From Fresh Produce to Fresh Ideas: AI & Digital Transformation in retail and private equity

Clément Marty (00:17)

Thank you very much for being here today with us at the AI x AI conference. That's the second edition of this conference organized by Ardian. And the objective of the conference is to basically help Ardian’s portfolio companies, but also our partners, the whole ecosystem, understand how AI can create value for the business. And one of the ways to do it is to showcase what we do, and more importantly, what our portfolio companies do. So you are a leading technology at Prosol. And today, we're going to share your journey, a long and very successful data journey that you've been leading at Prosol.

Fanjuan Shi (01:01)

First of all, thank you very much for inviting me to this second Ai x Ai conference, as this is the second consecutive year I have the privilege of joining, sharing and learning. So I really look forward to discussing with you and learning from our peers, their experience to benchmark so we can progress better and better use AI and data to create value for the company.

Clément Marty (01:27)

And how about Prosol?

Fanjuan Shi (01:28)

Well, Prosol is a very amazing company. It has a long history and is passionate about the quality and taste of fresh food. So Prosol is a specialist in fresh food, starting with fruits and vegetables, then moving to cheese, then to fish, and finally to meat and other categories. So today the group is generating more than four billion euros per year, representing more than 400 stores. And there are different kinds of brands in different kinds of countries. But not only that, we have strong teams to master the supply chain, and we have a very strong procurement team, a transformation team, and a logistics team. All this, for the quality of the products.

Clément Marty (02:16)

It doesn't sound like a tech company. But maybe you can tell us the type of data you have, in which areas, and on which topics your data covers. 

Fanjuan Shi (02:31)

First of all, for quite a long time, the company was obsessed with the quality of the products we produce and we procure. But recently, I would say 2 or 3 years ago, the company decided to accelerate the development by putting technology as an enabler of growth. So, a lot of investments have been made, new profiles have been recruited so that we try to put more and more, core KPIs, AI, and data, into our business because the top leaders of our company fully understand the importance of data in business decisions. So, this is a huge transformation that they sponsor, support, and trust.

Clément Marty (03:19)

And maybe you can share with us some of your product catalog. The data product catalog is very rich. Maybe you can share with us some data products or data use cases that you have worked on at Prosol.

Fanjuan Shi (03:36)

First of all, I think you are totally right. If we don't have high-quality data, we can’t work on use cases. So basically, we have created a very good infrastructure. Covering, recovering, transforming, transporting all this data and putting it into our data lake. And we have a system of three levels: bronze, silver, and golden data. Bronze status, our raw data. Silver data are transformed data that have already been cleaned. But they need to be enriched. And then the golden data is enriched data, ready for use cases. And on top of that, we also have data governance and a data catalog to measure the quality and ensure definitions and roles are correctly documented and then transferred to the use cases. So when we have high-quality have a couple of So the first family of use cases is the final products, which are the dashboards, the reports helping our stores, our logistics platforms, our factories to steer the performance of their day-to-day activity.

Clément Marty (04:55)

Every function has its own dashboard.

Fanjuan Shi (04:59)

We have created workspaces dedicated to each function.

We have role-based security so that each person has dedicated access to the report that they are entitled to. So, this is the first kind of product that we create: final reports. For departments that are more autonomous and want to go further in their data exploration, we help them create their own semantic models and datasets. So this happens in a couple of departments where they have data profiles. So, instead of creating dashboards for them, we create semantic models and datasets for them. This is the second level.

Clément Marty (05:40)

00:05:30,266 --> 00:05:40,833

And then, just to make sure that I understand well, they are autonomous in basically dealing with this data and extracting the information and the insights they want from it.

Fanjuan Shi (05:51)

Indeed, they are autonomous, and it's quite an interesting journey. And this is one of our strategic objectives. We would like our business units to be autonomous in the usage of data. So that's quite a fair comment.

Clément Marty (06:08)

How did you manage to train these business teams on data? But sorry, I interrupted you, on these types of data products.

