Achyuta Ghosh — Executive Research Leader, HFS Research[00:06]
Hello and welcome back to the HFS GCC Advantage. I am Achyuta Ghosh, Executive Research Leader at HFS, and this is the show where we get to speak with GCC leaders who are building the next generation of the model and share some practical stories on leadership, talent, AI, operating model, etc. Today I’m actually joined by Sameer Shaikh from XPO for a conversation on one of the most important topics enterprises are facing today: how do you move AI from pilot to experimentation to real enterprise impact. Sameer brings a very unique perspective across technology, financial shared services, and GCC leadership, and that makes this a very relevant discussion today. Sameer, welcome to the show.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[00:50]
Right. Thank you Achyuta. Great to be here, and thank you so much for having me.
Achyuta Ghosh — Executive Research Leader, HFS Research[00:57]
Great, Sameer. So, before we get into the crux of the discussion, can you give us a brief overview of XPO, the role of the GCC within the enterprise, and your own journey leading various technology and financial shared services, and what you’re doing today at XPO?
Sameer Shaikh — Senior Director, Technology & FSS, XPO[01:13]
Absolutely. Well, I’m Sameer Shaikh. I lead the GCC operations of XPO out here from India, based out of Pune. XPO is a logistics transportation company headquartered in Greenwich, US. You know, 38,000 employees, more than 400 service centers. We are basically one of the leading LTL operations across North America, spread out in Europe as well. Our operations are primarily on the transportation. So we move freight in the US. Effectively, everything that we do is around technology. Now our freight, our logistics operations, they move things from one location to another.
And I think from the India standpoint, we are, our GCC basically supports various functions that enable, or we are basically like a technology and functional backbone for our teams operating in the US. We have technology, we have financial shared services, people services, BI, and other functions that operate out of here. Primarily, everything is moving over to AI. So we do a lot of R&D and implementation and innovation out here from our GCC center here in Pune and Hyderabad.
Achyuta Ghosh — Executive Research Leader, HFS Research[02:39]
Great, Sameer. Many organizations today have access to the same AI platforms, tools, and technologies, right? But what we see is that some organizations are moving a lot faster than the others. And I think one of the key reasons why this is happening is probably the company culture, right? In your experience, what role does culture play in determining whether an organization successfully adopts and scales AI? You know, what cultural traits do you think accelerates adoption? What behaviors actually you think slows down adoption? And how do leaders frame this entire culture focus?
Sameer Shaikh — Senior Director, Technology & FSS, XPO[03:20]
I think, you know, probably more than leaders that understand that AI is just a tool, are probably the ones that are going to basically succeed in the race ahead. And when I say that, I think culture plays a very important role. And let me put it this way. Culture is basically, the cultures that succeed in AI adoption, the cultures that are succeeding in making a difference are basically those cultures where they have a learning mindset, where they have a mindset to experiment, where they have a mindset to fail and fail early in the journey. And most importantly, I think the drive for AI is not driven by technology but by business outcomes. So when those aspects are embedded in the culture, where the culture does not treat AI as a project but AI as a habit, I think that changes a lot.
So to answer your question, I think leaders play a very important role, and I think one of the most important roles that they play is providing psychological safety for the technology teams in making sure that the experimentation is not treated with a certainty aspect. I was just talking to someone a while ago, and it was being said that we are working on an AI project that is going to create a model, and it will basically help them make a very heavy impact on the bottom line. And we just want to be sure before we start the project that the ROI is very defined. And I was like, well, that’s certainty. You’re basically looking in for an outcome with a certain tool. AI is going to be experimentation, which will allow you to basically change the process.
And I think the leaders also need to play a very important role in removing the silos, the functional silos, especially when AI initiatives are driven. I think technology plays a separate role, they’re doing something in one area, and then there are functional requirements and business requirements that sit in the other case. That integration, collaboration does not happen, and I think that drives the outcomes of every AI initiative. And a lot of times when we see the way pilots shape up is amazing. You can see a POC or a pilot basically running with full success, and then when you talk about scaling it up, you realize that all those silos that you had created up front during the experimentation time are making an impact and not letting the project scale up because you have missed out a lot of things.
