Bekir Atahan on AI Transformation at Experis

Speaker A: Welcome to the podcast.

Speaker B: As a listener, you're invited to claim

Speaker A: your comp pass for our next live full day Software Oasis virtual AI event with the world's top B2B AI experts. By attending, you'll join our engaged community of founders and executives from over 200,000 leading B2B global organizations. AI won't take your place, but, uh, a competitor leveraging it to drive results will learn the latest in AI. First Thursday of every month. Tap green button@softwareoasis.com or use show notes link to secure your past.

Speaker C: Uh, you know, Akira's journey from Turkey to leading the Data and AI center of Excellence at uh, Experis is fascinating. He started his career in the US with a keen interest in data and, and it led him through roles at companies like Sun Microsystems and Oracle. And then after founding his own analytics company, he brought that wealth of experience to X Baris. What I love about Bakir's story is how he emphasizes that AI isn't just about technology, it's about transforming how people work. By the end of uh, this episode, uh, you'll understand exactly why that's the key to successful AI implementation. I am your host, Michael Bernzweig, and this is B2B Automation Spotlight.

Speaker B: I know a lot of uh, different executives that are listening in today have had their own journey getting to where they are. But I know that your journey getting to where you are is uh, an interesting one and I was hoping we could share that uh, to, you know, your journey prior to joining Experis, to give the audience a little bit of, uh, a little bit of insight into uh, some of our conversation.

Speaker D: Sure. Um, thank you for the opportunity again, Michael. I'm originally from Turkey, uh, and these days I guess Turkey is the one that everybody's trying to pronounce it but they changed the name on it. So I came to United States when I was just turning 17, uh, right after high school, um, decided to pursue a career in computer engineering. And um, I ended up having an interest in data in particular. Right. How the databases are functioning, how to kind of code that. And I got my first opportunity with an insurance software company down in Columbia, South Carolina.

Speaker E: And everything started from there.

Speaker D: Right. Um, and then I moved to MCI and WorldCom days, MCI and WorldCom, um, before they actually went bankrupt, which is an interesting story too, uh, which then become. They were bought by Verizon too. I worked for MCI and WorldCom for about five years or so. From there I went to Sun Microsystems, which kind of have the uh, Father of Java and the Solaris world. I worked at them with them and then they became purchased by Oracle. Right. Which we know the story on that. I have a different journey on those things. Um then I decided to do something like I was very into um by nature I like to learn every day. Right. Uh, every day is a learning journey for me. And um, I decided to work for a company called surescripts which is the E prescription network company that they were building at the time. I worked for them quite uh a bit. And then from there um, I decided to do something with the healthcare stuff and I work for Inovalang, the company called Inovalang. And from there I decided to say you know what, I'd like to try something by myself. Can I have a company and can I really focus on a data analytics world? And I started my own company um, uh with a partner of mine and uh, together we um, we stayed on, worked on data and analytics for 12 years. From there um, uh again the patent that I got in there, uh, during the time we were really focused on uh, artificial intelligence as well as the analytics uh, and data. So I have done a lot of different aspects of that. Then experience opportunity showed up and I happily took that and that included my partner as well. So that's how we became uh, I uh, became the Vice president of the Data and AI center of Excellence at Experts.

Speaker B: I love it, I love it. And I remember those early days of sun uh Microsystems. They were quite, quite the force. And um, I actually uh, had uh, you know some experiences with their early you know the Spark and a lot of the other, other systems and they were, they were unbelievably complex in terms of configuration and uh, all of that. But they ran the backbone of a lot of uh Internet organizations for many many years. So they were for sure ahead of their times. And uh, really an interesting part of that journey.

Speaker D: It is.

Speaker B: But uh, I know that uh, all that's going on over xperus is uh really exciting and clearly the origins of machine learning prior to AI really led into a lot of where we're at today. But um, can you start out by sharing with the audience? I know you're um, heading up the Data AI center of Excellence at Experis.

Speaker A: Can you share a little bit about

Speaker B: that organization within the framework of where it fits into Experis?

Speaker D: Right. So traditionally, so we're really um, a uh, staffing and services oriented company. So we were inside the IT world and um, the data actually foundation of everything including artificial intelligence, machine learning, everything Is really relies on a set of data, right? Um, the cleaner the data you are, the better the outcomes you can get. The our focus was really on understanding because I came from the data background in general, right? This included multi dimensional modeling, statistical, um, uh, the models as well and um, everything basically moving data from point A to B for reporting and analytics purposes was always played a key role in our world for AI to be successful. We really want to make sure that um, the data is clean at any given point and it is centralized in some shape or form. Um, and when the focus is that. When I joined the experis first I saw that there is a big opportunity here with our clients and um, even internal uh, focus was to be able to build platforms that can serve the needs of AI, right? Because again it's not just the demoing and from the get go I was never thinking about demoing, right? How can we efficiently really embed AI into our workflows to get the proper outcomes? So um, the first step for that was to set up our AI laboratory, right? AI labs and then the data lab as well, so, so that we can really service our clients. Basically software agnostic, uh, platforms and that's how everything really started for us. Michael.

Speaker B: Now I know a lot of um, organizations as they're rolling out AI, uh, you know, trust and safety and security is a huge um, huge area that especially in the enterprise space needs to be taken into account. And I know that um, you know, from our conversation before the podcast that you had shared a lot of that with me, but can you share with the audience um, some of your thoughts on that?

