Today we welcome Balaji Viswanath, Managing Director of Emerging Tech, Enterprise Architecture, Software Engineering and RPA at Tyson Foods to The CXO Journey to the AI Future podcast. He joined Tyson in June 2018 as Senior Director of IT Architecture. Balaji is a Technology and Data Science executive with 25+ years of experience in his field.
Question 1: First Job: Could you talk a little bit about your background and how you got to the position where you are now?
I’ve had the privilege of being in the technology industry for about 26 years or so now. The first twenty of which I spent in a core technology company, and the last six in Tyson Foods. If I look back at my career, I’d say I spent the first third or so building technology, products, and solutions that we could use in-house. The latter portion was predominantly spent trying to deploy technology in a way that added business value.
If you look at core technology areas like e-commerce, digital marketing, big data, or very recently, Generative AI, you see over and over again that everyone is spending a ton of time in analytics. Everyone wants scalable analytics platforms that enable the business to gain actionable insights from their data.
Over the years it’s been a privilege to be part of high-performing teams. Recently, everyone is very passionate about what Gen AI can do for the industry and what Gen AI can do to add value at Tyson.
Question 2: Generative AI: Everybody’s talking about it. But you, as a business leader, have seen other cycles before. Could you put a little context around Generative AI? Is this the highest priority for you right now
I’m glad you asked that, Gamiel. From my experience and understanding, a very significant amount of prioritization has to be applied to AI today. During my day job, I speak to at least ten vendors a week or ten new product companies a week that are talking about AI in some shape or form. I do believe that the predictive components of what we do in our day-to-day jobs are starting to be omnipresent and very ubiquitous.
AI is different primarily because it’s evolving so rapidly. There’s a sense of “Oh, am I missing out?” So there’s an evolution, a paradigm shift. If you look back at the last year after the introduction of ChatGPT and its capabilities, I think the leap that Gen AI has taken as a technology has led to new products and new offerings every day. So, it’s starting to become a huge priority for most of the companies I speak to, including ours.
Extra question: You’ve been in technology for a while now. Could you talk about some early learnings? What lessons or even advice would you give to somebody who’s in an IT leadership position when it comes to taking advantage of Generative AI?
There are a few things I could talk about.
First: Start with the business problem. AI and data science, in an unprecedented manner, are forcing a convergence between what was traditionally IT and what was traditionally business. I think data science is bringing together these two disciplines very efficiently. So, start with the business problem and don’t go after tech for the sake of tech, or AI for the sake of AI.
The second thing I would say is to chew in “bite-size chunks.” One of the ways we try to explore Gen AI and try to bring it into our company is by starting with a small business problem. Engage very closely with your business partners, deploy something as quickly as you can, and see if you can get value out of what you deploy. And if you see value, then you expand on it a little more and make it big. But if you don’t see value quickly, it’s easier to fail fast and move on to the next thing.
And keep your options wide open. Today, every hyperscaler that we work with has a significant AI offering. It would be, in my opinion, short-sighted to think that we go with one and only one. I think all these offerings that we need are very nuanced in themselves, and I think having MLOps and foundational platforms that allow us to exploit the best capabilities from each and every one of these offerings would also be a key learning that I’ve gathered over the last few months.
Question 3: New Metrics: You and I talked a little bit about your way of measuring impact or metrics. There are lots of use cases where Generative AI could be used. How will you go about deciding where to put attention? How do you cut through all of the potential use cases with a metric that matters? And what are those metrics that matter?
That’s a very good but very complex question. This is where I probably have the greatest number of discussions in my day job. How do you determine the value of something once it’s deployed? Today, there are broadly two kinds of value that we look at.
One is the value that impacts the bottom line, which is around efficiencies and manufacturing…anything we can do to get more efficient as a company. And what AI and Gen AI can do to help us get more efficient.
And the other area of value that we look at is what we call “top-line value and growth.” Does this help us improve our top-line growth?
There’s also a significant amount of thought around how we can employ AI to improve our employee experience across the company.
I met a few people recently that were specifically focused on using AI to do opportunity sizing. So we can say, “Hey, even before you embark on the opportunity, what AI can we use to size this opportunity?”
So that’s our approach. We engage with our business partners and try to get as close to articulating value as possible (in terms of a dollar figure). We approach problems in small iterative chunks, fail fast, move on to the next thing, and hopefully, we’ll arrive at a good value proposition for a large AI project.
I will tell you, though, that an integral aspect of value is essentially the benefit minus the cost. So one of the things that we’re working on actively with all our partners is: “What does Generative AI really cost?”
