We are joined today by Simon James and Terence Davenport. I will let them introduce themselves, and we will talk to them about what Publisys has to offer when it comes to their data and AI practice. We are going to specifically get into generative AI and their capabilities around Azure OpenAI, some of the patterns that are emerging, and how they can help our customers get to business value in an accelerated and agile way. So without further ado, I will hand it to our esteemed guests for a brief introduction. So Simon, we'll go with you first.
Thank you very much, Armin. Hello, I'm Simon James. I lead our data and AI capability in Amira and APAC at Sapien, and delighted to be here today.
And I'll go next. So I'm Terence Davenport, part of the Microsoft practice at Publisys Sapien, and I'm a technical director and tech SME in this space. Thanks again for making time today.
One of the first areas I think that makes sense to start with is really, Terence, if you could help us understand who Publisys is and what is the strength of your data and AI practice, give us a bit of an orientation about the company as well as your practice.
Yeah, sure. So let me first start giving a background view of Publisys Group and how Publisys Sapien fits within Publisys Group. So Publisys Group is based out of Paris, France, 100,000 plus employees. It's the second largest advertising agency group in the world. Publisys Group purchases over half of the media buys in the world and one out of every three media dollars in North America. And so how group is situated, we have four different solution hubs. There's Publisys Media, there's Publisys Communications, there's Publisys Health, and there's also Epsilon. And there's a number of different sister agencies under these different solution hubs like Leo Barnett and Razorfish and Saatchi & Saatchi, all these different sister agencies. And that just shows kind of the power of how we're connected with the CMO and media and advertising agencies. Publisys Sapien is the digital transformation arm of group. We drive with strategy and consultant technology and we have what we call speed capability. And as you see here, there's strategy, there's product experience, engineering, and data. And so we look at things a bit differently and how we deliver for our clients. So we bring strategy, we bring a vision, we make sure that we have value-based outcomes, we bring product where we think about how we scale and bring things to market with basically speed and quality and value. And then we look at experience, engineering, and data and AI as part of that. We're about 20,000 people across the globe, about 60,000 plus offices. We are basically consider ourselves leaders in data and AI. We have a joint venture called PSA Labs that we acquired early in this year. We have over 1,500 plus data scientists and strategy globally. And we also was one of the first to build a Genif AI playground on Microsoft Azure OpenAI called PS Chat. And we'll talk about that a little bit later through our webinar here. And we're recognized leaders. So we have met different benchmarks in the industry. We're also proud of our self-owned culture. So we have a number of different rankings that we achieve across various different agency comparisons. So we do have a lot of partners who provide AI offerings and solutions.
Please do share with us, right, what is your team's core offerings and how you go to market, how some of these offerings are structured. That'll help our teams understand what publicist has to offer when it comes to generative AI.
Yeah, thanks, Armand. So we would say that we're pretty versatile. We can tackle things bottom up or top down. And what I mean by that is some clients just want to get on and start experimenting and trying to see what generative AI can do for them. And obviously, that's great if you're trying to drive consumption out the gate, because in a couple of weeks, we can run workshops to prioritize use cases, get them organized into a backlog, and get working on them within a week. And some clients really like that kind of approach. In a two-week sprint, we can have a working prototype working live, production-ready code, enterprise scalable. Maybe it's not scaled yet because we don't have all the data. But it'll be live on Azure working for our clients so they can see. And I think that's one of the great things about generative AI is, like, historically, AI was, you know, it took us months to find the right data and months to train and optimize a model and months to deploy. But we've really sped that process up. So clients love that because they get the time to value is really decreased. So doing workshops, running hackathons, or getting straight to demos and POCs, like, we're very well kind of armed to work with clients to get going really quickly. And that's not right for everyone, because some clients want to take a step back first and say, well, what's my strategy? Where are my risks? You know, what are the other things? Who should own this in the company? Do I need an AI department or division now? Or should everyone do AI? And so stepping back and doing some of those strategy kind of questions, we also tackle that. And in specific, we've kind of really beefed up our approach to ethics, governance, and risk. Because none of our C-suite clients, whether CIO, COO, CDO, you know, they're not going to put anything live without making sure they understand the risks. So, you know, we find that's a big barrier unless you tackle it head on, that it's only going to come and bite you later on. So we do that. And also for clients who don't want to go through the learning and just want to kind of learn from other people, we've already thought of a lot of, like, industry-specific use cases. So in high-regulation kind of industries like financial services or healthcare, like, what are the use cases there? For retail and e-commerce, like, what are the e-commerce use cases? So we thought about that. So we thought about, like, the specific industry use cases and across industry kind of use cases. And then also for some clients, it's just about enablement. It's like if you're still a bit immature on your journey to cloud, you know, and moving all your data onto Azure, then how do I set my business up so I can, you know, start using generative AI on Azure? So we help clients with that kind of core mission as well, which is making sure all the enablement's in place, all the training's in place, and that they can feel comfortable their data's secure and safe on Azure, and then they can go and build some of these cool new kind of use cases for AI.
