data needs to be anonymized it is being

[00:18:06] [SPEAKER_00]: anonymized wherever data needs to be not processed or you know the obvious reasons like GDPR

[00:18:12] [SPEAKER_00]: etc that you're going to not process that data so that's equally important it falls

[00:18:18] [SPEAKER_00]: as part of this you know the data question itself. Another debate I'm hearing more and more about

[00:18:25] [SPEAKER_01]: is customer AI models versus off the shelf solutions to dig a little bit deeper on that what

[00:18:31] [SPEAKER_01]: you say are the advantages of leveraging customer AI models over those off the shelf solutions and

[00:18:37] [SPEAKER_00]: how can this provide maybe a competitive edge for brands it I always suggest in my conversations that

[00:18:46] [SPEAKER_00]: you look customizing the AI models you're largely so late as I'm certain at the brand

[00:18:51] [SPEAKER_00]: because of the cost associated with it it's it's a lot you know simple that you turn up and you

[00:18:56] [SPEAKER_00]: can you can you can create a customer and what do you get to that I think it's an important

[00:19:01] [SPEAKER_00]: there are two three ways you can actually do customer's issue without having to actually

[00:19:07] [SPEAKER_00]: customizing the model. One is a pattern called RAG pattern which is a retrieval augmentation

[00:19:14] [SPEAKER_00]: generation pattern so basically you're not by tuning the model itself you're taking advantage of

[00:19:21] [SPEAKER_00]: the let's say off the shelf or foundation model that you either with proprietary data or with

[00:19:27] [SPEAKER_00]: another cloud provider so what have you and then you're augmenting it with your company specific

[00:19:32] [SPEAKER_00]: data. So let's say search is a use case that you want to implement right or you know customer service

[00:19:38] [SPEAKER_00]: bought use case you want to implement the most important scenario there is that you won't the

[00:19:44] [SPEAKER_00]: answer to be grounded in your data and not some you know internet script data so that RAG pattern

[00:19:51] [SPEAKER_00]: very nicely without having for you to you know spend a lot of cost and fine tune in the

[00:19:57] [SPEAKER_00]: one you can solve that problem so that one way you can sort of use the the customer pattern in

[00:20:02] [SPEAKER_00]: AI scenario the other is the AI agents which are you know you can create an army of agents think about

[00:20:08] [SPEAKER_00]: you know a lot of your work currents you know sort of doing a specific task for you there there's

[00:20:14] [SPEAKER_00]: skilled with or you know they're given a specific tool to do summarization there's an example

[00:20:18] [SPEAKER_00]: or go and find a particular thing or go in the background and do the analysis for you and

[00:20:24] [SPEAKER_00]: then only present the report back to you right so that's another way you could you could do some tasks

[00:20:30] [SPEAKER_00]: without really doing the fine tune in but if we if we get to the specific question that you've asked

[00:20:36] [SPEAKER_00]: there are legit reasons when when you would actually fine tune or as your cognitive customer is

[00:20:42] [SPEAKER_00]: Jenny I'm all which is that like you said when you have a business specific use case like you've

[00:20:49] [SPEAKER_00]: got tons of your own campaign performance data from past from media campaigns or communication

[00:20:54] [SPEAKER_00]: campaigns you want to take advantage of that and then create your own brand voice let's say

[00:21:03] [SPEAKER_00]: Jenny I'm on so that you can generate your specific brand creative you want to be compliant to your

[00:21:10] [SPEAKER_00]: brand you know attributes or our purpose so to say right so you don't necessarily break the

[00:21:18] [SPEAKER_00]: the brain brand compliance or you have a domain specific understanding so your lexicon

[00:21:24] [SPEAKER_00]: has to be the same as you know you are your business users so there are general reasons when you

[00:21:30] [SPEAKER_00]: will want to be it the customization and then when you do do that and for the the reasons that I said