We have still yet to meet a company that has truly figured this out. Because there’s a computing aspect to it. There’s a training aspect to it. There’s a big data aspect to it. There’s an analytics aspect to it. So we’re hyper-focused on trying to understand what Gen AI really costs and use that to understand its true value.
You’ve been in an industry that has been in the process of transforming its landscape to a cloud-based, SaaS model, and that mostly doesn’t build custom solutions anymore. Do you believe that to be true? What’s the ‘buy versus build’ thesis that you believe you should be thinking about?
Being in enterprise architecture, this is a question that we are confronted with very frequently. Do you buy? Do you build? We started off with a very traditional base layer type approach where we said: “Hey, if it’s a system of record and if somebody in the industry already does it extremely well, then why don’t we go buy it instead of trying to build it?” Versus anything we think is differentiated, we try and build.
When you talk about this traditional tenant, as far as AI is concerned, especially for Generative AI, it’s too early for us to see if we have to buy, or if we have to build. I do think there are use cases where it makes a lot of sense for us to build. I think there are two ways of looking at this. One is either we send our data out or we bring the AI in.
So, there are a few use cases where it makes sense for us to bring open source or other models inside the company, train them with our data, deploy, and see value.
There are a few other bigger use cases where we need the firepower of externally well-trained, hyper-scaler hosted, managed models. We typically build a strong decision tree around this.
Question 4: What are some of the issues, or maybe even gaps in technology that you believe need to be addressed?
Like I already mentioned, I think the first thing we’re trying to figure out is: What is the true cost of Generative AI? What is the true cost of deploying and scaling a Gen AI product inside a company like ours and seeing value? So that’s our first area we’re trying to understand.
And from a skills standpoint, prompt engineering is starting to become an extremely important skill. I did speak with a few people recently who were starting to invest in AI through utilizing prompt engineering which I thought was very, very interesting.
Prompt engineering is a skill that we’re starting to understand and is starting to be a multi-disciplinary thing. It’s not so much only a pure technology play. It’s not like you get the best programmer to write the best prompts. It’s starting to become a very interdisciplinary thing.
And last, but not the least, compute. We have been in discussions with a few people who are willing to deploy AI models for us. The cost at this point is quite untenable without understanding the value. So, I believe that achieving a balance between cost and value is another area or a gap that the industry will need to focus on and address pretty quickly.
Responsible AI: How do you think about it as an IT leader? What does it mean to you?
There’s a lot of discussion about this, I’m sure, in most companies. For us, there are a few things that we go off of.
First, anything that we build, we very clearly say, “Do not let this be the only decision point. This is but an input into overall decision making.” So, in a sense, AI can possibly do a task, but a job is a very different thing. A job includes cognitive aspects and we don’t believe we are there yet with AI as an industry. The cognitive aspect isn’t close to where human beings are at.
So, we don’t put AI in places where we really need to make decisions. We look at AI as an augmented decision-maker as opposed to being a decision-maker in itself.
The second reason for concern is that we don’t indiscriminately send data outside the company to external AI models. We are very careful about how we use our data and treat it as an asset.
Finally, we focus a lot on explainability. We try to understand why an AI model, or a Generative AI model, provides the inferences that it does. There is a team that’s starting to look at explainability in AI.
Balaji Viswanath, Managing Director of Emerging Tech, Enterprise Architecture, Software Engineering & RPA at Tyson Foods to our CXO of the Future Podcast (“The Three Questions” edition). He joined Tyson in June 2018 as Senior Director of IT Architecture. Balaji is a Technology and Data Science executive with 25+ years of experience in his field.
He is a technology leader with a passion for applying technology to transform businesses at fundamental levels. He is particularly passionate about bringing in new technologies to solve business problems and build high-performing teams. Balaji has extensive experience working with business groups across verticals to envisage best-in-class digital experiences and translate them into robust Technology solutions.
His core competencies lie in working with business partners to identify large value-add opportunities by applying Data Science, Machine Learning, and other cutting-edge technologies.
He has led several Digital Transformation and Multi-country/Multi-Channel global platform rollouts in E-commerce, Digital Marketing, Web, Sales IT, CRM, and Supply Chain domains.
During his career, he has implemented state-of-the-art technologies such as Adobe Experience Manager, Intershop Enfinity, Drupal, etc. across large-scale enterprise IT. He is also a proven Change Agent, and has led several teams through grassroots organizational changes and transformations including transitions to managed services.