Speaker D: Yeah. So first of all I want to say one thing, right? The cost of not using AI, uh is bigger than all the risk that you just listed out. To be honest, in a statistical world that we work in, the sheer power of AI uh right now as of today is beyond anything that we have seen before. And if we fear it, we will not achieve where we're trying to go and then technology will leave us behind. So if it is good enough from other things that we should definitely, I mean the AI is, is the way for the future and it's going to be for a long, long time. But again the AI doesn't, it's not a magic wand, right? There is an input and there is an output and there is a weighted averages that that is being played here. Uh, so that being said, what we are really seeing, and this is one of the reasons, is that how can we really effectively apply AI to where we're Trying to go and what outcomes that we are looking. So you can't really just okay, everybody go use AI, right? Uh, um, or the tools. But it needs to be very targeted what problem that you are trying to solve or what efficiencies that you're trying to improve or what is the uh, I guess cost savings that you want to do. And then you look at your problems that your organization may be facing. How can I improve the efficiency, productivity and also the velocity of the things that we are doing maybe comes to question. And as I mentioned on our presentation as well and uh, on our discussions, it is really important to think about the governance piece as well. AI is extremely confident when it's wrong. I mean that is the reality of it. When you ask for something that is just strong, but it will take you for like it will take you for, hey, this. So we need to make sure that what we are really applying is governed and guardrailed very well. Test driven development is something that we really go by and live by. So um, if AI without AI, we cannot get to where we are going. We build certain applications under five to six weeks, but with the tight guard data and governance, you need to be able to be in front of it. And this is where AI is not doing it by itself. You have to have human in the loop, right? Actually human at the end kind of a way too right in the loop and at the end from the development perspective, you can really combine AI into the full flow of the development cycles. But your senior developers, your mid level developers, even your junior developers, they have to look and verify your pull requests. Like how is the code going to get promoted? Um, is the code is doing what it's supposed to do. It does a lot of self healing as well. So every outcome with AI, with the human together, right, it is the faster and the better outcomes. We just need to apply it without the fear and with the real use cases. That is the goal. And I think that everybody should be thinking that way. The days of building teams of like having 10, 15, 20 people, right to try to get to where you're going and six to uh, nine month projects are kind of shrinking, right? Those days are gone. Meaning that you really need to, you can get to the ROI that you're looking for or the outcomes that you're looking for very quickly and you can build the governance in the same time and you can also improve things as you're building or uh, as you're building and you're trying to get to that outcome.

Speaker B: Now I know you've un and the entire team have worked with so many organizations over the, over the years and most recently helping uh, organizations really find practical business use cases and outcomes for what they're trying to achieve.

Speaker A: How do you help them put guardrails

Speaker B: on this to really find those use cases that are going to provide ROI while still maintaining the structure that organizations need in this day and age?

Speaker D: Yeah, there are different, there are different governance tools in place, right. And observability becomes really important as well. So we work with the client based on the toolsets they have. If they don't have any tools that we have preferred tools, we have preferred tools that we can apply. But most importantly, uh, the context management is where the main guardrails and the governance really lies right now. You will always have to capture the inputs and outputs like I mentioned, at the end of the day it's the function, right? There is an ask, there is an output and how do you control that? How do you maintain the guardrails around this? Right? That this is not hallucinating or it's not giving you or it's not making things up and it is staying within the realm and it's returning to the right output that you're looking for. It may not be the same exact output you're looking for, but how is it similar to it? Again it depends on what part of the business problems that you're trying to solve. But there are very different techniques that you can apply uh, onto the governance itself. But the tools are really important to us. I mean in one case is then uh, I do want to, I don't, I don't shy away from this. But Boston Access governance is great. We are using observability tools with it as well. And so that we know that if there is a drifting or there's a different, different answers are coming through. We know better to strengthen the prompts or um, rerun the models and develop the models based on the outputs that it's producing. But the key to the matter is you need to capture inputs and outputs, uh, even if does come with a specific cost of maybe a speed. Instead of getting an output in two seconds now you're going to get it in three seconds, right? Um, but it is required that you capture the inputs and output so that you can take a necessary action accordingly.

Speaker B: I want to shift the conversation a little bit, um, to an area that I know touches on something that just about every organization listening to the call is working through and that's in the HR and recruiting and career space within Their organizations and bringing aboard and screening candidates. Um, there are um, a lot of off the shelf AI solutions. But I know something that's very special on the Experis end is your coding platform.

Speaker A: Can you talk a little bit about that?

Speaker B: And some of the use cases and some of the actual implementation of um, the model that Experis has pulled together.

Speaker D: Sure. Our idea was behind it is that hey, how do we build a platform actually will help our teams. Right. Teams is a cohesive way.

Speaker F: Right.

Speaker D: Cloud. I mean think about it. We have a lot of different coding platforms available today too. But everything is really individually based. Right. Individual targeted ways. Now cloud code is changing the way a little bit. But um, cursor is that way a lot. Low is like that. Right. Um, we thought that we sit down internally and we said that hey, we need our own CLI that we can control and then we can have our own MD files that we can put our guardrails, governance and guidelines in place so that everybody, whether I'm coding it, if you're Michael, if you're part of my team, for example, you can use it as the same way and then it will look, the code will look like very similar to whether. But he wrote it or Michael just produced it. Right. Because it has the same standards and coding. This is the idea, what the whole idea was. How can we have a tool that we get to control the inputs and outputs of it and we can enhance it at the same time apply our guard lays and governance while producing custom applications for our clients. So that was the whole set. And how can we improve our code quality and test driven development approach? That's what it got us to the point that where we need to build our own command line interface. Once we build that then we said we looked at it internally as well. How can we apply AI in such a way that it will improve our internal use cases and provide value uh, to our clients at the same time? Because there are inefficiencies that we are seeing in certain processes that we have as a business. And we applied this to the. And we built a software Recruiting is one of the places that very, as you were mentioning. Right. Um, that we wanted to increase um, um, basically effectiveness of IT and how can we get to the best talent with the least amount of time. And what this provided Sophie code with that command line provided for us is that we deployed our three to four engineering team members in IT utilizing that. And then we produced a platform where we can utilize the AI for our initial screenings with AI because again if you think about that. An AI can interview up up to infinitive amount of you know, um, um, candidates within the hour. So but if you have a recruiter who needs to do the initial screenings, they have a cap, which is a linear process. So we wanted to make that an exponential because that's the business that we are in today. And the whole idea behind is rate and rank these um, uh, candidates so that we can and our recruiters in the morning when they come in, they can focus on the top five or top 10, whatever that may be. So it increased efficiency in tenfold. And because again we wanted to, we want to get the best candidate that is available outside looking for, looking in and then we want to get to it utilizing the AI. Obviously we have resume matchings and then we have also resume rankings but we combine m that also with interviews. Our recruiters are also using the platform as they are interviewing the candidate live and our artificial intelligence is helping the recruiters as well. This also gives a capability as such that the recruiters, they don't have to be versatile on every technological stack either.