[00:21:37] [SPEAKER_00]: advantage is you will have it's a now you have something that you've created as your IP so now

[00:21:42] [SPEAKER_00]: you can call it your own technology tune be you're not producing something that everybody else is

[00:21:51] [SPEAKER_00]: producing so if you and I go on you know let's say the Gemini and start to create a new creative

[00:21:58] [SPEAKER_00]: image or create a new breed both of us are probably going to get similar answers slightly random but you

[00:22:04] [SPEAKER_00]: know similar answer but now that you tune your own model with your own creative from past or your

[00:22:12] [SPEAKER_00]: own campaign performance like you're going to get your specific output so you're not going to have

[00:22:17] [SPEAKER_00]: the sameness as you know any other brand so you're going to look unique you will be differentiated

[00:22:23] [SPEAKER_00]: you will be talking to your customer with your specific brand purpose voice and that and that

[00:22:33] [SPEAKER_00]: the most important thing like I kind of called out in the previous questions and response is the

[00:22:40] [SPEAKER_00]: as the performance of your model so if you're using something generic it is going to give you

[00:22:45] [SPEAKER_00]: vague answers because there is no grounding you do not know what is the basis of its world view that

[00:22:51] [SPEAKER_00]: it is going to be responding with but with the fine tune where in you have the ability to control

[00:22:58] [SPEAKER_00]: the temperature, the waves, the biases, the tuning etc you can actually create something which is

[00:23:04] [SPEAKER_00]: very specific to your department your organization so you don't have to worry about whether it is

[00:23:10] [SPEAKER_00]: going to give me you know made up fact or it is going to give me some competitors data because

[00:23:16] [SPEAKER_00]: it was trained on something so you can avoid all of those pitfalls and have a very performant model

[00:23:22] [SPEAKER_00]: which then allows you to do the personalization for your customers or do productivity efficiency

[00:23:27] [SPEAKER_00]: for your market yours in your own specific way so that's the advantage of the customer

[00:23:34] [SPEAKER_01]: I'm on. A few years ago did you tell a system or the rise of them and everybody using those

[00:23:42] [SPEAKER_01]: voice assistants many wanted the dangers of actually handling over the voice of your brand to

[00:23:47] [SPEAKER_01]: Alexa, Siri or Google etc but now I'm seeing no same concerns being voiced with the use of

[00:23:53] [SPEAKER_01]: AI especially if everyone is using that same generic language that you can vary often spot

[00:23:58] [SPEAKER_01]: so how can brands balance the efficiency of AI but also maintaining that unique voice and

[00:24:05] [SPEAKER_01]: customer connection to avoid that losing that voice? Again fantastic question so I want to answer

[00:24:12] [SPEAKER_00]: that little bit with with say pressure in class so so this is the problem real so far as

[00:24:19] [SPEAKER_00]: as a services organization we are also going through the same disruption that everybody else is

[00:24:25] [SPEAKER_00]: going through that how does AI change our job and as we concerned as you talked about right we

[00:24:32] [SPEAKER_00]: we consult with brands we help them create a unique experiences create that bond design that

[00:24:38] [SPEAKER_00]: relationship with that customers and consumers with the brand which is unique right?

[00:24:42] [SPEAKER_00]: Now when we were thinking about how we will use AI and how our method or our approval

[00:24:49] [SPEAKER_00]: changes that allows us and so successfully in past years to create these unique relationship

[00:24:58] [SPEAKER_00]: bonds between the art points and their customers we we sort of you know looked back at our

[00:25:05] [SPEAKER_00]: own foundational methodology we call it relationship design and we said actually all of these answers

[00:25:11] [SPEAKER_00]: exist in that relationship design methodology what it is is actually forward different lenses with

[00:25:18] [SPEAKER_00]: which we examine a problem or a business imperative those are brand business customer consumer

[00:25:26] [SPEAKER_00]: and the technology in data so we always look at all the problems that we're trying to solve with