Fanjuan Shi (06:19)

It's quite an interesting comment because we organized different kinds of training. Like Power BI training, like different kinds of training. Helping the business team take the tools on by themselves. You can see how impressed they are when they see that they can do all these kinds of things by themselves and not only rely on some experts. One of the philosophies that we want to put forward is that we would like the data to be accessible and also usable for teams, regardless of their business and their educational background. This is one of our philosophies. And if we go a little bit further for the teams who want to go even further, not only have descriptive and explanatory data, but also to have prescriptive and predictive things. So we work hand in hand with them to build models: predictive models, optimization models, assortment optimization, and price elasticity. We have different kinds of use cases co-developed with them. And to finish, there are some use cases which might require a very consolidated and high-level coordination at the group level, for these  use cases, like the GenAI use case, like a chatbot, and all these kinds of things. So, we produce in collaboration with different business units, ensuring we have a range of products in place.

Clément Marty (07:42)

Impressive. And I like very much the idea that one of your goals

is to put the data in the hands of the operational teams, and I think that's a great vision. And I believe that should be the goal for any company and, for sure, for Ardian’s portfolio companies. So, I love that. Maybe if we get a little bit into the technology, it looks like you have a lot of use cases. How does the person in charge, the leader of the team in charge of tech at Prosol, you, ensure that your infrastructure is at the right level of performance, but also of security, to handle all, this quantity of data, of use cases and of users? So how do you scale your infrastructure? Do you build a huge infrastructure in the beginning, or are you scaling it? I mean, how do you approach it?

Fanjuan Shi (08:48)

Thank you for your question. I think, first of all, when we started our journey, we had to review our legacy. We do have quite old-fashioned systems, which are performing very well, but they cannot be scaled very easily, and they cannot fit into the new use cases. So the first question is how to make a good inventory of these technologies and develop a strategy. Should we keep it, should we improve it, or should we get rid of it? So that's very important because we have a very pragmatic philosophy, which means we cannot break something that works. So if we want to break it, we have to make sure that there’s something better to ensure the continuity of the service. The second thing that we try to do is to choose the right talent, because ,as you said, in all the journeys, the most important thing is the people. So, in this journey, we identified the infrastructure manager. We invite the data platform manager, and also a couple of business users who are working together, so that we will be able to design, from the beginning, what future IT infrastructure is needed in terms of data ingestion, data transformation, storage, computing, governance, And, in the end, visualization and modernization. So all these kinds of things are worked on with anticipation because it's our philosophy to prepare the infrastructure a little bit in advance for the business. Otherwise, we are going to be a little bit in a rush, and the choice might not be optimized.

Clément Marty (10:34)

This can be a challenge. The infrastructure follows up on the usage. It's hard.

Fanjuan Shi (10:40)

In some cases, that can be. Maybe I would also like to add two little elements on this infrastructure topic. We try to put in place a concept: cybersecurity by design, governance and cybersecurity by design. Because in a lot of cases, even in our organization, before we develop a product, we submit it to the cyber security team, and they tell me, “Oh, sorry Fanjuan, this one is not compliant and this one, you have to rework it. ”So I tell the team working on the product and the team managing the cybersecurity: “Let's work together from the beginning.” So cybersecurity becomes a concept and also implementable actions in the product design an example can be BI tools. We, in the beginning, clearly specify the governance of the workspace, the governance of roles, the governance of the capability to share and integrate once the tool is in place.

Clément Marty (11:44)

Very interesting. Let’s discuss people again and starting with the top of the leadership. So, as a tech leader, what type of support do you have from business leaders? From the CEO, but also from the heads of BUs? How is your collaboration with this group of people?

Fanjuan Shi (12:10)

So I think it's a combination of governance and personal relationships. Both should work. Let's start with the governance. In our company, it's a joint effort of everyone. We try to govern our tech topics in four levels. But I will go very quickly.