Speaker B: So yeah, it sounds like you're really using the AI as a tool for the existing recruiters to magnify how efficient and how. So it's really helping an organization make the most of the assets that they're paying for.

Speaker D: Absolutely right. And the quality has increased tenfold as well for us.

Speaker B: That's absolutely amazing. And I think you must have so many organizations coming to you day in and day out asking for a lot of very similar types of solutions. So having the sofa code that you mentioned really allows you to take that area of subject matter expertise and in your own day to day organization extend it to many, many clients.

Speaker A: Um, so can you talk a little

Speaker B: bit more from the experis standpoint? The types of um, organizations and industries that are relying on experis for solutions.

Speaker D: Uh, energy. Yeah, that's fine. I can definitely use. So we are in the different sectors. Right. Our really strengths are in the financial sector so we are really focusing on the industry retail space as well quite a bit. Right. Uh, also we are also helping our technical giants as well. Right. Um, we have clients for Microsoft AWS and as well as the Microsoft and Meta for example. These are the kind of the clients and they always ask us how are you helping us improve our workforce utilizing AI and Sophie code. Plus the SoFi recruiter is our answer to those clients. So pretty much we are involved in everything.

Speaker A: Yeah.

Speaker B: And it's amazing you're not only helping with the staffing and all of that of IT and tech. But you're also utilizing the tools internally and helping clients with these same tools.

Speaker D: Yes, that's exactly right. And that's the whole purpose of the whole platform as well. Uh, increase our initial internal efficiencies utilizing AI and then showcase that to the clients that who are asking for it as well. Because we're not going to be able to scale without the help of AI. And I believe that every single company out there right now trying to embed AI into it.

Speaker E: Right.

Speaker D: I have the same focus on goal now.

Speaker B: We uh, always like to leave um, organizations with some, and especially a lot of the listeners with some actionable uh, takeaways. But as far as you know from where you sit, are there certain types of enterprises that see maybe an um, unfair advantage in terms of rolling out some of these capabilities that you have or are some of the organizations seeing just a way higher level of ROI or results from rolling out some of these expera solutions than others?

Speaker D: Um, can you clarify that a little bit more for me on the question?

Speaker B: So you're working with so many different kinds of organizations and while it's a wide range of solutions that you're helping them with, are there certain applications of the AI and, and the tech and the experis solution that seems to really help organizations off the charts compared to others. I know a lot of organizations in the early days of AI said hey, we can write all kinds of content. And another um, organization said hey, we can analyze all kinds of data but clearly AI provides amazing results in certain areas. Are you seeing something like that with, with the clients you're working with?

Speaker D: Yeah, yes we are. The biggest thing that we are seeing with the clients is a few of our clients. Right. Are um, the capability, first of all they wanna really prove is the capability there and then is the value there. Right. But some of the clients are really focused on the outcome. However they don't have the infrastructure to support it or the data is not ready. Right. So when they see the solutions that we provide because we kind of really augment ourselves and uh, prepare the data in such a way shape that the outcome is really right there. Um, different clients are seeing a huge potential on the way that they are doing a lot of stuff with manual, for example, taking pictures of specific tags or anything like that and they will basically somebody's looking at these pictures and typing the stuff in manual. So those are like a lot of manual tasks are being applied which are error prune and when we show up and then we show them the art of possible. They are extremely wild and they are also seeing the similar, um, efficiency gains as well as understanding that basically the evolution of the AI requires a platform, um, knowledge, understanding how do you go from point A to point Z, understanding how the data has to be flowed through properly as well. So there's a lot of opportunities that we are seeing with the clients that we can help them get to where they're trying to go to.

Speaker B: So a lot of times it sounds like having the right foundation in place before undertaking, uh, an AI project.

Speaker D: That is correct. That is correct. They are asking us, hey, can you help us stand up? Because again, it is an ocean, right? Whether it's an on prem or cloud, the setting up the environment properly is the case.

Speaker B: So. And I guess, you know, maybe as a final thought, which I think will be helpful to a lot of organizations, M. You know, it's uh, no big secret that so many organizations took a first stab at AI over 2024, 2025, and as we're heading into 2026 and beyond, a lot of organizations are saying, I need to find some success in this space. Are there certain things that you've seen, clients that have succeeded with AI, which is a sub segment of those that have tried, uh, to roll out AI, that they're doing differently or that you're helping them along the journey that um, you might care to share?

Speaker D: Yeah. So, uh, we are seeing certain things especially on the help desk, um, stuff this includes healthcare, uh, hospitals and then also the IT help desk stuff that we are seeing that they're implementing these changes where the AI is, especially in the voice AI piece. Like one of the partnerships that we have done is the soundhound. And um, that is the one that I'm really proud to say that they really improve. The AI is improving the interaction, um, with the patients and their needs. Right. Um, the patient can call in, schedule doctor appointments and it's not just a traditional press one, press two kind of thing. You're just having a conversation just like you and I are having in an ejectic framework. And that's being implemented by us as well, help of our partner as well. Right. But, but that is really life changing. Uh, things are happening. Right. Anything that is task oriented is going to be done in the future by AI. But this particular piece is dear to my heart as well because the platform is allowing you to chat and very accurately getting you where you're trying to go to. Right. And improving and having the other, you know, healthcare, uh, instead of somebody's answering the phone, which the human is in there. They're really focused on the better part of the healthcare pieces.

Speaker B: I love it. Well, I'll finish with this. Is there something in the AI area that uh, you're really personally excited about as we're looking towards the future of where things are heading?

