Barilla’s Kester Fernandez: AI’s Weakest Muscle Is Organizational Readiness

Kester Fernandez:

AI is not a solution to all tech problems. I don’t think we’ve necessarily cracked the code in sizable business value. It’s less about trusting the AI, but trusting the data that it sits on. We are products that people put in their mouths and feed their kids. We take that seriously.

Jeff Kavanaugh:

I’m Jeff Kavanaugh from the Infosys Knowledge Institute. I’m joined today by Kester Fernandez, Vice President of Digital and Business Technologies at Barilla Americas. Welcome.

Kester Fernandez:

Thank you, Jeff.

Jeff Kavanaugh:

Today, we’re talking about unlocking AI value, and I’d like to come at it from a few different angles. First of all, what is your team doing with AI today that may have seemed impossible just 18 months ago?

Kester Fernandez:

Well, I think just the speed of exploration has been fascinating. This is not just at our organization, but what we’re seeing across the industry is just the access to platforms and tools and the spirit of innovation. I think the experimentation that we’ve done in the recent past, just the pace has been quite fascinating to watch.

Jeff Kavanaugh:

With this pace of innovation, how has it helped you deliver value, and especially where your board’s prioritizing?

Kester Fernandez:

I don’t think we’ve necessarily cracked the code, I’ll be honest. I think most CPG companies like ours started with exploration of productivity enhancements, right? That’s where you typically get the low-hanging fruits. And we did the same.

But what we wanted to do was try and do so with some decisiveness around how do we start? Which platforms do we make available? So we wanted to make sure that we stood up a light but flexible governance model, which said, “Okay, here is how we’re going to make platforms available for all our employees across the organization who have access to these platforms.” Put in place AI policies, because that’s another hard part. For a European-headquartered company with compliance and GDPR and data protection, that was extremely paramount for us to figure out very quickly.

And then make sure that we’re not slowing the pace of innovation down. So while governance had this monolithic approach of well, you’re going to create this bureaucratic process to stem and stop projects from going through, how do we still keep the pace of innovation, but making sure that we not only provide access to platforms, but recommendations on architecture, and more about getting value really back into the hands of our business teams. Because running pilots is fun, but trying to do it at scale at enterprise level is a whole different game that you’ve got to navigate through.

Jeff Kavanaugh:

Speaking of that, how far along are you in moving from pilots to scale, and what’s been your biggest challenge?

Kester Fernandez:

We’ve created our AI journey into three phases. Starting with productivity and enablement, and then the third phase being transformation. I would say we’re at the juncture of migrating from phase two to phase three, where at this point, we’ve got a lot of AI capability that have been put in the hands of our employees, and a lot of production-deployed agents at this point. But I don’t think we’ve necessarily cracked the code in sizable business value. So at this point, we look at it as the investment to learn and get that muscle memory in.

But I think from a standpoint of getting real enterprise-grade use cases, we’re at the early stages of defining that. So we’ve done some work with domains such as revenue growth management or supply chain forecast improvements. Where some of those decision enablers on making smarter decisions have already been put in place, but I think there’s a lot more value that lies ahead of us for us to unravel and figure out.

Jeff Kavanaugh:

You mentioned your employees, your most important asset. How do you help these employees know when AI is the right tool, and how to use it well?

Kester Fernandez:

AI is not a solution to all tech problems. And I think the key piece there is to determine which ones is it appropriate for and which ones they aren’t, is looking at it from a standpoint, if you look at a business process which has some semblance of pattern recognition, some semblance of just high volume, routine decision making that you would want to activate, it makes absolute sense to be an AI-based solution.

But some which don’t have predefined patterns, where you need human judgment, ethics, some sort of a greater accountability impact to customers and consumers, where reputation is at risk. You want to make sure that the AI can be an enabler, but it still becomes a human and accountable process.

So making decisions on which ones can potentially be solved with just a Power BI analytics dashboard, and you don’t necessarily need an AI solution for it, versus some, and it all, I think, to me, stems down on data and the quality of data that is sitting there, and the ability for it to be governed through some sort of an architectural and security tenets. I think that’s paramount for some of these systems to be successful.

Jeff Kavanaugh:

As AI changes the work itself. How are you bringing people along for adoption?

Kester Fernandez:

I think we view the AI onset as another business transformation program. So it’s less about the tech, it’s more about the employees and the people that are impacted that embrace it. And this is a significant change management program like anything else.

So making sure that we have awareness training, a community where we can share knowledge and bring in true business cases where value is being added, and have that percolate across the community is extremely important as an ongoing initiative. It’s not a phase towards the end of a program. This is an always-on campaign that has to be done, and we’ve taken the same approach with AI.

