AI Accelerates Insight, Experimentation, and Measurement at Kimberly-Clark

Jeff Mosier:

Hi, I'm Jeff Mosier, the marketing content lead for the Infosys Knowledge Institute. I'm here with Saurabh Gajjar, he's the marketing analytics director for Kimberly-Clark. Saurabh, thanks for joining us today.

Saurabh Gajjar:

For sure.

Jeff Mosier:

Tell us a little bit about how AI is changing, how your marketing organization operates.

Saurabh Gajjar:

So for us, as a CPG company, we are not at the forefront of adoption of AI. However, AI, of course, like with other companies as well, has already made inroads into our organization as well. So the most basic application of AI across all of CPGs is using AI for creating for creative purposes. And that's the same thing for us as well. That's where we've started. A chunk of our creatives are now being created using AI and being put to test and then out in the market. The other application would be mostly operational where we are trying to save time, where we are trying to cut down on time and gain efficiencies on things like brief writing. There are a lot of other data and analytics related applications as well within the organization where we would lean on AI to be able to sit on our data and give us agentic answers based on what it has learned from our data.

Jeff Mosier:

So where are you getting the most value? Is it the content creation? Is that the greatest value or are you getting more value from other areas?

Saurabh Gajjar:

Me not being a marketing person, I'm on the data and analytics side, so it would be hard for me to speak towards creative and all. But for me, as a data and insights person, we have a lot of value derived from the immediate answers that we can get out of this from the visualizations, from natural language, querying that we are able to do out there. And on the insight side, the additional application that we now have is AI can create synthetic groups and can act as a synthetic control group.

And what it does is it learns from our data, social data, other publicly available data, ratings and reviews and all those places, and pretty much act as its own focus group. At that point, we can start asking random questions to them and utilize it for innovation, utilize it for messaging, utilize it for creative content, utilize it for trying to figure out what would be the claims that we could put on our products. And it will answer using all the intelligence and all the data that is gathered, it'll be able to answer all of this and act as a human panel for us, which has probably been the most useful area for AI in terms of insights and analytics.

Jeff Mosier:

How difficult has it been for you to evaluate how well that's working and how effective it is, especially whenever you're working with synthetic data?

Saurabh Gajjar:

So for us, this is in nascent stages. It's not something that we are inherently trusting yet. And I don't believe that any organization or a CPG would in today's date. You would still use it as your initial frame of reference to probably winnow down from a large list to a smaller list. You will then use that with an actual panel, with an actual human running the analysis or something like that and try to triangulate before completely trusting the AI. It will probably take us months of testing and months of understanding the kind of output that we are getting from AI before we fully trust it.

Jeff Mosier:

So what kind of surprises have you seen in AI adoption and marketing as far as values, challenges, things like that?

Saurabh Gajjar:

There has been several areas. So one is on the fidelity. Initially, when AI came out, it seemed to be like a solution to all of our problems that has not been the case for us, at least on immediate basis. Like I said, the fidelity has been an issue. We have seen a lot of hallucinations when it comes to data and all. When we are feeding our own data for AI to work upon, it comes back with lot of incorrect responses. If we were to go back and check, so just the fidelity was an issue. The second surprise has been the application of it. The application has been much broader than what we had perceived it to be initially. Initial perception was pretty much on the creative side. Now you can see that the application is everywhere. With the evaluation agentic AI, there's so many other applications that we've been able to derive out of it.

And then lastly, the adoption within the organization has been surprising to me as well. From zero, it seems like we've gone to almost five or six where everybody has started to talk about it. In every portion of the organization, there is at least a work stream or two that has started that focuses on driving AI into our day-to-day work.

Jeff Mosier:

How do you think about the ways to scale it up and find these new use cases and expand them? Do you have a specific experimentation program to figure out what's going to work, what you think's going to work and evaluate it?

Saurabh Gajjar:

So more often than not, what we've been doing is putting it into the hands of the people that use it and try to figure out if they are able to replicate the results that we've been able to replicate outside of AI. So for example, as me having an analytics team that tries to parse the data, dissect the data, run data quality, create data visualizations, run models and whatnot, as I start putting some of those tasks in the hands of AI, I'm doing the triangulation. Those people who are working on it day-to-day, they are doing those triangulations as well and trying to make sure that it is providing us with the right answers.

Jeff Mosier:

Do you see AI changing how you measure marketing outcomes or measure marketing value?

Saurabh Gajjar:

We certainly do. So at some point, the pace of how we measure is definitely going to get affected by AI. So in today's date, our standard measurement, for example, MMM, would come back within a certain amount of time. Once the MMM comes back, a portion of my team will start working on it, start delivering it to the brands. And every time a brand has a question, there's come back to my team, probably ask that question. Now, imagine an agentic AI sitting on top of all of that MMM results and answering all of those brand questions without any intervention from my team, that all of a sudden that interaction which lasted from two to three weeks after the delivery now gets shortened to zero time for my particular team. And for brand, they can go to any extent they want in terms of the questions they want to ask out of it.