[00:25:32] [SPEAKER_00]: these four lenses this allows us even if you know Ravi and you know Nila looking at the same

[00:25:39] [SPEAKER_00]: problem this allows us to have diverse interviews because the way you may interpret data maybe

[00:25:44] [SPEAKER_00]: different as I would the way you will look at customer segments would be different but at the

[00:25:50] [SPEAKER_00]: same time because we are combining them with you know customer and the business ecosystem data

[00:25:58] [SPEAKER_00]: the brand's purpose and all the data that you have might have collected business or you know

[00:26:04] [SPEAKER_00]: the consumer data or transactional data or behavioural data you obviously are going to come back

[00:26:09] [SPEAKER_00]: with very different diverse and integrated views that would automatically create a unique perspective

[00:26:16] [SPEAKER_00]: so this methodology obviously there is a lot of detail behind this methodology but this allows

[00:26:21] [SPEAKER_00]: us to create that you know unique blend of outwards and that's that's something we are

[00:26:27] [SPEAKER_00]: codifying in our day-to-day working that's something that we're codifying into and actually

[00:26:32] [SPEAKER_00]: AI driven tool and we're seeing the outfoot that you know we're able to keep that human

[00:26:38] [SPEAKER_00]: creativity the potential that humans have that you know ingenuity that humans have allow with

[00:26:44] [SPEAKER_00]: the efficiency and the scale that the genuine I can give us that's that's how I will respond to it

[00:26:50] [SPEAKER_00]: I'm sure different different clients will find their own method or approach which might be

[00:26:55] [SPEAKER_00]: giving them a similar sort of an output but I find the relationship design very very

[00:27:02] [SPEAKER_00]: productive and useful to sort of you know solve this challenge or seamless or see

[00:27:08] [SPEAKER_01]: of seeing this and where I just feel that we're in somewhat of a learning phase at the moment

[00:27:13] [SPEAKER_01]: and trying to work out what will work well can given that so many workers are interested in learning

[00:27:18] [SPEAKER_01]: more about AI other any strategies that you'd recommend for organizations to invest in AI

[00:27:25] [SPEAKER_01]: education and train staff across all levels of the organization and ensure that everyone moves forward

[00:27:31] [SPEAKER_01]: together we don't really very well be heard any strategies around that and getting everybody on

[00:27:35] [SPEAKER_00]: boat I'm going to give you a summary of what I've been through in last year or what our

[00:27:40] [SPEAKER_00]: organization been through in last year and therefore you know people can learn from our you know

[00:27:46] [SPEAKER_00]: what we think actually works. A like any other learning you have to have a tier approach to learning

[00:27:53] [SPEAKER_00]: AI as well but that means is that not everybody would have the same news some will be advanced users

[00:28:00] [SPEAKER_00]: some would be you know some would remain let's say you know basic users of the AI technology and

[00:28:06] [SPEAKER_00]: that also depends on the function they are part of number one. Number two when you start you

[00:28:11] [SPEAKER_00]: won't start with foundational elements which are same for everybody and foundationals are like

[00:28:18] [SPEAKER_00]: you know understanding this technology without being you know geeky or nerdy but understanding

[00:28:22] [SPEAKER_00]: it enough to see what are the challenges it brings as well as what are the advantages it brings

[00:28:26] [SPEAKER_00]: so that you don't have the fear of unknown because most of the resistance comes from that fear

[00:28:32] [SPEAKER_00]: of unknown in the organization. The third aspect is that once everybody's gone that foundational

[00:28:39] [SPEAKER_00]: learning you want to start to do purpose driven training which is to say let's say in an

[00:28:45] [SPEAKER_00]: organization you have a be business analyst you you have someone who manages your data data

[00:28:52] [SPEAKER_00]: you have an engineer who does coding you have a manager who is looking at let's say the dashboards

[00:28:59] [SPEAKER_00]: and Beatrix etc and then you can you know keep going through the hierarchy. Each one of them

[00:29:05] [SPEAKER_00]: will have a very different problem they will solve in their day to day work.