Speaker D: Yeah, I am. Uh, again everybody, right? This includes today, even the people who are recent graduates and everything like that. Look for a problem, right? And you can solve it. It's not going to take months, it's not going to take years. Right now we are in an era where we can really solve problems and build applications in a light speed with the accuracy and um, the efficiency. So that um, I'm really, really excited about, right? Build something, show it right, and come help the legacy systems to become the future systems. That is what really excites me. The speed, the sheer quality. Um, but don't talk about it. Just find the real thing and don't build something. Build something meaningful is right now everybody's reach. And that really excites me.

Speaker B: Well, on that note, I think we've uh, dived into a lot of different areas and I think we've absolutely uh, given a, uh, lot to ponder. To those that are listening to this segment and from my end I'm very much looking forward to your presentation at the live event.

Speaker D: Thank you very much. Thank you for having me Michael and have a great day.

Speaker B: Thank you.

Speaker D: Thank you.

Speaker C: Bakir just shared how his diverse career journey led him to focus on the intersection of data and AI at xperis. Now he's going to walk us through how Experis is using AI not just as a tool, but as a driver of operational transformation. It's a crucial piece of understanding how AI can truly impact business outcomes.

Speaker E: Stated. I'm, um, the VP of the Data

Speaker D: and AI center of Excellence at Express Services.

Speaker E: I have a computer engineering degree from University of South Carolina and I have spent more than 25 years across data analytics, enterprise technology. Along the way I picked up a patent in virtual assistants using state based artificial intelligence.

Speaker D: So this space, the AI space, is not really new to me.

Speaker E: I have been thinking about how machines and humans work together for a long time. Today my focus is helping our teams and our clients use data and AI in practical ways. Ways that actually improve delivery, operations and business outcomes. And that is really what I want to talk about. For us, this hasn't been about one tool or one pilot. It's been about learning where AI actually fits into delivery, engineering, recruiting and sales. The value was never in experimenting with models. The value came when we started applying. When we started applying the AI, uh, to real work things like our AI workbench for secure enterprise AI use, Sophie Code for application development acceleration, and Sophie Recruiter for recruiting efficiency. The phrase human plus AI model is deliberate. We don't see AI as replacing people. We see it as changing how people work, how teams are structured, and how outcomes get delivered. Here is where I will take you why this is an operating model shift. The five stages of our journey where AI is reshaping our business today and the principles that separate hype from outcomes. For us, AI stopped being just a capability discussion and started becoming an operating model discussion. Let me start with why I keep calling this an operating model shift. AI transformation is just as much about people skills and operating models as it is about platforms. The real shift happens when workflows change, not just when tools get installed. Our president, Kai Mitchell put it in a perfect code. Basically, AI transformation is just as much about people skills and operating models. It is about platforms. One of the most important things we learned early, AI transformation is not a platform decision. You can stand up models and buy licenses, but if the team doesn't know how to work differently, you don't get much of value. AI, uh, workbench is a good example. We didn't build as a chatbot. We build it as a secure enterprise environment where teams can use LLMs, generate content, work with internal knowledge, and do it in a governed way that change how we knowledge work happens inside our teams. We built Sophy code. Sophy code is not just a coding assistant. It brings in our own standards, our own patterns, and different agents for different parts of the development lifecycle. The changes of how developers work end to end. These aren't tool rollouts. They are workflows, chain changes, role changes, and velocity changes. Sophie code brought us a velocity we never imagined. AI transformation sounds like a technology topic, but it becomes an operating model topic very quickly. So how did we. How did that operating model shift actually,

Speaker C: uh, Beaker, I'd love to hear how you approached the initial stages of your AI journey at xperis. Can you walk us through what that

Speaker E: looked like unfolded for us? It came in five different stages. Our AI journey involved in these five distinct stages. And it didn't happen in a one big leap. It evolved in these stages and each one built on the last. Let's look at these stages. Staging 1. It was the placing our looking for our artificial intelligence and machine learning talent. Early market demand was almost entirely about talent. Clients Needed data scientists, machine learning engineers, AI architects, the people who could actually do the work. That's where our actual journey started. Helping clients stand up to human capacity before anything else could have happened. From there, we looked at it and go, uh, you know what? We need to build our own AI solutions. We realized placing talent wasn't enough. We did our own internal proof points. Real experience, not theory. Clients always ask, how are you using AI in your own organization? And that's where AI, uh, Workbench and Sophie Code came in. We wanted to build and use AI ourselves, not just advise clients from outside. You cannot gain credibility or guide someone through a transformation if you haven't done it to yourself. And then we looked at orchestration of AI and humans together. Just like I mentioned, the next phase was the orchestration. How do humans and AI actually work together in practice, not in theory? Uh, and then we build Sophie Recruiter, which is a great example for us. AI handles the screening, scoring and front end efficiency. The recruiter stays focused on judgment, relationships and closing. Then orchestration done right, AI takes the repetitive lift. Humans take the decisions that need humans. Solving industry challenges with AI based on our clients ask came next. Because we had the foundation, and once we had that foundation talent solutions and orchestration model, we could apply it to the industry's specific challenges, the problem our clients actually care about in their language and their context. Generic AI doesn't win. Applied AI does. In our stage number five, we are growing through our partnerships. Once we solve the industry challenges and

Speaker D: once we can determine that we can

Speaker E: solve the industry challenges with AI, you really need to look for the partners who can really help us grow. No one solves this at a scale alone. You need the right ecosystem platform, partners, model providers, industry specialists to take these solutions into the real world. We have a solid relationship with our Soundhound team with the company called Soundhound and also with IBM.

Speaker D: What's the next platform?

Speaker E: Our journey moved from placing AI talent to building our own solutions like AI, Workbench and Sophie Code to figuring out, uh, how people and AI actually work together in the real world with the right partnerships and right solutions. That journey led us to three places inside our own business where AI now shows up every day. Enhanced rcoe delivery, sales, execution and recruiting efficiency. This is where the journey becomes tangible. AI is reshaping three core functions, like I mentioned, our delivery, our, uh, sales, our recruiting with concrete products and real outcomes behind each. This is where AI moves.