I love that. I love that. The capability of publicists to go in and really meet the customer where the customer is. And you're right, right? Sometimes it is a digital transformation conversation and there's lots of use cases, but your ability to go in and do a quick workshop, a quick start, build out a quick demo so that they can really get a taste of things, but then also stepping back and saying, okay, you know, do we have all the right elements in place so that, you know, this aligns with the overall customer strategy and where they are trying to go? Ethics is a huge piece for us too. We are seeing the same thing. Our teams are being challenged by our customers around a lot of questions around data privacy, responsible AI, how does our services from Microsoft, how do we exactly take into account some of these concerns that our customers have? So to have a partner like Publicis who can really partner with our customers and help them understand truly where the risks are, but then where there is just noise. Because we as partners, we will take care of some of the risks and will mitigate some of that from within the Azure platform, but then there are going to be other things that the customer is going to have to think about and they are going to need your support in figuring out what that ethical AI strategy needs to be. So I just love the fact that you are meeting your customer wherever they are, right? So whether it be at the point of strategy or whether it be at the point of enabling them to just see what the art of the possible is. Also, we get a lot of questions about generative AI and how that could drive industry specific use cases. So really would love to get your thoughts on from a use case perspective, what are you seeing? I think you've done some work around enabling generative AI just even within Publicis, but then making it available to customers as well. So maybe we could talk about that, Simon. What are some of the use cases, patterns, trends that you are beginning to see with customers?
Yeah, and if we, I think, go on to the next slide, Terence, please. So, you know, the way we frame that up is like everyone wants to run and do things really quickly, but we know there needs to be a balance in terms of those guardrails and those risks. And I think there's a big difference between risk, which is something you know and you can deal with, and uncertainty. And I think clients have risks. It's fine. We can deal with the risks. It's the uncertainty that no one likes. In terms of where the use case and the value is, I think there are three broad categories. First one is how can I do the same with less? And that's really, you know, an automation kind of play. How can I take some cost out of something and still deliver the same results? I think the second one is what we generally call augmentation, which isPoint about customers might want to, you know, take the first step with something like PS Chat because, you know, they are hesitant to go big with something external, but they can do a trial, expose it internally to a set of select group of people. And then from there, you know, once that is successful, they can expose it externally to their customers. And to Simon's point, from there, it's not just about the chat use case. This is just the beginning. Once they get comfortable with that, then it opens up the possibility of other web applications that are using Azure OpenAI in the back. So I love your thoughtful approach. And I also think the fact that you've done this for yourself is going to give our customers a lot of confidence. Why don't we talk about maybe some more customer success? I think, Terence, you have some examples to share. So yeah, I think this one with the UK RHS group, I think is an interesting one. And there are many more, but walk us through what your team did and the success that you've seen. Simon, you want to take this one?
Yeah, sure. So probably a very British example, the Royal Horticultural Society has been around for hundreds of years, as you'd expect. They look after the gardens and the heritage of plants in the UK. And one of the things they've built up over the last 100 years is a huge data set on answering members' questions about gardening. And so they have the perfect data set for LLMs. They have Q&A around any question you've ever seen around gardening and write the expert answer to this. So we've taken that data and built a bespoke LLM for them and put that into an app that allows anyone to ask a question of the bot. And it can answer any of your questions around gardening. I know many countries in the world don't have the fascination with gardening that the UK people do, but it's a very British kind of case study. But yeah, it's a tremendous kind of knowledge bank for anyone who's trying to grow anything in their garden.