[00:29:10] [SPEAKER_00]: Other used the AI tool for productivity solving something as simple as doing competitive research

[00:29:17] [SPEAKER_00]: doing summarization of hundreds of documents that they might have seen you writing some

[00:29:23] [SPEAKER_00]: user stories with you know steps that can be auto generated with the productivity and super

[00:29:28] [SPEAKER_00]: then so on right. So each one of them will look at a different year code and will look at it very

[00:29:32] [SPEAKER_00]: differently so we need to have that purpose driven as you go around in the journey of learning

[00:29:38] [SPEAKER_00]: and eventually each one of them will we'll get to a point where they will start to do cross learning

[00:29:45] [SPEAKER_00]: what that means is that a be able then understand the power that actually I can now be a

[00:29:52] [SPEAKER_00]: quota because now I can use this tool to say hey give write me a script a JavaScript or a

[00:29:57] [SPEAKER_00]: Python script for me to automate something and then they will be able to actually deploy that by

[00:30:03] [SPEAKER_00]: asking that how I deploy it in my environment and then that's when you will be able to unlock the

[00:30:10] [SPEAKER_00]: through productivity from from the learning but but more importantly you will start to develop

[00:30:16] [SPEAKER_00]: new skills and individuals that I will caveat with the expertise will not be replaced.

[00:30:23] [SPEAKER_00]: So a creative will know exactly what's the right creative if I go there and I generate an image

[00:30:29] [SPEAKER_00]: and I do product placement and uncertainty we're that's the need I actually understand what's

[00:30:34] [SPEAKER_00]: the right creative like so the expertise is never getting replaced but I can actually try and

[00:30:39] [SPEAKER_00]: you know I don't have to know wait for someone if I was doing a POC I can actually

[00:30:43] [SPEAKER_00]: go and generate something on my own right so that that becomes important but having said that so

[00:30:50] [SPEAKER_00]: this is what we've learned and you know it has worked for us and especially at WPP we've done

[00:30:56] [SPEAKER_00]: a massive you know educational program with the tools that we've made available but one thing that

[00:31:03] [SPEAKER_00]: worked best for us was that the tools that we bought the investment we made we took away all

[00:31:10] [SPEAKER_00]: the fundamental challenges so as an example if we made the creative studio that is one of the

[00:31:17] [SPEAKER_00]: tools available to everyone we took care of you know challenges like you know confidentiality bias

[00:31:27] [SPEAKER_00]: toxic content etc so that you don't have to be with it in day to day so

[00:31:32] [SPEAKER_00]: you know that is actually equally important because when you expose such a large if you got

[00:31:37] [SPEAKER_00]: thousands of your work force suddenly you know learning that you don't want to every day get into

[00:31:42] [SPEAKER_00]: a new challenge somebody gets exposed to toxic content or somebody's you know doing drift with

[00:31:49] [SPEAKER_00]: you know getting exposed to drift on the model and you know writing a song or what have you

[00:31:54] [SPEAKER_00]: right so you want to keep certain controls controls over what kind of content is being produced

[00:32:01] [SPEAKER_00]: I think that's that's something that that we learned that that was one of the one of the reasons why

[00:32:07] [SPEAKER_00]: we could freely distribute the tool at such a large scale for people to experiment

[00:32:12] [SPEAKER_00]: but that's I think the simplistic way but hopefully that gives people an idea that you know how

[00:32:19] [SPEAKER_00]: they can go about creating the training or learning programs and their organizations

[00:32:24] [SPEAKER_01]: and in the past we've seen when markets is maybe pick up the the wrong image of some

[00:32:29] [SPEAKER_01]: or a person or a celebrity or a business leader or a location on Google images and there's