Speaker C: Can you tell me a bit about how AI is impacting delivery costs and the development Process at Xperis from strategy

Speaker E: to day to day execution. 3 Rs. Again, outstanding for us is that delivery, sales and recruiting. The first thing is the SoFi code and legacy modernization. On the delivery side, Sophy code is the clearest example. We are seeing up to 50% reduction in delivery cost of a solution and that's because it accelerates development, testing, consistency and quality of the code that has been being developed.

Speaker D: The key thing with the Sophie code

Speaker E: that we decided to focus and build, it brings our own coding standards. It is focused on our development team rather than a development individual. Our preferred models, we get to choose and select and control and different agents depending on the task. Whether you are trying to build or whether you're trying to plan, you get

Speaker D: to choose which agent you like to use.

Speaker E: So it's not just autocomplete, it is a jantic system that reflects how we actually want to work, get the work done using Sophie code. We wanted to also utilize it for legacy modernization purposes. In a world like where the SaaS conversions can happen to the to R. The kind of work where AI shines reading legacy code which was developed in another platform like SAS and then um, converting that into R, the legacy code, understanding patterns, helping convert logic and speeding up that modernization that would otherwise be manual and slow. So delivery isn't just about writing new code faster. It's all about transforming older environments and faster with efficiency and quality. Then we looked at it and we go like okay, how can we use AI in our sales agentic engines for proposals and prospecting? On the sales side we build secure intuitive agentic engines for prospecting and proposal automation. A uh concrete example is our proposal support. As you may imagine that our company does get a lot of RFIs and RFP requests. So we will have to build proposals day in and day out. AI helps create first drafts, organize inputs and shorten the time that it takes to respond with quality. This is a huge unlock for our sales and proposal team. Another example is the thoughtful outreach. How can our sales personnel use starting instead of doing everything from scratch every time, how can instead of starting from scratch every time the seller start from the AI prepaid context? The point isn't to remove the seller, it is to remove the repetitive setup work so the seller can focus on the relationship and the strategy. That brings us to uh, our. One of our biggest winnings that we always achieved is the recruiting. Where we built SoFireCruiter using SoFi code in recruiting, SoFi Recruiter is the best example. The slide references, AI video interviews as you can see candidate scorecards and uh, recommendations. That is exactly where the value lives. Recruiting is often very manual on the front end. A lot of recruiter time gets burned on screening software. Recruiter scales that the screening and ranking. So the recruiters spend their time on the highest priority candidates and then close on them faster. The common thread isn't the replace the human it. How do we make the human more effective and efficient? That's where the operating model shift becomes real. AI showing up in core business functions and not just in innovation labs. Across delivery, sales and recruiting, the pattern is the same.

Speaker C: What principles do you think are key for companies to move beyond just AI demos to achieving real outcomes?

Speaker E: AI handles more of the heavy lifting, uh, up front so uh, our people can focus on their high value part of work. So what actually separates the companies getting outcomes from the ones who are still stuck in demos? A handful of principles is the answer. And what are those principles? If I had to boil down what we have learned is capturing these four principles and each of each one of these ties back to something I have already mentioned. Principle number one, begin with the end in mind. Experimental AI has to begin with the end in mind, tie it to the real outcome. Revenue, growth, efficiency, customer experience or risk of reduction. Every one of our solutions has a clear outcome. Sophie, Recruiter for instance, faster screening, better recruiter focus, getting to the best talent in the least amount of time. So if you come on faster delivery, lower development cost and higher code quality. SAS to our conversion modernization on us. Uh, again with the speed and then reduced manual effort. AI workbench safe govern enterprise use of AI. If you cannot name an outcome in one sentence, it's going to stay as a demo. AI is a core skill, not a specialty. AI fluency cannot live in one innovation team forever. Engineers need to use it, recruiters need to use it, sales teams need to use it. Leaders need to understand where it fits. Fluency doesn't mean everyone becomes a data scientist or AI architect. It means everyone understands how AI improves their function. Our principle number three is embed AI into engineering and operations. The real return shows up when AI is inside the workflow, not alongside it. Look at where ours are embedded. Sophie. Code is embedded into engineering. AI workbench is embedded into the knowledge work and enterprise workflows. So if your recruiter is embedded into recruiting operations and measure the right things, the delivery velocity, quality, repeatability and enterprise scale, those numbers prove AI ah, is actually embedded. And our last but not the least principle is a secure, governed and responsible. This one is Non negotiable. A real risk is data exposure, hallucinations, governance gaps and model abuse. One reason AI work match matters is because it gives the teams, our teams, a governed place to use AI without exposing data to public tools. If you don't solve for governance, early adoption gets blocked later or you have to rebuild. Responsible AI is not a separate conversation. It has to be built from the start, like we have done to bring this back to where we have started. The human plus AI operating model isn't a slogan. It is the difference between the AI theater and unreal transformation. Real transformation changes skills, workflows, measurement and governance all at once. That's the shift.

Speaker B: Thank you, Pakira.

Speaker C: That was fantastic.

Speaker B: And I know everyone that's listening into this live has, uh, questions that they,

Speaker A: they will have for you.

Speaker B: So if you're listening live and you want to ask a question to Picare, just type that into the live Q and A box that you see in front of you during the panel. Uh, coming up in just a few minutes, Be care will be on that

Speaker A: panel answering questions from the live audience. And we will be there in just a moment.

Speaker C: Uh, Bakir's presentation really highlighted how AI is reshaping experis business operations. This is the live panel from our software Oasis basis AI Summit where real audience questions delve into these transformations. It's all about understanding how these insights apply to your own business challenges. So, uh, what Beer's presentation made clear is that AI transformation is as much about changing workflows and roles as it is about the technology itself. I think a lot of listeners are now wondering how these principles apply to their own organizations and how to overcome specific challenges in AI implementation. And that's exactly what this audience brought to the table.