Jeff Mosier:

You mentioned agentic AI a couple of times. How far are you into your use and experimentation with agentic AI and where do you see that going in the near future?

Saurabh Gajjar:

So for us, I mean, that has been the major application. I mean, all of the things that I've been telling you, when we lay over AI on our data, that is an example of agentic AI. If we think about, for example, Amazon Marketing Cloud, historically you would have needed analysts who are at least knowledgeable in SQL to be able to extract data out of AMC, versus now AMC themselves provide you with an AI agent on top of AMC that cannot create queries for you. All I have to do is formulate my question in a particular way and do the right prompt, and it'll create the query that is required. An analyst job can now be done by a marketer. So as we are discovering all this use cases on day-by-day basis, the penetration of agentic AI is just continuing to increase within our organization.

Jeff Mosier:

What's the next horizon for agentic with you guys and what do you think might be the next horizon for marketing in general?

Saurabh Gajjar:

So it could lay over everything and anything, any intelligence that we as an organization have and that could derive multiple use cases going forward. The extent to which we can utilize it is not something probably we are still realizing. For us, we are still starting to dabble into this arena and try to drive some of the use cases, but it's hard for me to comprehend where this would go from here, honestly.

Jeff Mosier:

Obviously the big benefits of agentic is being able to have these agents take multiple steps and actions without human intervention, making decisions, taking actions, things like that. How difficult has it been, or do you think it will be to figure out where to insert a human in the loop and make sure that it doesn't go astray?

Saurabh Gajjar:

That's a great question. For now, I did mention fidelity earlier in this interview, and we see AI going pretty astray with our data. So unless we are able to rely on it and we have triangulated the results over and over again multiple times, we would not trust AI with our data and the results that it's producing.

Jeff Mosier:

More broadly, looking at responsible AI, how do you and your company approach that in putting up guardrails to ensure that you lessen the risk out there?

Saurabh Gajjar:

Well, 100%. So one of the applications that we have today is generating PDPs for us. When AI looks through the reviews and the publicly available data and create those PDPs, I would imagine there have to be certain regulations to be put that ensures that the PDPs contain information that is true to what the product is. There are multiple lawsuits that we have seen out there in the marketplace where there is overreliance on something like this and it provides falsified information. So, I mean, I'm not very close to the legal side of all of this, but I would imagine that us as an organization is doing everything in its power to make sure that all the information that's being provided by an agent like this would or by a creator like this is all true to the product that we are creating.

Jeff Mosier:

On the people end of it, what do you see as the kind of need to re-skill or upskill employees or get them to understand what is going to be needed for the new jobs that they're going to have as they evolve pretty quickly?

Saurabh Gajjar:

Yeah. Understanding the technology itself is such a huge challenge. There are so many solutions out there. So one, just understanding the range of solutions that are available for your roles is a big one. From there, as you start utilizing some of these solutions, there is prompting that needs to be right. Once the results come through, the validation and QCing of those results has to be another step that they need to do. And once you start using it, it has more to do with what are the right applications and figuring out those applications as they go.

Jeff Mosier:

What are your change management needs like? Are people embracing AI? Is this something that's exciting? Or are they worried about it? Or what kind of combination is out there?

Saurabh Gajjar:

I would say a little bit of both. I mean, even in this conversation, we talked about some of the areas where that analyst role gets eliminated, where some of the basic needs that come from data visualization, data parsing, and all of that gets taken care of by AI. So there's definitely some fearfulness about it. At the same time, lot of the folks have adopted it given the kind of efficiency generates in your day-to-day operations. It's a mix of both across the board.

Jeff Mosier:

What are some of the emerging challenges you're seeing with AI? What are going to be new challenges that you're going to have to overcome to continue? I'm assuming y'all want to continue bringing more AI and weaving it into the company and marketing operations. What challenges are you going to face with that?

Saurabh Gajjar:

So fidelity is something that's always going to be an issue, then comes adoption. Even though the fidelity is going to be there, it's hard for people to just switch over to something like this. So adoption is going to be another one. And then the last one is both scale and imagination. So you could go to any scale with this and only our imagination limits us from what you're able to do with AI.

Jeff Mosier:

Thinking more big term overarching, what are the major implications for AI and marketing? How do you think it's going to change it and potentially restructure it?

Saurabh Gajjar:

To me, it's both scary, but at the same time quite exciting. The scary part of it is the limits of this is far beyond what I can think through or anybody can think through in terms of creative, in terms of audiences, in terms of the scale at which this can grow and excitement is pretty much stemming from the same stuff as well. Our jobs can be much more easier from where they used to be. The content that we can put out there could be much more effective and more personalized than what it used to be.

Jeff Mosier:

In many ways, it seems like you're running a race and you have no idea where the finish line is or how far you're going to have to run to get there.

Saurabh Gajjar:

That's a great way to put it.

Jeff Mosier:

Well, thank you so much. We really appreciate your time. Thank you for joining us today.

Saurabh Gajjar:

No, thank you for having me. This was a great conversation. I appreciate it.