And I think the third aspect, which is governance and guardrails, basically the whole governance and data element. I think AI needs to be seen from an angle of guardrails, and leaders need to be providing that rather than being gatekeepers. Like, I think, let’s not do this because it does not have an ROI, let’s only use X technology or Y. I think those are the statements that basically hamper the AI experimentation. Obviously, I’m a strong advocate of putting in guardrails in place and governance should be looked at from a very high depth, but at the same time, I think experimentation has its own place too.
Achyuta Ghosh — Executive Research Leader, HFS Research[07:10]
You’ve touched upon AI, Sameer, and we’ll move into that. Right, many enterprises have experimented with pilots for a long time. The bigger challenge is now basically scaling those success stories across the organization, because pilot to production is a very low ratio, often 10 to 15% in the most successful organizations itself. So what lessons have you learned about moving from isolated pilots to initiatives that deliver enterprise-wide impact? What really separates successful ones from failed ones? What organization barriers are actually emerging? How do you prioritize use cases? And feel free to share examples from your journey at XPO also.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[07:49]
Oh, certainly a question. The first and most important thing is lack of data readiness. Let’s start with that. When you do a pilot, you do it in a very isolated environment. When you start talking about volumes and the scale of operations, when you put that to an enterprise scale, you realize that, oh, your POC was not addressing that aspect. So the whole element of data readiness, the volume, the type of data that is spread out within an enterprise is tremendously huge, and pilots do not address that problem. Pilots are very focused on the happy path scenario of it.
So it starts with that. And I think this is something that we have faced. When we look at implementing an AI initiative or a process redesign using AI, we figured out that when we start working on projects, take an example of invoices, the automation of invoices that we have executed using AI, we figured out that invoices vary in their forms and shapes and the data they bring in. And we are not even talking about language and currencies and all those complications that are involved in just doing AI and OCR.
In other words, the second element that I feel actually helps the scale very easily is the governance of it. I think when you do a pilot, you don’t look at it as a wholesome initiative. You basically look at how it works in a very restricted environment. When you go into production, you have factors like volumes, you have factors like speed at which things should work, you have factors like performance, and all of that starts coming into play. And I think those conversations need to be done up front, especially to make sure that your pilot can scale up at the right time in the right way.
The third thing is waiting for perfection. And I think this is something that I experienced. Your accuracy level is 80%, and I think it needs to be 99.9%. And that’s, I think, going to drive your pilot away from being successful. I think this has to be a continuous process of develop, measure, redefine, and launch again. So I think that is probably one of the other elements that I feel takes the pilot back.
And in my experience, a couple of pilots that could not make the same impact as some of the other successful initiatives we did was that we start treating AI as a technology initiative. And I think that is one of the biggest, how I call it as a mindset, culture, and leadership barrier. You need to treat AI with a business outcome, the thought that it is going to make an impact on your process and business is basically how the initiative has to be led. So you need to have proper stakeholders being put up, or people who would own the governance, people who would own the initiatives, people who would own the outcomes. So from my perspective, I think these are some of the examples where AI pilots are forced into failing rather than putting a very deliberate effort into making themselves right.
Achyuta Ghosh — Executive Research Leader, HFS Research[11:43]
This is a good segue into our next question, Sameer. At HFS, we often talk about how we need to move from experimentation to industrialization as AI becomes more embedded in overall business processes. You touched upon governance, talent, operating model, stakeholder alignment. What changes do you think need to happen to make this work at scale? And how do you measure success? How do the different teams collaborate? What role does the GCC play in the orchestration across the enterprise? Please share some examples from your organization also and how we plan to go ahead.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[12:18]
Well, yes, let’s talk to GCC as an element. I think that the word is over and overused at this stage. Everyone is talking about GCC. But let’s treat for a moment that the centers out here are basically just extensions to the organization and operating on the same principle, same outcomes expected.
The first lesson that we learned is that collaboration is key to success out here with any of the AI initiatives. I think when we treated AI in silos, which is like, hey, let’s pull a tech team and that will basically come up with an AI project, and let’s go back to the business and tell them on what value we can deliver. I don’t think that is working. It needs to be thought through in a reverse order. Let’s figure out on what business impact, and business outcome, and customer outcome we are looking to make. That thought is very important. I think that has been key to our success because we started with that as an operating model, that business is as much involved into the process as technology is, and it was never a tech-led initiative. It is always a business-driven initiative.