Speaker A: Welcome to our final Q and A session. So on this panel we actually have, uh, all of the speakers that you've heard over this past couple of hours. We have, uh, Ryan McElroy, who is the vice President of technology over at Hylane. Welcome Ryan. Uh, we have Pritash Lal, who's the digital and AI, uh, lead over at Baringa. Welcome Pritash. And finally we have Jeremy McLeod, who's the founder of the Search Experience. And I know we have uh, quite a few questions that came in from all of the, uh, attendees today. So I want to, uh, do my best to get through as many of these as we can so we can get all of your questions answered. So the first questions, uh, came in for Ryan, so I'll start there. Uh, Ryan, um, this first question that Came in is from Jacob, who, who is in Raleigh, North Carolina. So Jacob is asking, as you help enterprises build data reliability practices, how do you prioritize which pipelines to harden first so you maximize AI readiness without boiling the ocean across every system at once?

Speaker G: It's really important to not try and do this sort of work across a lot of different pipelines. So, um, systems that support AI are obviously a very high priority right now. So whenever we're working with leaders and subject matter experts to figure out what to tackle next, if something happens to flow there, then it has a higher priority. The other thing that we really want to look at is actual, uh, business value. If there is a pipeline that supports something that we, uh, consider a data product, for instance, that really makes it, um, you know, something that we want to focus on. And then last but not least, there's that feasibility question. Um, we can add data reliability to legacy technologies, but there are often hard stops at certain points. We were working with a healthcare client recently and they mentioned the, uh, ETL technology they were using, and there was just no way to add the sort of observability we wanted. So, um, we'll be prioritizing other data ecosystems for them.

Speaker H: I can't hear Michael. Michael, I think you might be muted.

Speaker B: Yep.

Speaker A: Uh, the next question came in for Jeremy, and this is from Olivia, who is just outside of Chicago, Illinois. And Olivia is asking, when you say most sales hiring still over indexes on storytelling, what is the single most revealing tests you use to see whether a candidate can actually sell in your specific environment?

Speaker H: It's a very good question. Um, I think the number one way we test in terms of sales ability, it's really teaching coachability, I think, is the most important thing when trying to figure out if somebody can sell in your space. And the way that we usually run that and the way that we recommend clients run it is you get somebody to do a presentation and they can present on their existing technology or they can present on their previous technology, uh, and, uh, or they can present on your technology, it doesn't really matter. You want them to present on something that they're reasonably confident on. And then you give them coaching and you give them some advice and say, hey, I want you to think about this in terms of selling to this particular audience or selling to this particular stock stakeholder. And how would you change that? And here's some things that I think you could do differently in the way that you approach this. Then you run it again and you just see how they change things. Because the Number one thing you want to figure out is am I able to impact the way in which this person sells? And the only way to test that is to coach them in real time and see if they actually implement your feedback into the next presentation that they do. That will tell you more than anything. Because the thing I always tell people about doing a presentation is you could have somebody do the best presentation they've ever done or the worst presentation that they've ever done and you have no context for that. So you want to test on if it was the worst presentation they've ever done because interviews are really nerve wracking and people have stress and all the rest of it. If I make a couple of tweaks to them, do they get better? And if it's the best presentation they've ever done done and you make a couple of tweaks, can you make them an even better seller? So that's usually the way we try and test for it because I think it's, it's very difficult to, to get them to, to sell something and test in real time. But I think the coachability piece will tell you everything because then you can, you can learn a product uh, if somebody is coachable.

Speaker B: Yeah.

Speaker A: And I think that's relevant for just about every organization on the event because

Speaker B: sales is uh, clearly uh, something that

Speaker A: every organization touches on. So uh, the next question came in for Priytush. Um, the question comes to us from Madison who is right in New York City. And Madison is asking when you help enterprises de risk AI transformation, what is the first conversation you have with the C suite to align on business outcomes before anyone talks about models or tools?

Speaker F: That's an easy one. Um, Michael, that it's about adoption. Right. Risking de, risking AI return on AI investment is all about how much the AI machine will be used. Right. Um, we've seen this time and time again where the business case comes in where um, there's everything in the kitchen sink is promised in terms of value. And reality is once it's, once it's designed, implemented, built, ah, the teams end up using a fraction of the capability. So um, that's the hardest hurdle to get over.

Speaker A: Makes good sense.

Speaker G: I was going to mention Michael on that. We have enterprise client who's rolled out uh, AI tooling for their developers and they're very tech forward, very excited about it. But just last week they were mentioning that what they projected in terms of token usage and everything like that is 10% of uh, what they were targeting. They're six months in at least. And so even for people who are very into it, um, it's, it's definitely a challenge.

Speaker A: Yeah, I mean it's, it's so new for everyone.

Speaker B: You know, this is not, uh, not

Speaker A: something that anybody can say they've been doing for years. Although, you know, a lot of this, um, has its origins in machine learning. And a lot of the fundamentals to these rollouts are very similar to traditional, uh, rollouts, but the tech is absolutely different. Um, Ryan, uh, the next question came in for you and it's from, let's see here. This is from Megan, who's in Denver, Colorado. And Megan is asking, in organizations where data ownership is fragmented, what governance models or moves have you seen actually change behavior? So teams treat data quality as a shared ongoing responsibility.

Speaker G: Uh, there's something to be said, you know, about even if data quality is a shared responsibility, having a specific name attached to a data product or an ecosystem, for instance, even, or a collection of names, even is really important. And I think that that's foundational. Uh, that's what you start with. The other thing that we've seen, I do want to tie it back to data reliability here, but there's a lot of different actors in a data ecosystem. There are vendors that are sending you data, you have data engineers managing how it moves throughout your organization, and then consumers. If everyone has a very clear zone of control and you have, for instance, notification systems that let you know when things go wrong, and data owners who know who to contact when they encounter an issue, that really helps. It lets you build a no blame or a know where the problems are to where you don't have to spend time figuring out who did what or assigning a lot of heavy blame and doing a bunch of root cause analysis exercises that take up um, a lot of extra cycles. So when you have that clean, simplified, uh, explicit zone of control setup, uh, it really helps it be a team problem overall.

Speaker A: And I can see the, the map

Speaker B: lighting up in Canada. We have so many people joining us

Speaker A: from, uh, just over the border in Canada. So I want to get to a few questions from there.

