What Success Looks Like When Enterprises Get Their AI Implementation Right

0:09 Welcome to NASDAQ trade talks where we meet with the top thought leaders and strategists in emerging technologies, digital assets, the regulatory landscape and capital markets. I'm your host Jill Malendrino and joining us this afternoon we have Jes. He is CEO of Highland as well as Shimon Ben David's chief technology officer of Weta. We're here to discuss what success looks like when enterprises get their AI implementation right. It is great to have the both of you with us. Certainly a key topic as we head into 2026. Jesh, there's been massive investments in enterprise AI, but we're hearing mixed results. What's your take on where the industry stands today? >> Yeah, you know, um first Jill, thanks uh thanks for having me. Um obviously very exciting topic. I would say uh we're

0:55 transitioning from prototyping uh to to outcomes uh tangible business outcomes uh depending on your favorite hype cycle or whatever where uh some organizations are in the trough of disillusionment as they you know refer to it they've gone from the peak of inflated expectations to boy I'm not really seeing the outcomes there's so many uh of these pilots that have failed or they're interesting but not value accretive. Uh what's exciting is uh working closely with our customers in you know very regulated mission critical industries healthcare insurance government uh banking etc. uh we are seeing some really high high ROI uh uh use cases and

1:40 outcomes around intelligence and automation from that intelligent agentic automation. Um so I think 26 is the year where we start seeing those tangible uh business results around productivity and efficiency. >> Yeah. And Shimone from your perspective what emerging challenges do you expect um customers to face as they're scaling AI over the next couple years? Mhm. So actually I'd like to join GS's point where we do see first of all that the first challenge is customers moving from these experimental environments to a real life environment and we we look at 2026 as the year of um return on investment. So it's not only about doing it but doing it in a costefficient

2:26 manner which makes sense. Um the the challenge is that a lot of customers are finding that a PC environment uh use case my PC AI environment is not equivalent to my production grade uh full scale out blown full full-blown environment. Um this is where we see actually challenges around the infrastructure where infrastructure that used to be very solid and valuable for my enterprise use case suddenly is not optimal for my AI use cases. things like storage, networking, obviously my compute. >> Yeah. And and Jesh, what are some of the biggest barriers that's preventing enterprises from achieving AI success? >> Yeah. Yeah. It's it's a great question

3:12 because we all see it, live it, are benefiting from it from a BTOC set of use cases. you know, we're all using our favorite uh Gen AI tool, whether it's Perplexity or Chat GPT or Gemini or maybe rotating through them. In the enterprise, you know, the buck stops with your data. You know, to really drive useful insight, it has to be from your data about your enterprise and to support your business processes. And so really the key areas it's not really about whether you're using anthropic or mistral or or an open source model. It's about how you're applying that model to your data for your processes for your

3:58 intelligent insights and from it agentic actions and outcomes. That is the key unlock that uh that we're enabling at Highland. We're enabling uh for our customers as a as a leader in all things unstructured data, you know, whether doing so so that the right agents have access to the right data and are taking the right actions in the appropriate uh business processes and workflows. >> Yeah. Well, touch another thing I want to address is system fragmentation as well and the continuation of operating in silos. Is it the point of this emerging technology to remove those silos? >> It's exactly right. You know, it's uh we've gone through many many waves of

4:43 technology uh innovation and disruption. Uh all of it accelerating fragmentation. You know, we have a system of record for sales. We have multiples of systems of record in sales, in in finance, in marketing, etc. They all have valuable data and context about the operations of your business. If you want to agent enable your enterprise, you need to contextually understand your business. You do that through your data. And you do that through your business processes and how you onboard customers, how you support customers, how you bill and invoice customers, how you do so for patients and healthcare by assigning them to beds and so forth. All of that, as I was mentioning earlier, comes down to being able to connect to all of these

5:29 systems. And if you if you know the data space 20% of an organization's data is structured 80% of it is documents. It's purchase orders, invoices, patient records, referrals, etc. Uh various contracts of of of various forms. What's exciting is now with large language models, you can give structure to these documents, give structure to these images, and from there you can make decisions and take actions. >> Yeah. Well, Shimone, what about the cost and economics of employing AI? >> So, so there's this um famous saying that I love quoting that says, "Last week in AI was the longest year in my life." Right? It it looks like every day there's new innovations, new frameworks,

6:16 new accelerators, new um development environments that are built in order to accelerate the workloads. However, one thing that stays constant is the infrastructure that's constantly being innovated on. But we know that what we're seeing more and more is that users or AI users are are hitting the infrastructure wall, right? Your your uh GPUs are not fully saturated. Um Jes mentioned multi being able to ingest and and query on multiple different types of data. So it's how do you orchestrate the entire infrastructure so that you utilize that in an efficient way. uh how do you make sure that your GPUs are constantly fed on constantly um generating tokens that are valuable for your customers instead of uh just pre-calculating things that they already