[00:32:35] [SPEAKER_01]: a huge amount of embarrassment that happens from there but of course fast forward to present days

[00:32:39] [SPEAKER_01]: all about AI hallucination so how can marketers ensure transparency and explain ability in their

[00:32:47] [SPEAKER_01]: use of general particularly in terms of understanding and interesting somebody was out but

[00:32:52] [SPEAKER_00]: that provided by these systems I actually the answer is in the question that you're asking

[00:32:57] [SPEAKER_00]: well try and elaborate so I think we discussed trust sometime sometime earlier in this

[00:33:03] [SPEAKER_00]: conversation think about a situation where a market here use or you're a deploy AI and this AI is

[00:33:12] [SPEAKER_00]: sort of you know recommending vouchers or discounts to Neil and Robin and suddenly you learn that

[00:33:19] [SPEAKER_00]: it has started with distribute 60% discount everybody comes and interacts with it

[00:33:25] [SPEAKER_00]: that'll be that'll be a huge you know break up trust because that's not how it is supposed to

[00:33:30] [SPEAKER_00]: function so then what do you do right so that's that's a very legit question everybody's asking

[00:33:36] [SPEAKER_00]: right so one of one of the things like you said XER so there are techniques like like

[00:33:43] [SPEAKER_00]: explainable AI so when we're implementing AI we need to make sure that we implement it with these

[00:33:49] [SPEAKER_00]: techniques that are available some of them are lying an LIME and SHAP so those those techniques are

[00:33:56] [SPEAKER_00]: available but you need to you need to cap it with a couple of additional things one that you need to

[00:34:04] [SPEAKER_00]: make sure that in production there is enough telemetry and transparency that you are auditing on

[00:34:11] [SPEAKER_00]: on a regular basis that you know how is this AI making decision which part of it will come from

[00:34:17] [SPEAKER_00]: explainable AI but part of it has to come with you know audit that the AI tool has to generate

[00:34:23] [SPEAKER_00]: and then you have to review like any other you know data on it then you want to make sure that

[00:34:29] [SPEAKER_00]: you have documented your data sources you've documented the architecture that you used to train

[00:34:36] [SPEAKER_00]: this AI and then make sure that is reviewed through a governance process and that governance process

[00:34:43] [SPEAKER_00]: can actually call out early enough if there are any gaps whether in the way you are implementing

[00:34:49] [SPEAKER_00]: the architecture which could lead to an issue like this wherein you know AI just starts to make

[00:34:55] [SPEAKER_00]: some random decisions or it could avoid you know if it is going to create some biased output

[00:35:02] [SPEAKER_00]: if let's say you're generating some creative or it avoids you know some sort of data breach

[00:35:08] [SPEAKER_00]: because you know in past we've seen somebody uploaded a document and then you know one of the

[00:35:14] [SPEAKER_00]: AI engine actually verbate and cite it the same documents you want to avoid all of those right so

[00:35:21] [SPEAKER_00]: between these three things if we do them right I think we can we can avoid creating this trust

[00:35:28] [SPEAKER_00]: to deficit that is one other thing that that is emerging now it's not very prevalent but I

[00:35:38] [SPEAKER_00]: think AI as we were building software in the past right we created some best practices we call them

[00:35:45] [SPEAKER_00]: extreme programming practices or different termologies and things like unit testing run whatever we

[00:35:52] [SPEAKER_00]: wrote we tested before you know it was even sent to holiday assurance teams etc for AI as well

[00:35:59] [SPEAKER_00]: though techniques are available they're not well known so there's a technique on evil

[00:36:04] [SPEAKER_00]: so as you are implementing AI step by step through your you know let's call it AI as DRC

[00:36:10] [SPEAKER_00]: then you should be implementing a well as part of your your methodology of development it should not

[00:36:17] [SPEAKER_00]: be enough up the thought for for an AI engineer or an AI's find test if we do that right I think

