Inside P&G's AI Factory and the Push to Operationalize AI at Scale
How do large enterprises move AI from experimentation into real business transformation?
Table of Contents
- [00:00:00] Introduction
- [00:00:32] Jeff Goldman's Career and the AI Trajectory
- [00:02:21] P&G's AI Organization Structure
- [00:05:42] AI Engineering and the AI Factory
- [00:08:03] Keeping AI Tied to Business Value
Introduction [00:00:00]
Donna Medeiros: Hello and welcome to the CDO Magazine interview series. I'm Donna Medeiros, vice president of AI and data advisory at The Data Society Group. And today I have the pleasure of speaking with Jeff Goldman, vice president of enterprise AI at Procter & Gamble. Really glad you can join us today, Jeff. [00:00:05 → 00:00:30]
Jeff Goldman: Wonderful to see you and we look forward to the discussion. Thanks for having me. [00:00:30 → 00:00:32]
Jeff Goldman's Career and the AI Trajectory [00:00:32]
Donna Medeiros: So, Jeff, I know you have spent much of your career in analytics and data science. So, as you and we see large enterprises really trying to move from AI experimentation and to that operational state where we're seeing value, what's your perspective on the role of AI and really that business transformation? [00:00:32 → 00:00:57]
Jeff Goldman: When I look at my motivation and my interest in analytics, my interest in AI, it's really been my entire career. And if I reflect back on my time in university, I started by studying the combination of economics and operations research. And those two really hit the intersection of how engineering approaches can change business outcomes. [00:00:57 → 00:01:17]
And even at the very start of my career, you realized our entire industry transformed that way. If you look at something like airline pricing or some of the early transformations in yield management, it was analytics and AI fundamentally transformed the way our business operated. [00:01:17 → 00:01:29]
And I came to P&G because I was excited about that possibility. I was excited about the possibility of applying AI at scale and really operating on a global basis, applying algorithms to change how the company operates. And I think you're seeing that come to life more and more. [00:01:29 → 00:01:46]
And I know a few years ago my dad asked me was I did I get lucky or did I know that this analytics or AI things would come so big? And my comment was a little bit of both. You saw the potential, you saw the seeds of what the decision making at scale could look like, and now the technology is finally catching them up with the potential the math always held. [00:01:46 → 00:02:04]
So, I think it's a really exciting time. I think the types of business problems we can solve today are orders of magnitude more interesting, more complex ones we could solve a generation ago. I think the trajectory of that was always clear, and I think we're on a path to really transforming how companies operate. [00:02:04 → 00:02:21]
P&G's AI Organization Structure [00:02:21]
Donna Medeiros: Yeah, that's a great perspective, too. In the your career journey has really just enabled you. I know it's as it's an incredible opportunity to lead AI for a global enterprise has taken hold. So, I want to have a follow-on to that and if we looked at for us how you've had to structure your AI capabilities in the organization to scale value and lead what you have there is the one AI organization bringing together different parts of the organization, whether it's engineering or data scientist, and I think there's mention of Procter & Gamble and AI factory. Could you tell us a bit more about all of those things, what prompted that structure so you could really have the AI capabilities to scale AI effectively? [00:02:21 → 00:03:10]
Jeff Goldman: Absolutely. So, as I think about my entire philosophy of how I set up my team to succeed at a at a company like P&G, there's almost two critical aspects. One is local relevance. Second is global scale. And I think an organization structure has combined the best of those two elements. [00:03:10 → 00:03:27]
And as you play it out with my team reflects the combination of almost every aspect of what you need to bring AI to scale to company. So, we have data science team that's really about the innovation and what the new algorithms we want to build and how we can transform business processes with AI. You have the AI engineering team that's all about how do we scale those algorithms and make them truly production-ready, how do we operate them reliably, and how do we integrate them to downstream digital products but within within the product organizations we partner with to bring those in into our business. [00:03:27 → 00:03:57]
Then you have the AI factory team which is all about the platform and the capability we need to truly operate that AI at scale and do it dependably, reliably, and with as much automation as possible. And each of those three is structured in a way to really maximize what we can deliver for the company. [00:03:57 → 00:04:14]
So, I'll start with the data science team which is really where I've been the longest and we started the current data science organization about 11 years ago. And we did it with hybrid design where we were looking at both and both how can we be globally relevant, how can we be locally relevant, and how can we have the required expertise in the business areas we need to transform to be designing the most relevant algorithms. [00:04:14 → 00:04:37]
And so, with that, if you look at our design, we have regional leaders in every major market around the world whose day whose job is to really embed themselves to the local business team, understand the priorities, and customize any algorithms we're building to make sure they're as locally relevant as possible. [00:04:37 → 00:04:55]
Then there's three primary domain areas my team operates. One is marketing, one is retail, and one is our supply chain. And I have global leaders of each of those domains whose job is to make sure that they're partnered with the the marketing leaders in the company, the supply chain leaders in the company to transforming and building and partnering on algorithms to transform their business processes. And they then collaborate with our regional data science leads to make sure the algorithms they're developing have that local relevance. And if you look at the combination of those two, I think it really sets us up for success with transforming business outcomes. [00:04:55 → 00:05:26]