Clément Marty (12:28)

But it's interesting because I   have an extremely structured approach. I think it's very useful for the people who listen to you to basically understand the frameworks and the structure to use them. So feel free to get into it.

Fanjuan Shi (12:42)

00:12:32,333 --> 00:14:38,633

And feel free to challenge if you find something that we can improve. Well, the four levels. First of all, portfolio. Portfolio means that all the tech topics are related to a business unit. So we regularly have a governance review with the business unit to review priorities track progress and make decisions collectively. So the second level is program. A program is a collection of tech topics that might be transversal and impactful for a couple of business units. It's not as strategic as a portfolio, but it's quite important. For example, if we are developing a chatbot, it might be of interest to several services teams. So, if we neutralize some requirements, we will be faster in deployment, and it will be more relevant.  these are the programs, and we have program leaders talking to different BUs, presenting prototypes, inspiring people, and collecting feedback so that when we develop products, they’re more relevant. Third very classic: project. So I'm not going to go into the details. And last but not least, we build, we transfer, we operate. So we also have tickets and tasks. So if an IT solution doesn’t work, they come to us, and we handle the tickets. So these are the four levels. So thanks to these four levels of governance with our business leaders, we have transparency, and we have efficiency. At the same time, my teams and I try to create personal relationships with the business leaders, try to understand what their priorities are, and try to understand how they operate their business. We put a very important focus on mobilizing ourselves to their sites, working with them, on their sites, and talk to them on a day-to-day basis so that we understand better. So this kind of relationship helps when we sometimes need to better understand the priorities. So I think it's a combination of these.

Clément Marty (14:48)

And I think this mix of the very structuredapproach and the common vocabularyand people know what is what, as well as this personal relationship that really makes problem-solving easier. It’s the golden combination. Question on the teams. We talked about the leadership. How about the daily users? Have you organized training? How do you ensure that they understand the data products you guys are building, and how do you make sure that they have the right skills to use them?

Fanjuan Shi (15:32)

Well, I think on this point we are still in the middle of a journey, so we can still improve a lot.

Clément Marty (15:40)

Prosol is a big company. You have many employees. It’s people heavy.

Fanjuan Shi (15:45)

Yes, exactly. We have more than 9,000 employees today, and they are from different professions. They have different backgrounds. So, it's very hard for everyone to have the same level of understanding of technology. This is not what we try to do. So, I think the first philosophy that I tell my people, my colleagues, working on the IT side ,is: imagine that if finance people are talking about their jargon to you on a daily basis, can you really adapt to their proposition of a product? It will be hard. So let's do the same. Let's try to avoid too much jargon and the very specific IT stuff for our people. Let's simplify. Let's use a metaphor. Let's use images. Let's use easy ways. So this is the first thing. The second thing is that, when we design our tools, let's try to be more user-centric. So, try to make the tool by design easier to understand, easier to use. Even for those who don’t speak French. I was very impressed by one of our operational directors. He's managing stores. He told me, “Fanjuan, we are employing people who don’t speak French”. It's quite common for people to not speak French but still need to use the tool. So how can they use the tool? How can you help us design some tools that are independent of the French language?” This is very interesting. Things that I also ask my team, we try to work on that. Now, if we say that we try to adapt ourselves, the tool is simplified. It doesn't mean that people will, do you say, buy it naturally. So now it comes to transformation, training, and so on. So on this point, we try to do two things; one is that we do training with anticipation. For example, with data: On topics like data and AI, we gave the team different levels of training quite a while ago. And we repeat this kind of training to them. So even if they do not have day-to-day experience with this topic, they already heard it from the news, they heard it from their partners. Now we give the training to them. This makes things more concrete. And then, when we develop tools and deploy them, we do have a lot of transformation workshops, where we present how they work. I still remember one thing that I did recently, we wanted to deploy some BI reports and tools to a department, which told me, “Oh, I just want my PDF. I just want my Excel because it's so much easier.” I didn't say no, but I said, “Maybe you give us a chance to show you something differently”. But before doing that, what we do is try to talk to the head of this business unit and convince him. So when he creates a workshop with us, he becomes our sponsor. He is also there promoting the tool with us. So it's much easier. So that's one of the things. And I was very appreciative of this business unit leader who is taking on this role. And it helps a lot. So when we present our BI tool to the users, they totally get rid of PDF and Excel because they say, “I didn’t realize how convenient and efficient the tool can be”. I think this is one of our roles, because sometimes we can't ask people to imagine something that they have never seen. Sometimes we have to be a little bit more of a pioneer, by trying to be an ambassador, trying to convince, trying to represent this kind of technology so that people, when they see it, they can adopt it and recognize that it’s useful. 