Speaker F: Um, yeah, Michael, just want on that sort of data governance. Uh, yeah, you know, it's um, it's absolutely critical, like Ryan says, to have a name or a bunch of names who are ultimately responsible. But at the end of the day, the usage, the consumption patterns for that data should be what defined governance. You're ultimately as the person or team or function using those data elements responsible for creating value for your business. Uh, and if you can't do that because of XYZ reason. It's sort of on you.

Speaker D: Right.

Speaker F: So the governance model needs to be pliable enough where you can extend governance for certain data elements, uh, and have, you know, fragmented, segregated, whatever different model you want. But at the end of the day, consumption should drive ownership.

Speaker B: Yeah, and I think that's a very

Speaker A: important, um, detail for everyone that's on the event today. I mean, I think that really covers an important, uh, concept. Um, the next question came in for Jeremy, um, and as I had mentioned from Canada, this is from Sophia, who is right in Toronto. And, uh, Sophia is wondering, for companies expanding into new regions, how should they adapt their sales hiring playbook so they don't just clone their home market profile and hope it works everywhere?

Speaker H: Yeah, another really good question. Um, I think the biggest piece is understanding how different your customers buy cycles are in those regions. And once you understand whether they are different or not, because in a lot of cases people actually buy in a very similar way regardless of geography. So you would sell a product in a very similar way in Australia as you would in the UK and in the us. So I think you want to understand if there is genuine difference in the way in which your customers engage in any kind of sales process and understand whether you need to have that local knowledge to then move the sales cycle forward and uh, ideally be more successful. But the best way to test that is to sell with your existing team that you have and understand where things are falling down. So have come, you know, the more conversations you have with customers, the more informed you become, the more informed your hiring decisions then are around building that team and what skill sets you need to actually be successful in a market. I think people make a lot of assumptions and go and hire and think if I go into this market and I just go hire the people that were working for my competitor previously, then they obviously know this market and they're going to be immediately successful because it's a space they understand. And you don't even know if your customers like the competitor. They might think the competitor is terrible and you've just brought the problem in house into your organization. And they want something completely different and they want to be treated in a completely different way. So I think until you understand that, you can't really shape the profile of the people that you're going to go and recruit. And so it's all about that kind of preemptive knowledge piece to then shape. Okay, if we were going to go and hire some people to look after this region, what is it that our ah, customers really want. And then you can go and target those profiles specifically target the background that you're looking for, target locations. Because sometimes it's proximity, sometimes you can actually cover pretty wide geographies from central locations as long as people are uh, willing to travel. But I think I would always say to people, understand what you want and what your customers want before you make any decisions because otherwise you're just going to be unwinding those and you'll be figuring it out and real time that the highs you've made aren't actually going to make the difference that you really want them to.

Speaker A: This next question comes to us, uh, for pray to, uh, this is from uh, Anika, who's just outside of London in the uk And Anika is asking in global organizations, what have you found most effective for aligning regional leaders around a common AI transformation vision while still giving them room to adapt to their local realities?

Speaker F: That's a good question. And you know, there's it the answer lies. It's actually a cultural answer more than a technical answer.

Speaker D: Right.

Speaker F: Um, different regions react differently, different cultures react differently to different types of leadership. Um, at the end of the day, as a global organization, you do need to have some alignment, some standardization on what value is and what the definition of value is. And using that then you drive um, the prioritization of initiatives in local market. Because if I'm defining value differently than my region over the border, um, I'm going to do different things to maximize my value. The common understanding here is assumption here is that everybody wants to maximize value, uh, and there's no bad actors and all that fun stuff. But uh, what we want to get to is how do we use AI to maximize the return on our investment and maximize uh, how much value it's creating for the organization. The realities I'm coming to question that Jeremy was just answering, right. The realities of go to market vary across each region dramatically. Right. Buying factors, uh, your value chain, your supply chain, your talent pool, all these things are different. And all these things should be predicated and factored into how much, how quickly and what the things are that you're doing to drive value.

Speaker A: So uh, the next question came in to us for Ryan and um, this question comes to us from Lucas, who's in Munich, Germany. And Lucas is asking for global firms operating under strict regulatory regimes. How do you design data reliability programs that both satisfy auditors and give AI teams the flexibility to experiment quickly?

Speaker G: Awesome question. Uh, there are two things that I think, um, two specifics that I would call out. The first is when you look at making sure that data movement is generally secure and complete. Client One thing that we've done multiple times is basically ensure that data is if there's sensitive data, pii, um, phi, whatever it may be that you know we're looking at those regex values and seeing and tracking when it shows up.

Speaker H: Right.

Speaker G: And then if some version of that needs to make it to another system, is it tokenized correctly and is it masked? There's a lot of great tools out there that can be integrated into your ETL process like um, uh, Perforce's Delphix and everything like that that are great. But making sure that it actually happens is a key part of data reliability. Then for that secondary part, by doing ah, great tokenization and masking, you can sometimes still let AI systems perform to a reasonable degree. Um, you can give them uh, access to the non tokenized and non mass data if it's necessary. But um, what we're starting to see a lot of clients do recently is push synthetic data further. So that's separate from data reliability completely. But um, for instance today Ally bank got an award for generating synthetic customers. And that's what they're using to test a lot of their new products. They generated it, they have like six different profiles or something like that and they were able to do so because their customer360data product was so rich. And so if you, if you do that well then uh, being able to generate safe data, uh, it becomes possible.

Speaker B: Fantastic.

Speaker A: Um, Jeremy, two great questions that came

Speaker B: in for you, but I'm going to

Speaker A: get to this first one here. Um, this is from Ryan who's uh, just outside of Atlanta, Georgia. And Ryan is asking when you assess a salesperson that keeps missing targets, what are the most common gaps you uncover and how they were either hired, ramped or tested before being put in front of customers?