There's also an additional organization we partner very closely with which is our research and development organization and they actually have another team of data scientists who are really focused on molecule product discovery using physics-based models and AI-based models and combining them together to really change how we design and bring our products to market. [00:05:26 → 00:05:42]
AI Engineering and the AI Factory [00:05:42]
Jeff Goldman: Now, behind all that, you have the AI engineering team that I talked about at the beginning. And AI engineering team was founded a couple years after we started the data science team and came into my organization much more recently. But they've always been our critical partner in how you bring AI to market. [00:05:42 → 00:06:00]
And so, I learned very on very early on in our data science journey that as we're building groundbreaking algorithms, the biggest challenge often wasn't the algorithm itself, but how do you scale it and how do you operate reliably in business processes. And I remember many times the AI engineering team quite frankly did the impossible in making something that would be impossible to solve in any reasonable business timeframe and make it something that we could bring to market with the speed and reliability we needed. [00:06:00 → 00:06:27]
Now, that's all backed by AI factory team. And if you look through our AI journey, we start we started proposing AI factory around 2021. And at that point, we had a very large team of data scientists that was trying to collaborate globally on algorithms, but there was no standard way we were building algorithms and building and deploying things. There was no no standard infrastructure, no standard way that algorithms supported production. And there was a lot of time lost in provisioning environments and replicating best practices and other things. [00:06:27 → 00:07:00]
So, that led us to create the AI factory, which we actually took the slogan live life model more because it was really a push of how can we get our data scientists modeling, how can we get our engineers scaling, and really automate all the toil around bringing algorithms and models to market. And that's been a great ride. It's been a very successful one. We actually did a case study with Harvard Business School on that about a year ago just detailing that journey of from concept to factory. [00:07:00 → 00:07:26]
And so, when you put those three together, you have the platforms you need to scale AI, you have the engineers to do that scaling, you have the data scientists you need to drive the innovation in partnership with the business. And then most importantly with the or equally importantly with our product partners in a number of organizations throughout the company because the best algorithm in the world not integrated to a digital process doesn't deliver any value. It only delivers value we can partner with organizations to really bring that to bear on our business. [00:07:26 → 00:07:52]
Keeping AI Tied to Business Value [00:07:52]
Donna Medeiros: Wow, that's quite a journey and by the way, I think you've had to orchestrate a lot, right? And through the years building new teams and so on is just amazing. So, I'm going to ask this question that kind of goes to the heart of the partnership model that in orchestration that you lead there, which is how do you make sure that it's leading to that this transformation? A lot of what you just discussed in the value discussion and that makes sense. But we see a lot of companies that are running things as a technical project in the end or they just happen to run. So, how do you ensure that the partnerships lead to that transformation continuously rather than becoming a technical project like it does for some? [00:07:52 → 00:08:36]
Jeff Goldman: Well, that's a great question and when you're dealing with people working AI, they often want to use the most complex model or the most advanced analytic technique. And sometimes that's needed. Sometimes that's what actually solves the business problem. In fact, often it is. But it should always start with a business problem and a business problem that you're trying to solve that's, you know, going to deliver material value to the enterprise. [00:08:36 → 00:08:56]
And so, I go to couple different directions. One is if you look at our structure where we have our our regional leads embedded in the business and we have our you know, we have our global domain leads partnering with both the product teams that are scaling applications for those for those domains and also partnering with the business on how to transform business outcomes. You start with a mechanism that almost injects you into that business conversation to begin with. [00:08:56 → 00:09:24]
Second I would say we're very focused on co-locating our data scientists with the business. So, if you look at where our data scientists are located, they're in our business hubs and that's a very deliberate choice. Every mile of distance of separation between your the people trying to transform business algorithmically and the people working on the business leads to disconnect. It leads to chasing things that aren't going to deliver the maximum business value. [00:09:24 → 00:09:44]
I think it's a third thing I would say is we have additional mechanism which is almost our entire organization is funded directly by the business. So, we engage with them on the projects they want to work on that are going to solve their business problems and they fund us directly for those. And as a result it's almost a foregone conclusion they're choosing to invest their dollars in an algorithm transformation of their business or they can choose to invest their dollars elsewhere. So, it gives a continuous reinforcement measurement every time I start a new project is this where I want to invest my next dollar to deliver the maximum business value. [00:09:44 → 00:10:17]
Donna Medeiros: Right. Okay. Well, Jeff, this has really been just packed full of great information, discussion and you're truly a business executive, right? And a positive one. So, thanks so much for the conversation today and being with us. [00:10:17 → 00:10:30]
Jeff Goldman: My pleasure. Thank you for having me. [00:10:30 → 00:10:31]
For more interviews and insights, please visit cdomagazine.tech. [00:10:31 → 00:10:38]