Jeff Kavanaugh:

AI, powerful tools. And they’ll create recommendations whether you want them or not. Lots of them. Can you share a time or when maybe humans had to overrule an AI recommendation that was made?

Kester Fernandez:

Yes. A lot of this is, stems on institutional knowledge that sits in the heads of people. It’s not in a SharePoint document. It’s not in a repo that you can point an AI model to. I think that’s where most companies struggle with kind of it’s not structured or unstructured. It’s sitting in people’s mind. It’s the years of experience that they’ve had, and having models get to that same level of proficiency.

I think for us, particularly being a consumer-facing brand, reputation, whether it’s interactions with our customers, those are elements that we don’t want to take a chance. It’s not a, “Oops, we made a mistake, and do over there.” We really may not have the opportunity to do that in some scenarios. So we want to be extremely cautious where judgment and accountability matters in some situations, where in some other scenarios, it could be low risk.

You’re dealing with a large volume of transactional information with some sort of a pattern that you can make relatively straightforward decisions. We would give it to a more autonomous AI process management, versus some that we still want the accountability and judgment and ethics to be decided by the human-in-the-loop in those cases.

Jeff Kavanaugh:

Regardless of the degree of either risk or the area, if AI does get it wrong, who’s accountable?

Kester Fernandez:

I think in the end, the organization and the brand is, because you are, for a food company like us, we are products that people put in their mouths and feed their kids. We take that seriously. And we can’t afford to be cutting corners in the quality and the brand that we stand for in terms of the high-quality experience that we want our consumers to have.

Jeff Kavanaugh:

At some point, AI will act more independently. What would it take for you to trust it to act independently?

Kester Fernandez:

I think it’s less about trusting the AI, but trusting the data that it sits on. Because I think the technology is there. We have autonomous cars. We’ve got autonomous deliveries that’s happening. I think it’s a matter of the maturity of the data quality and the transparency of the decision making that comes with it, and more around organizational readiness.

I think that’s the weakest muscle in this entire equation, that most organizations aren’t quite ready. And at the pace of change coming in the tech space, and the organizational readiness doesn’t quite match, and I think we’ve got a clear role to catch up.

Jeff Kavanaugh:

Well, fast-forward a couple of years. At that point, what will AI have done for your company, and what will the new frontier be?

Kester Fernandez:

I think we would have a lot more production-enabled automation that AI would be doing for us, where we would have a much more clearer definition and separation of across all the end-to-end value streams. Which are the pieces that fall into that repeatable patterns that AI would be doing a lot of the automation in terms of communication and notification to our suppliers and customers on timeliness of deliveries, and when orders are being shorted and substitutions are being made. Some of that is very transactional, which you have customer service teams today doing it manually. We could do that a lot faster.

But some of the other ones around decisions on what proposals do we take to retailers on cross-category promotions that we want to run, which requires their buy-in. We know a certain posture where the buyer has switched in the last six months, and they come with a different motivation on high protein product offerings that you may have, and you maybe want to bundle your offerings. Some of that judgment call on offerings and what the customer is willing to accept will need some human intervention and accountability, because you’re talking about opportunities, you’re talking about reputation, you’re talking about consumer quality and food. You don’t want to compromise on those aspects.

Jeff Kavanaugh:

Got it. Well, taking a step back. You’ve been a client of Infosys for some time. Talk a little bit about the work or the partnership and what that enabled you to do.

Kester Fernandez:

I would say the Infosys partnership has been relatively recent. We’ve done some early work with Infosys, particularly around shaping our revenue growth management capability. And this was a new domain for us at Barilla, where we were trying to figure out what revenue growth would look like.

We had historically, obviously, running on an SAP ECC environment, a lot of the back-office operations already in place. We had trade promotion system, trade promotion optimization. But moving the needle to say, “Where can we do what if scenarios on opportunities for us to maximize our revenue growth?” I think that was where the Infosys partnership helped us figure that out and create those fundamental knowledge platform blocks.

So we’re in an early partnership journey, and hopefully, and I think from what I’m seeing at Confluence, there’s a lot more capability that Infosys has to offer that I would love to have the rest of the organization learn from to say, “Here’s the art of the possible.” Not just limited to AI, but even beyond.

Jeff Kavanaugh:

Yeah. And hopefully it is a bidirectional relationship where there’s a lot of learning going on. Right. Well, listen, thank you so much for your time.

Kester Fernandez:

Appreciate it, Jeff. Thank you.

Jeff Kavanaugh:

You bet.

Jeff Kavanaugh:

I’m Jeff Kavanaugh from the Infosys Knowledge Institute. Until next time, keep learning and keep sharing.