United Airlines on Agentic AI, Architecture Assurance & Enterprise Scale

Ashiss Kumar Dash:

Hello, agentic AI is emerging as the essential next step for most enterprises as they look at AI for scale, as they look at AI for creating value, as they look at AI for defining and reimagining who they are in the AI native world.

Ashiss Kumar Dash:

I'm very pleased to have Ninos Gabriel here in conversation with me. Ninos is the Managing Director of Enterprise Architecture in United Airlines. And he wears multiple hats, and you’ve been transforming through the AI journey at many things in United. So thank you for joining me today.

Ninos Gabriel:

My pleasure, thank you very much.

Ashiss Kumar Dash:

Ninos, one of the things I wanted to start off with is like I said, agentic is the next frontier for most enterprises. And you’ve done quite a bit of work in this space. Would you like to share some of your learnings, all the things that you find to be, you know, very relevant for our clients, and overall in this journey, things that you’ve observed?

Ninos Gabriel:

Well, I think quite a lot in this space have evolved in the last couple years since we started our journey. I would say the one thing that sticks out the most, as part of beginning in this having this conversation, is really around, how do we make sure that we are positioned in the way that we could evaluate these technologies, the proper way, right? And this is where, I believe, the value in standing a nimble team, that has multidisciplined subject matter expertise, that are working with architecture engineering, application SMEs, security, governance, to make sure that we're baking in the proper best practices as we evaluate these tools. So I think that incubation allowed us to evaluate quite a lot of tools. And those tools have eventually migrated to capabilities, and then outcomes from the business perspective. So one of the areas that sticks out the most to me is our ability to inject AI throughout the lifecycle of SDLC. And in this case, it has really provided us an end-to-end type of view, but also efficiencies across the SDLC.

Ashiss Kumar Dash:

Wonderful. So, talking about SDLC. One of the things I know you're passionate about is architecture assurance. Can you elaborate a little bit on that? Because I've not heard that from too many people. But the way you are doing it, the holistic approach you have taken on AI, seems like, you know, our clients and our friends at Infosys could learn a lot from that. So can you elaborate on this?

Ninos Gabriel:

It's not a new concept. I think artificial assurance have existed in the last 20 years or so, and it's been done manually for a long time. Which is basically the ability to guarantee, or in this case, you have the assurance that your design matches your production. And in some cases, there's going to be some deviation between the design phase and what's being deployed in production. One of the things that we've done at United Airlines has been very helpful, is to marry those two together. We want to make them closer together. And the reason we're doing that, twofold, really, is we want to make sure that deviations are not happening from a technology stack, and or applications that are pumping out into infrastructure that potentially could be costly in the cloud. So we have a set of patterns from a security, from a cloud FinOps perspective that we inject in that. And I think the second is making sure that our incident management and problem management teams are also very proactive in identifying what is in production, by looking at documentation that have produced as part of the playbook of the application. So both of those areas have significantly increased in value and efficiency when we included architectural efficiencies there. But it has also given us the opportunity to actually review any Snowflake deployment and determine where we could focus our efforts.

Ashiss Kumar Dash:

Wonderful. And as you deploy these at scale, one of the things that probably will be focused on is, through this SDLC lifecycle, what requirements you capture is not necessarily what gets deployed at the end, right? How are you engineering that aspect of no loss or no drift from your department?

Ninos Gabriel:

Yeah. And that's a tough one, right? Because requirements come through all the time, as part of iterative process, which we allow that, because that's where innovation is mostly there. When we're going through an iterative process, we want to make sure that we have a good set of requirements in the beginning. And we know that there's going to be some drift that's going to happen. We want to not only record that but also be aware of it and maybe get in front of it. So one of the things that we're doing at United is, we have a repository, an AI repository of what we consider very good business requirements. And as part of the business requirements capture, we actually make recommendations, hey, we hear you talk about these business requirements. By the way, historically, other business requirements are part of that. So we're making those recommendations and allowing the subject matter expert to select them, which has really impacted how many changes we go through this cycle. So before, maybe we'd go a little 50/50 on business requirements, now we're seeing an upward, a lot more 85% or 90% accuracy on the first phase, which allows us to not do the boomerang effect and go back and look at the requirements and redesign it.

Ashiss Kumar Dash:

That's awesome. So you can actually point out if some requirement is missing. Because the cost of retrofitting requirements at a late stage, SDLC is much higher.

Ninos Gabriel:

Absolutely. As you go down the stream itself, right? You're going to hit development and the coding, right? It was based on that requirements, then it goes into UAT testing, it goes into design. And some of the regression testing, all of that is based on those requirements, both technology requirements, and business requirements. And if there's a deviation in the future, then all of that chain would have to be redone.

Ashiss Kumar Dash:

Absolutely. And you have put together an ecosystem of partners, and very thankful for Infosys being one of the partners. Can you talk a little bit about what's your expectation from partners like Infosys, going forward, what has been some of the things that we’ve done for you, and just, you know, setting the expectations with our team?

Ninos Gabriel:

Yeah. Well, I think Infosys is a fantastic partner. And I say this because you guys have helped me achieve a lot of the objectives that we wanted to hit. Not from an organization perspective, but the entire technology teams and the business. For example, you guys were on the ground with us when we started establishing some of the enterprise standards about how to actually look at solutions delivery, including DevOps platforms, like Harness. So, you guys not only had the intimate knowledge of how our applications run, but you had the resources knowledge, you had the subject matter expertise to bring the technology to bear. So I would say, continue to be that partner that understands our business, right? We're a unique business, and airline business is definitely a complex, heterogeneous environment on the back end. And we have to be very sensitive in when it comes to reliability, security, resiliency, and even latency, right? There's a lot of sensitivities there. You guys understand that very well, and everything that we've done from an engineering and pattern’s perspective has been very accounting into all these conditions that we need. So it's been fantastic.

Ashiss Kumar Dash:

Thank you Ninos for being here. And thank you for sharing your insights. I'm sure that the journey has just begun. There will be many more milestones of great success in the future.

Ninos Gabriel:

Thank you. My pleasure. Thanks for the partnership. Cheers.