The second most important thing I feel around making that industrialization bit more successful is your ability to experiment and evolve. I have one example where one of the AI initiatives that we had included a segment of OCR, and we went in live, and I think it all was working fine, had a good business outcome, saved a lot of time, and automation allowed us to work faster. And at a certain point, I think if you are not going to change or redesign your process, you’re going to basically keep automating things that you have been manually introducing into the process. So I think our lesson was really that let’s not think about a left-shift automation strategy. Let’s go out and redesign the processes as needed. And that’s where I feel AI is very impactful.
I think the third area that I feel all organizations, and especially on our side in the GCC where a lot of innovation and drive into the AI is coming in place, is our ability to make AI a part of a habit rather than treating it as a project. And I have an example. We are talking about, let’s say, the efficiency in terms of the testing process, the QA process, where a lot of code automation is happening and AI automation is helping us automate test scripts and whatnot. And we learned one thing: that if your process is not ready to adapt AI and embed it within itself, all you are going to do is a developer will write the code, deploy it to a certain environment, you’re going to look at it, write the test cases or get AI to write those test cases and test it, and then 15 days later you will come back and say, “Oh, something changed in the UI and now my test cases are failing.” So let my team go fix it, or let AI do something to fix it or write the next script easily. If it is not a part of your embedded system, if it is not a part of your process, it’s not going to succeed.
And we are talking about self-healing tests now based on our experience of how test case automation can be improvised over a period of time without having that human in the loop as needed for specific scenarios. Right? Not everything is going to be led or AI-driven. So I think that was our example where we thought that okay, until and unless we go and fix the process or change the process or evolve it, your AI is going to only play a role of doing a very small part.
Achyuta Ghosh — Executive Research Leader, HFS Research[16:42]
I see that as a challenge in most organizations, which are bolting on AI to existing processes. So you’re right that way.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[16:50]
True. And third is the measurability of it. Every time you start talking about an AI initiative, I think our experience is that someone got a bright idea in the shower, got it out into a technology team to implement. The team spent weeks and months and years to do the build the product. And then in the whole process, we forgot that in an attempt to look great on the presentation, we forgot the business outcome aspect of it. Right? And then that’s when you realize that yes, well, you spend millions of dollars in taking an AI tool and an initiative, and then the value it drives is almost zero.
And I have a very interesting example that came in from one of my friends who was talking about an HR process they had, driven by an AI thing, where the ATS process was automated to read a resume, do an initial screening with the candidate, figure out the interview process, make an offer, all automated, all AI-driven. And it was like all amazing, one click, your entire process can be done. Only to realize later that the organization was probably hiring 50 roles in a year. And that’s basically, you spend millions of dollars in building up a system, and you later realize that, oh well, the impact that it is going to make, or the outcome it is going to bring, is probably much less.
Achyuta Ghosh — Executive Research Leader, HFS Research[18:23]
So, last question to you, Sameer. A future outlook for your center, for the XPO GCC, what are you doing today versus what you think you will be doing in the next 3 years? And what capabilities are you investing in? How do you think the role of talent and the work you do will evolve, and how will success look like differently? Quick 30-second take.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[18:49]
So the first most important thing that we intend to do is create that AI fluency in our organization, right from leadership to our teams. They all need to be AI trained, future ready, AI trained, I would say. It’s not going to help if you only have your technology team trained. For us, success is going to be on how AI is embedded within our processes. So we don’t want to see that years later, to change something or automate something, we bring in an AI project. I think we see that AI is going to be part of our day-to-day function, right from a functional requirement emerging. We would like AI to embed in itself and start working on things that would basically drive a business outcome, a customer outcome. Our measuring criteria is not going to be on how much money we saved or how much money we generate using an AI tool, but what is the impact our AI solutioning is making on our business outcome and our customer focus.
Achyuta Ghosh — Executive Research Leader, HFS Research[19:52]
Sameer, thank you for joining us and sharing your perspectives. What I found very useful in this conversation is the emphasis on culture. AI scale will not come only from tools and pilots. It will come from leaders creating the right environment for experimentation, adoption, accountability, and business impact, as you pointed out. Thank you again for being a part of the HFS GCC Advantage.
Sameer Shaikh — Senior Director, Technology & FSS, XPO[20:14]
Absolutely. Pleasure having been here. Thank you.
