As vice president and CIO at PPG, Bhaskar Ramachandran is spearheading a multipronged transformation strategy to fuel innovation, customer growth, and new capabilities through cloud technology and AI. With a track record of delivering business impact across several industries, Ramachandran leads with a people-first mindset that helps accelerate strategy.
On a recent episode of the Tech Whisperers podcast, we unpacked Ramachandran’s leadership playbook and explored how the people-driven culture he’s built at PPG has aligned a team of 1,500+ digital IT professionals behind his ambitious vision and goals. Afterwards, we spent some more time talking about PPG’s journey to the cloud and how the paint and coating manufacturer is using AI to improve efficiencies, speed, quality, and customer value. What follows is that conversation, edited for length and clarity.
Dan Roberts: What did you do foundationally to set your cloud journey up for success?
Bhaskar Ramashandran: The cloud journey here at PPG is the result of lessons learned from my roles prior to PPG. Everyone talks about ‘cloud first.’ The problem with cloud first is it leaves a lot of room for interpretation, and every large organization has pockets that have little incentive to make a change. Those pockets resist the pronounced change, and cloud first becomes less effective.
Because of that, at PPG we changed it to ‘cloud only.’ There is no room for interpretation when it’s cloud only. ‘Only’ makes the mission extremely clear: This is what we are going to do, and this is all we are going to do. That moves the organization. There will always be a small group that thinks this is a fad, this will go away, but once you start gaining momentum, that fades away fairly quickly, because there is no alternative.
Then you start measuring it on a week-to-week basis, and you talk about it. ‘This week, we shut down X number of servers.’ That’s the level of granularity we track this with. I can pull up a Power BI chart that’s updated live when servers go down, when workloads go down, that tells me the status is 93.7%, which is what it was when I checked it last. In reality, we are at about 97%, but for one reason or another, the old servers haven’t been powered down, so they don’t get credit for it until they do that. Workload is out, but the server still has its lights on.
What are some of the business outcomes and benefits of being at 97%?
Those numbers have little meaning if it doesn’t make a business impact, and the business impact is sheer agility. Any respectable software vendor these days does not develop on-prem software unless it’s something specific in a regulatory environment or something very niche. Outside of that, very few people do, so all these innovations and capabilities are happening in the cloud.
If you are not in the cloud, it’s very difficult to consume it. If you are in a hybrid environment, it’s very difficult to consume it. And a hybrid environment, by definition, is too much plumbing. You are unnecessarily complicating the architecture. When you do that, you have no choice but to make compromises on cybersecurity, and it’s never a good idea to compromise on cybersecurity. Or you overspend on cybersecurity, so the benefits of having something on-prem goes away because you’re not saving anything.
When you talk about the cloud, it’s not just the workloads and the flexibility; it’s also the control plane. You have a single control plane to enforce security controls, SOX controls, all the controls we need these days. If something is wrong, push a button and it goes across all the workloads. You have a level of control that is unbelievably sophisticated, and it’s getting better.
The only exception to the rule is speed-of-light issues. What I mean by that is when you have SCADA systems or safety systems in a production floor, and you need to have guaranteed latency. As good as the latency is in the cloud these days, it’s not guaranteed, and you never want to take any chances on those kinds of systems.
Those are the only exceptions at the edges where we have local workloads. Outside of that, there is no data center. We will be literally powering down our data centers. At our global town hall this morning, there was a presentation to the entire global digital IT team about how we are going to do deconstruction of the data center. It’s a huge project, in and of itself, deconstructing a data center.
When it comes to the AI journey, there’s almost a tale of two CIOs right now. Some are going on the offense as early adopters, but a lot are not, due to risk, uncertainty, and other issues. You’re an early adopter who’s been at this for a while. Can you share some ways you’re using AI?
We’ve had our AI COE about five years now. We look at AI as a science-based organization. There are parts of our business that need deterministic answers. Every time the question is asked, the answer needs to be exactly the same. As a coatings company, when you come up with the formulation, the primary question is, how do we create new formulas faster? We invest about $700 million yearly in R&D, so how can we make our chemists a little bit more productive? A single percentage point is a huge win for us. We can be faster to market. We can be faster to serve our customers.
That’s one space. The other is quality specs. Let’s say you buy a car and your neighbor buys a car, and the OEM has a plant in two parts of the world or two parts of the US. You both bought a red car, same brand, same model, same everything, but they just happen to be manufactured in two different sites. How would you feel if your red is slightly different than your neighbor’s? The question for us then is, how can we guarantee the characteristics of the paint, the color of the paint, and the quality of the paint are exactly the same, no matter what the weather conditions were in plant one and plant two?