Speaker H: Um, I guess there's two ways to look at this. So um, if we're looking at from a recruitment point of view where you're looking at somebody that's applied to a role or has been um, submitted to you that has missed quota. The first call out, I would say, which is a massive green flag, is if this person has admitted that they have missed quota. That already is a win for me because so many salespeople miss quota and they don't want to talk about it. They will fluff over the fact that quotas have been missed. And I actually think the best salespeople learn the most when they haven't hit quota. And so what you really want to dig into is their understanding of why they missed quota. What are the factors that have contributed to their miss. And then, uh, as I said in my presentation, it's all about relativity compared to their peers. So you want to understand if they, even if they miss quota and they are at 80% and they've been at 80% for the last three years, where did they stack rank against their peers in their organization? It can be sometimes that they're just selling a rubbish product. And I know that's a funny concept to think of because everyone likes to think that their technology is the best thing that ever existed. There are some rubbish products out there that are really hard to sell. And if this person has been selling against a really large successful incumbent and they've still been landing deals and, and you give them a great product, they actually are more than likely going to be very successful with your product. And it's for a couple of reasons. They already know how, like resiliency and how to battle through things. So when they come up against somebody saying no to them, that's not going to shock them and throw them off their game. But it also just means that they have to be more creative as a salesperson. They can't just rely on a big brand to go and sell something. They've had to really understand their customer. They've had to work far harder to close every single deal that they did close. And I think you can dig into all of those things in an interview with people where you want to understand, okay, what was your process? Who are you competing against? Why do you think you didn't close as many deals as you should have? How did you rank against other people? That'll give you a really good story for this person. If you know they were hitting 40% and everyone else on the team was at 120%, then it tells you everything you need to know more than likely about this person's suitability to be the hire that you're going to make. But I think numbers are one factor. And I think if you get the story behind the number, you learn a lot about the seller to then understand, okay, how does this person fit into our organization? Rather than just judging, you know, purely based on, ah, a number that in a lot of cases and sales leaders and rev ops leaders are going to hate me saying this, these numbers get pulled out of the sky like, you know, they don't know whether somebody's going to hit this number. So quotas are a very magical Theme that I think, uh, sales leaders think is some exact science. It really isn't. And so I think judging people against uh, a very binary number is not the best way to be successful in your sales hiring. It's understanding more of that story which I think tells you, tells you a lot more.

Speaker B: Yeah. So what I'm hearing is it's important

Speaker A: to be able to peel back and really understand which sales people are really order takers versus selling the value and, and figuring out, you know, how to land accounts that, uh, that are really challenging. So you're. Yeah, that ends your co. Yeah, yeah,

Speaker H: yeah, yeah, yeah, yeah. And have they taken lessons from where they failed? So did they learn something? What, you know, what would they do differently in the next place because you know, they, they learned this, that or the other thing by failing. And so here's how they've changed their style, here's how they've changed their approach to things. Those people quite often will be absolute rock stars when you bring them into a good environment.

Speaker A: So hire the RC Cola sales guy

Speaker B: that could get shelf space, uh, working his way.

Speaker C: Exactly.

Speaker G: Yeah.

Speaker H: That's it.

Speaker B: I love it. Uh, Praytos, I know we're coming to the end here.

Speaker A: I want to get this last question to you. Um, this is from Daniel in Dallas, Texas. Daniel is asking for enterprises juggling multiple AI initiatives. How do you recommend they sequence investments across data talent and change management so they don't stall out halfway?

Speaker D: Uh, good question.

Speaker F: Sounds like, sounds like our question askers is uh, going through some of this right now. Um, it's a common, common affliction. Uh, you know, a lot of times the four or five things they mentioned are done in silo. Right. And you're not, you're not, you're not building an integrated view into creative creating value for the organization. So you've got the tooling, but you haven't got the change, or you haven't got the operating model and you haven't got the talent to actually go use the tooling or to manage it moving forward. So a well built strategy and a well built operating plan will cover for all of those work streams and all of those dimensions to take into account at the end of the day the project from a, uh, from a client's perspective, the project doesn't end when you go live. The project actually really begins when you go live. Right. Because that's when the value creation is going to start for them. Um, and once you have that kind of a mindset about. And my, my finish line is actually when I get to value, not when I get to go live or not when I get to vendor selection or whatever the case may be. Um, that's when you get a fully integrated view and do all the moving components that are needed to get you to success.

Speaker A: Well, I think we've tackled a whole bunch of real use cases that organizations are struggling with not only on this panel, but throughout the course, uh, of the event today. Ryan, Pritash, Jeremy, uh, thank you so much for the wonderful presentations, uh, during the day and uh, your time on this panel. And uh, for everyone that's listening in live or even on the uh, pre recording, um, if you'd like to reach out to any of the panelists on this panel, any of the other panels, or any of the speakers at the event, that big orange button at the top of the website will get you connected. I know we did not get to every question, but I think we uh, answered, uh, quite, quite a few. And uh, once again coming up next month we will have our security boot camp followed by our uh, Compliance and Risk Summit, uh, coming up in August. So, uh, thank you very much for everybody that joined the panel today and I hope everybody enjoyed the event.

Speaker H: Thanks Michael.

Speaker F: Thanks Al.

Speaker C: Today we learned that AI transformation isn't just about the technology. It's about changing how work gets done across the organization. Bekir showed us that the real value comes when AI is integrated into workflows, not just used as a standalone tool. By focusing on operating models, Xperis is unlocking new efficiencies and outcomes. Bekir, thank you for sharing your insights with us. If you want to connect with Bekir, you can find him on LinkedIn. And if you haven't already, come join us over@softwareoasis.com. that's where all of this really comes together. If you enjoyed today's episode, consider subscribing or leaving a review. It really helps us out. Until next time, take care.

Speaker A: Thanks for tuning in to the Software Oasis Podcast Network. With more than 200,000 listeners each month, you're invited to join our engaged community of B2B founders and executives at our next live full day Software Oasis Virtual AI event with the world's top B2B AI experts. Reserve your compass today. Get ahead with real AI strategies. Join the first Thursday of every month. Tap green button@softwareoasis.com or use show Notes link to claim your spot today.