Clément Marty (19:52)

That's a nice story, the idea of “Please, make my PDF” and, well, give me a chance of changing your mind. Maybe one word on the collaboration with Ardian and how, hopefully it has helped you. If yes, in what way?

Fanjuan Shi (20:16)

Well, first of all, I really appreciate, your support and all the support from Ardian on the data and AI journey. I think the collaboration is quite interesting from both a strategic and operational level. From a strategic point of view, you give us a lot of support overall in the direction and priorities we are defining together. And, we are your first member in the data and AI maturity assessment. So that's really interesting. Maybe you can also, highlight a little bit on this topic.

Clément Marty (20:53)

One thing we have built and tested with you, because you've gone on a long journey and you are very sophisticated today, so we thought you would be a great company to test it, is a data readiness assessment model that we've put in place. And this is basically ,a set of questions that we ask to, of course, tech people, but also business people at our portfolio companies, to obviously understand where they are with data, both technically, but also in terms of use cases, in terms of business, in terms of tech infrastructure. And then from there be able to design with them; their journey, where and how they could improve. So, we believe that this tool is very useful for increasing the data maturity of companies, and your support and feedback on this tool were super useful. So, thank you very much for that.

Fanjuan Shi (22:01)

And for us, it's quite a useful tool because we can benchmark ourselves in the journey but also benchmark against the other portfolio companies so that we know if our auto assessment is really reality or it's different. So, this is quite interesting. The benchmark is always something that adds a lot of value. On the strategic part, I also appreciate that we regularly review our priorities on a quarterly basis. And also, our budget. Your team and you give us a lot of useful suggestions on how to prioritize our technology, investment and it’s quite an interesting collaboration.

And then on the operational part, I think this is also quite impressive. I think, first of all, Ardian connects us to a very large pool of experts, partners, who are providing software and expertise. You are also helping us get very good and ideal arrangements. So all these kinds of things are helping us use the least budget to maximize results. So this is a very, very useful help. And on top of that, we launched a couple of projects together. You have GAIA, and we were inspired by GAIA, and we worked with you, based on your suggestions, on a personal GPT and based on our own chatbot tools, which saves us a lot of time because you have already gone through all the journeys, so you know, what the optimizations can be used for. And when our engineers and I have interviews with you, we save time thanks to your experience. So these kinds of operational collaborations are very helpful. And last but not least, I still remember that Ardian is working with schools, the best class schools, on student projects. And we have the privilege to be chosen and to be part of this journey, supporting the schools on the projects that we are interested in, and also with the help of Ardian. So, this is also very interesting because the results delivered by the students becomes part of our inspiration to work on pricing elasticity projects.

Clément Marty (24:30)

I like the examples you gave out very much. The things that illustrate very well the support we tried to bring, which goes from strategy, efforts, and budget allocation strategy, efforts, and budget allocation like people, providers, and experience-sharing. And I take this opportunity to thank you for your very active participation in the various experience-sharing events that we organized, where we try to have portfolio companies, obviously not competing and kind of, part of the same family, to meet, talk together and learn from one another. And the objective is that with Ardian, you go faster and more efficiently that then you would alone.