Yes, Simon, and I think, yeah, even though this might be, you know, horticulture and more oriented towards plants and horticulture, etc., I do think it, you know, the type of interface that you are sharing here, you know, on the mobile app is very relatable, right? So these types of applications, mobile-based, very, very applicable to shopping experiences. So I do think this is a great example of a knowledge base that can have a user interface that is really a differentiated experience.
Yeah, and for any client that has their own, you know, their own data, really bespoke knowledge that maybe only they have that nobody else has, proprietary kind of data, and it's a field that doesn't generalize well for large language models. And so specific knowledge about, you know, what you do with specific plants, that's either fine-tuning or building your own LLM. Now, some clients will have the resources to that and some won't have the resources for that. So you can do it the heavyweight or the lightweight way, but either way, it's going to drive a lot of consumption, a lot of usage, because you get 100,000 users using that, you know, regularly, then that's going to generate a lot of kind of consumption. So really good, like technically proficient, using cutting-edge technology to give, you know, people in this field who actually, you know, love gardening, you know, something that they really want. It's really valuable. It's not a technology looking for an application. It's just a really practically useful tool for people who have, you know, gardening in their life.
Wonderful, wonderful. And I think if we look at beyond these use cases that are more horizontal use cases, I would say, right, like chat and these types of web applications, with your focus on marketing, I'm sure you're seeing a lot of applicability around content generation across the board with the marketing function. Is that a fair assumption?
Yeah, we're seeing a lot of those examples. And I think at the moment, like chat is kind of nine out of, so text is nine out of 10. It's pretty good. You can roll that out immediately with very little work. Imagery, it's probably six out of 10. You know, it's good, but it needs a lot of hand-holding and development. And so, yeah, we're on that journey with clients kind of helping them mature and maturate through the process of that. And starting, I think everyone starts with language and text and moves on to vision. It's probably the next barrier. And then obviously it gets more and more complicated after that. But yeah, we're definitely seeing a lot of that. That's the next horizon.
Yeah, and I know we're also seeing, and we have a number of different use cases built out for this, is around the personalization as it ties to AI, right? And so you think of areas of taking a mobile application, maybe say retail or travel hospitality to kind of have that conversational experience. So what you're seeking is personalized towards you and what the output may be. So for example, if you like a specific hobby or a specific area, then that response may be more personalized to you. And maybe also even with the imagery may be more personalized to who you are. So we have built different demos for that, for different retail clients, where we have taking something where we have looked through product reviews or product description, and then tying that to an actual customer and what their personalization is. So if they like a specific thing, then that imagery and what they see on that landing page may be related to them. So those are just some examples as well, kind of how you can kind of tie this into personalization and loyalty and how AI can be used in that space as well.
Those are excellent examples. And I do want to reiterate for our audience, I think there are some key industries from what I have seen with publicists where you all do extremely well. I think what you are able to offer in CPG and retail, very differentiated with your focus on data, the focus on being able to do these very applications that require personalization, I think you all do a fantastic job at that. So retail, CPG, definitely. I think the other two, and tell me if you agree, the other two where your work really stands out, I would say is energy and commodities. And so Terence, is that a fair statement?
It is. And so we've done a ton of work in data. I mean, that's our bread and butter. I mean, we do a lot of the data and app modernization. So this is an example of a case study where we've done for a large oil and gas company where we went in and implemented basically a CDP, a customer data platform. So the customer was seeking to get a bit more understanding about, capture more information about their consumer, build loyalty, increase sales, right? And so as part of that, so we implement a CDP on Azure and be able to enable, activate that data to be able to understand and get a 360 view of the consumer. And so for them to also establish profiles and establish marketing campaigns and to be able to increase sales. And so in the result of that, sales increase, they was able to create more brand awareness and also have more conversions.
Thank you for the partnership that we have had for all these years. And I think this is a brilliant opportunity for us to provide some differentiated value to our joint clients. Appreciate both of you, Simon and Terrence for taking the time and sharing your thoughts with us. I wish you all the best and we will definitely be in touch as client opportunities come about and we'll be working very closely together. So thank you.
Thank you.
Thank you, Han.