Let’s say it takes two days to make the paint, and three or four days to qualify the paint. If we can reduce that cycle time, that’s a huge win. We basically increase our production capacity without doing anything. So, instead of making 10 adjustments, can we do two or three to get to the golden batch? That’s where we use AI to reduce the cycle time it takes to qualify the paint — because we have a rich history of data to come up with modeling on taking samples from the existing batch and identifying those corrections we need to make to get to the golden batch.
What can you tell us about your generative AI journey?
There’s a lot of buzz around gen AI because it’s accessible to everyone, so people think it’s easy. When we talk about GPT, everybody gets the G part, because we interact with the generative part of it. But the PT — pre-trained — part gets forgotten. There are years it took these LLMs where somebody tagged the data and somebody tested and built these algorithms and made sure the right things show up when you ask the questions. That pre-training and the adjustments they made gets forgotten.
It doesn’t happen overnight. I think that’s the piece we had to educate a lot of people on. For example, for one of our gen AI projects, we took the rich data we have on the IT service desk and fed that to an LLM and tagged the data. We took eight months tagging all the data and having the internal help desk agents use the chatbot to make sure it’s giving the same answer they’re expected to give. If not, then they would troubleshoot it. This is the PT part of the work that gets it to a point where it can go mainstream. Ultimately, the first point of interaction for all our internal service desk calls will be the AI agent, and then if for whatever reason the AI agent cannot solve it, it’ll go to the human agent behind it.
This has given us a model for how to do this, so for any point of customer interaction, we now have a template for how to take the data we have and convert it into an AI agent.
Christina Cassotis, your fellow S&T Bank board member, had a question about the interplay between humans and AI. In a world where AI is more prevalent, which human skills do you see rising in value?
Geoffrey Hinton, who’s been called ‘The Godfather of AI,’ simplifies AI as glorified pattern-matching. The reason I think he uses this phrase is because the judgment aspects are not there. You give it a pattern and it’ll follow the pattern and it’ll make some variations and make a decision based on whether it’s A, B, C, or D. But when you introduce a completely new pattern, we don’t know how it’s going to react. Maybe it’ll hallucinate, maybe it’ll be right. We talk about deterministic AI. This is probabilistic AI. There are many times we want it to be deterministic, but it’s going be probabilistic, and that’s an issue.
Human judgment and the heuristics we’ve developed from our experiences enable us to make the call the first time we see something novel and enable the likelihood of us being right or the likelihood of us being able to take the responsibility when we are not right. AI doesn’t have this. As somebody once told me, you take chances in life, and if you are right, you win; if you are wrong, you coach. We have that ability. AI does not. So what differentiates humans is that judgment skill and being able to accept the results of that judgment.
The second part of it is just because the AI can do it, should it do it? I want my doctor to tell me I’m sick. I don’t want an AI agent to tell me I’m sick. At the same time, I want the doctor to use the AI agent to inform, to find this obscure pattern that the doctor may not have thought of. But I still want the doctor to tell me that this may be a problem with you, or here’s the treatment or whatever. So just because the AI can, I don’t want the AI to.
PPG has a 140-year track record of staying ahead of the curve. What excites you about the future?
Gary Cantrell, who I worked for for many years and who has been a huge mentor to me, likes to say, ‘If you’re not the lead dog, the view is not pretty.’ If we lack motivation, we have to keep that image in mind. Are we being the lead dog? And if we’re not, are we willing to accept that?
We as an organization have in ourselves that motivation of always wanting to be the lead dog. That drives us, but it doesn’t drive us blindly. We want to do things for the long term, because that is important. Anyone can win in the marketplace in the short term, but more often than not, it is not sustainable. The reasons don’t matter; it’s not sustainable, and history has shown that. You see it with companies that are in the S&P or the Dow or the Fortune 500 and then fall out of it.
We’re there and we will be there, and it’s because of the perspective of our executive management and our board. I give them a lot of kudos for managing us for the long term. It’s great to be part of that culture. It’s great to be part of that thought process and the freedom it gives you to do the right thing for the long term.
For more insights from Bhaskar Ramashandran’s transformational leadership playbook, tune in to the Tech Whisperers podcast.
See also:
Read More from This Article: From cloud first to cloud only: How PPG’s Bhaskar Ramachandran drives bold transformation
Source: News