[00:36:24] [SPEAKER_00]: you can avoid lot of these challenges at the later stage so I think that's one one tip I'll

[00:36:31] [SPEAKER_00]: anybody we've been interested in exploring the new dark techniques especially in there

[00:36:36] [SPEAKER_01]: in their development methodology and despite the constant delays we'll must talk of course about

[00:36:42] [SPEAKER_01]: the death of third party cookies and unlike third party day other shared among companies

[00:36:47] [SPEAKER_01]: first party day is unique to your business and something that's rising it importance and

[00:36:53] [SPEAKER_01]: ultimately of course for anyone outside of the marketing space is the information you are own and

[00:36:58] [SPEAKER_01]: directly from consumers maybe through your Apple website interactions as well as emails and

[00:37:03] [SPEAKER_01]: loyalty programs etc but with all that in mind what role does first party day apply do you think

[00:37:09] [SPEAKER_01]: in creating effective personalised marketing campaigns when using Gen AI and how can brands

[00:37:16] [SPEAKER_01]: maybe better leverage this day or after spending too long addicted to the third party day yeah

[00:37:23] [SPEAKER_00]: I isn't it interesting right so a couple of years ago we were all worried about hey third

[00:37:28] [SPEAKER_00]: party cookies are going how are we going to do certain certain things that we are so used to doing

[00:37:33] [SPEAKER_00]: as well to yourself and here comes AI and then you know the the question changes from saying

[00:37:38] [SPEAKER_00]: so what can my data do now right because I'm sitting on and and all the brands are sitting on

[00:37:44] [SPEAKER_00]: on treasure trove of data there is a lot of data and not that the time clients tell me

[00:37:48] [SPEAKER_00]: we got to was data but we don't know what to do with it right and I think I think the Gen AI and

[00:37:55] [SPEAKER_00]: I'm going to actually add a caveat to it and AI together solves that problem I don't think

[00:38:03] [SPEAKER_00]: Gen AI alone can solve all the problems but we need to look at the predictive AI we need to look at

[00:38:10] [SPEAKER_00]: the classic AI methods as well the broader data science as well a combination of that becomes

[00:38:17] [SPEAKER_00]: very powerful so so as you know in the previous section we were discussing right if I'm

[00:38:24] [SPEAKER_00]: I'm trying to tune in the model I need to use my own data which is which is my first party

[00:38:28] [SPEAKER_00]: data on my business data on what hand view and that allows me to create that unique competitive

[00:38:32] [SPEAKER_00]: advantage that also allows me to answer you know create generate content if I'm using Gen AI

[00:38:39] [SPEAKER_00]: or personalize something at a very different scale and an AI can operate at a scale that

[00:38:46] [SPEAKER_00]: we couldn't earlier from a personalization standpoint so that becomes usually valuable right

[00:38:53] [SPEAKER_00]: but what I would add to that is that while you know we have all the marketers and most of

[00:38:59] [SPEAKER_00]: our clients had some system where all of this data is sitting so marketing organization today has

[00:39:04] [SPEAKER_00]: customer data platforms where they've brought in not a marketing related data put it in one

[00:39:09] [SPEAKER_00]: single system some clients have what we call composable data platforms where you know all the business

[00:39:15] [SPEAKER_00]: data and other you know cross discipline data are also stored so you can kind of do better data

[00:39:21] [SPEAKER_00]: science models but either either model that you're sitting on or either data back from that you're

[00:39:26] [SPEAKER_00]: sitting on I think the important thing would be if you look at the left-hand side of this data

[00:39:33] [SPEAKER_00]: that you're sitting on which is let's say the media which is the customer acquisition side

[00:39:38] [SPEAKER_00]: and you were doing your spending a lot of effort and time in doing media campaigns it is imperative

[00:39:44] [SPEAKER_00]: that you bring whatever data you can bring from your airport the third party over back into

[00:39:51] [SPEAKER_00]: your CDPs composable or otherwise and therefore that allows you to then look at the additional

[00:39:58] [SPEAKER_00]: dimension laws whether you're bringing in linked in data, fees put data allows you to see

[00:40:04] [SPEAKER_00]: your data with different feature attributes and start starts you to think about can I create a new

[00:40:14] [SPEAKER_00]: different insight a unique insight so that's that's one aspect of it on top of your

[00:40:21] [SPEAKER_00]: you know first party data the doubt first party data none of this is useful because first

[00:40:25] [SPEAKER_00]: party data is actually going to give you that unique identifier in in in in terms of you know

[00:40:30] [SPEAKER_00]: having that postlines conversation with your customer on the right hand side using personalized

[00:40:36] [SPEAKER_00]: or or first party data you are doing at scale personalization so you're sending out communications

[00:40:43] [SPEAKER_00]: you're serving a personalized product description page or information on your website

[00:40:48] [SPEAKER_00]: now feed back from all of those analytics or streams of analytics coming back into your CDP

[00:40:57] [SPEAKER_00]: and then allowing you to do this three way enrichment of data to then think about new ML ones that's

[00:41:03] [SPEAKER_00]: actually the where you start to unlock the true power of data so so in short the first

[00:41:10] [SPEAKER_00]: party data is very important the data that you're sitting on is really the core engine with which

[00:41:15] [SPEAKER_00]: you can drive your own initiatives but don't forget to sort of connect the dots between feedback

[00:41:21] [SPEAKER_00]: loop as well as you know the performance data that you can get from other sources for you to

[00:41:26] [SPEAKER_01]: create new segments and new insights and test it advice and you've shared so many golden insights

[00:41:33] [SPEAKER_01]: today I'm going to be great here and say before I like you guys I'm going to ask you to leave one

[00:41:37] [SPEAKER_01]: final gift for everyone listening we have an Amazon wishlist where I ask my guest to share a book

[00:41:42] [SPEAKER_01]: that might recommend or mean something to them so what book would you like to add to our wishlist

[00:41:49] [SPEAKER_00]: I'm going to actually go slightly I mean it's it's actually relevant but it's not it's not directly

[00:41:57] [SPEAKER_00]: linked to AI topic that we're discussing in a way engineers and I consider myself a technology product engineer at

[00:42:05] [SPEAKER_00]: what's so used to looking at the world through that you know the the matrix I can imagine

[00:42:10] [SPEAKER_00]: the movie rights of the bits the bytes to have a different perspective or explore a different

[00:42:17] [SPEAKER_00]: dimension you need to sometimes read stuff that gives you new dimensions or parallel universe so to

[00:42:24] [SPEAKER_00]: say for for you to let's say imagine possibilities with technology so there are two books actually

[00:42:31] [SPEAKER_00]: that that opened that perspective for me the first one was the the source source and their

[00:42:39] [SPEAKER_00]: apprentice I think it is by Frank knows he was the MIT media director and what he did was he sort

[00:42:47] [SPEAKER_00]: codified all his experiences and experiments you know in the digital AI MIT media lab was

[00:42:53] [SPEAKER_00]: this magical thing they were doing things that nobody else was able to do and what was fascinating

[00:42:59] [SPEAKER_00]: reading through that book was that you know honey innovate because all of us and you know in

[00:43:04] [SPEAKER_00]: day to day so stuck with the same processes then sort of you know technology and tools then you

[00:43:10] [SPEAKER_00]: know where is that one thing that you should change that allow you to innovate so that that has

[00:43:15] [SPEAKER_00]: got some real knowledge nuggets that that can open up that perspective for you without having to change

[00:43:22] [SPEAKER_00]: any you know substantive or fundamental things in the way you operate and that was very useful

[00:43:28] [SPEAKER_00]: and it's relevant because I was building an innovation lab that I found that book and you know

[00:43:33] [SPEAKER_00]: I was experimenting myself and trying to figure out his process there isn't for innovation are

[00:43:39] [SPEAKER_00]: people the catalyst to innovation or the tools and technology that pushes the envelope and you

[00:43:45] [SPEAKER_00]: find that actually either of them can change the output so what are my favorite examples from

[00:43:52] [SPEAKER_00]: that book me and probably giving you a longer answer but what are the favorite examples as when

[00:43:57] [SPEAKER_00]: designing a car for the city they actually got an environment engineer and quality to design

[00:44:03] [SPEAKER_00]: the car not an automobile engineer to so that it'll give a different perspective. The second book I'm

[00:44:09] [SPEAKER_00]: going to recommend is enchanted objects I forget who the author is now but enchanted object also

[00:44:17] [SPEAKER_00]: gives you a new portal of sorts right it allows you to imagine use of technology through the lens of magic

[00:44:25] [SPEAKER_00]: you know you're sure familiar with chronicles of now or any or Harry Potter wonder what

[00:44:30] [SPEAKER_00]: had you that book allows you to imagine how we are could be used to open new

[00:44:36] [SPEAKER_00]: new portals for you to experience you know new that's the information gateways and information

[00:44:44] [SPEAKER_00]: starting to raise of whatever you want to call it but they they kind of give you a new perspective

[00:44:49] [SPEAKER_00]: if you're a technological artist or a market year and then you're struggling with the how do I

[00:44:54] [SPEAKER_00]: use the technology or new technology in a magical way so those are the people books I I would

[00:45:00] [SPEAKER_01]: recommend I think they're fantastic books. I got both of those books added to the Amazon

[00:45:06] [SPEAKER_01]: wishlist and for yourself if I know wants to maybe connect with you on LinkedIn find out more

[00:45:11] [SPEAKER_01]: information about our gov or anything we talked about today where would you like to point everyone listening.

[00:45:19] [SPEAKER_00]: So as an individual I'm more active on X which is you know formerly known as Twitter so you can find

[00:45:27] [SPEAKER_00]: me as Ravipal at a VIP air 1 2 1 tour that's my handle but otherwise for all business formal

[00:45:35] [SPEAKER_00]: conversation or maybe one or otherwise just playing me on LinkedIn you can search with Ravipal or

[00:45:40] [SPEAKER_01]: maybe writing on Shangui. Next then we'll allow the links to everything there I love chatting with

[00:45:46] [SPEAKER_01]: you today a big message there and how Jenny I can produce those personalized marketing brands are

[00:45:52] [SPEAKER_01]: covering but also what customers expect and they suddenly big talking points around that and I

[00:45:57] [SPEAKER_01]: did buy everyone listening to let me know their thoughts but more than anything just thank you for

[00:46:00] [SPEAKER_01]: bringing this topic to life today Ravi thanks for joining me. I'll take a new thing for adding me.

[00:46:05] [SPEAKER_01]: I think talking about how strategic personalising presentations can significantly enhance things like

[00:46:12] [SPEAKER_01]: customer engagement and brand loyalty. It opens up so many different opportunities and as we wrap

[00:46:18] [SPEAKER_01]: up the show today remember that integrating the technologies that we've talked about today isn't

[00:46:23] [SPEAKER_01]: just about adopting new tools but it's also about evolving with them creating meaningful and impactful

[00:46:29] [SPEAKER_01]: customer interactions bringing your staff along for the ride as well ensuring everybody understands

[00:46:35] [SPEAKER_01]: technology but what changes do you see Jenny I bring into your marketing strategy please join the

[00:46:42] [SPEAKER_01]: conversation by emailing me techbrugwriteroutlook.com lovely your thoughts on this one and we'll be back

[00:46:48] [SPEAKER_01]: bright and early tomorrow so let me know your thoughts but that is it for today so thanks for listening

[00:46:54] [SPEAKER_01]: as always going and until next time don't be a stranger