7:03 did. How do you make sure that you can burst into different environment and consume them in an efficient way? So there's no one answer there but there's there's a lot of challenges and eventually uh there's the SLA that you give your users. What one thing I was about to say also is that we've seen in 2025 that there's a lot of internal AI projects where your users you're very forgiving right maybe have my internal chatbot my mean internal HR policy but in 2026 we we see that going outward we see a lot of these projects going outward and obviously the SLAs that my external users and the accuracy that I'll need to provide to my external users increases so suddenly overcoming my infraure structure limitations and being able to be more accurate with my

7:50 answers in an ROI makes sense way is is paramount. >> Yeah. Well, you also make a great point in your notes here too that it's not just about technical efficiency. You you want to you don't want to fall into the trap where you're being identified as a cost center versus a profit driver. >> Brilliant. Yes. you you want to be able to actually accelerate the workload and and make sure that you're generating revenue like uh we have this mandate and we're seeing it in many places where um can I use a person to do it or is it much more efficient to do it with AI? We actually also see AI being multi being utilized to to skip time phrases. So not only people I know that there's this whole uh discussion around replacing people with AI, but it's also to accelerate time for example cast cancer research centers that are using AI to

8:37 predict cell growth and and um predict cancer um activity. So so yeah, you want to make sure that you unlock new capabilities and provide profit eventually. >> Yeah. >> Yeah. You know Shimon to Shimon's Buena as as an enterprise software leader we are the benefactors of the GPU optimized infrastructure compute storage uh that uh Shimon and team and others in the infrastructure are building out what I you know what's generally referred to as the standard compute and then accelerated uh compute. Um it's on that accelerated compute that uh the outcomes around uh AI integrated workflows to

9:23 reduce diagnostic um you know clinical workflows to reduce diagnostic turnaround times uh AI integrated uh workflows and processes that uh u accelerate time to loan approvals, mortgage approvals u and and so to your point it's it's not just about efficiency. about productivity that results in greater volumes and and greater transaction and greater effect on the top line. >> And by the way, there's the there's the thing that sorry for barging in that nobody really talks about, but it's the DevOps of AI. So whenever we're imagining AI workloads, we we imagine a model and it's running somewhere and it's answering and everything is hunky dory. Um the the truth is that uh as I mentioned there's so many innovations in

10:09 AI, so many frameworks versions, GPU kernels, um operating system versions, drivers. Eventually there's a massive DevOps uh and cost center that needs to operate that and and part of the challenge is how do you simplify it? How do you make an AI environment become almost invisible to the users to consume AI and just work with it? >> Yeah. Well, I'm I'm glad you brought up the um DevOps scenario here because you know certainly over the past you know let's call it since late 2022 when the you know AI entered everyone's vernacular there's been this rush to get internal and external product out there Shimone right and um I would imagine that impacts the process that developers engineers sisops devops and the way they

10:55 test and and you know go through their methodical procedures and I wonder if some of those gaps and vulnerb vulnerabilities are being exposed because to your point there's just projects happening all over the enterprise. >> So, so I jump into it. Yes, we we're seeing that initially when organizations are just jumping into AI and understanding the the land of possibilities, running projects, code development or running agents that are uh performing multiple actions. Security is is an afterthought. It's it's a proof of value. Let's get it right. Let's get it working. And then comes the um the how do we secure this environment? And actually what we're seeing maybe as a result or not is is the rise of more soft rain clouds. So obviously we we know about the hyperscalers, the neoclouds and there's recently there's

11:42 all of these sovereign clouds that we've been fortunate actually to to work with and you see that part of that is is a is a motion to secure the data to per uh country to allow and accelerate innovation on that country and and make sure that all of these security gaps are covered at a higher level even. >> Yeah. and and just ask actually let me I want to address a question to you and expand on this topic for a moment in thinking about divi designing with governance transparency security and you know regulatory compliance in place right um in industries such as healthcare as an example or financial services we are one of the most heavily regulated industries out there you have to think from an ethical perspective so rather than anti-inovation why not pro-

12:27 responsible innovation by designing with governance transparency and security first. >> Yeah, it's paramount. It's paramount. U you know, AI is only as good as the data that uh that you are able to work with. What's unique about uh how we're serving our you know 14,000 customers globally is um if you know enterprise content management it is the management of all enterprise unstructured data across your fragmented systems providing unified views of it and ensuring governed compliant access to that data. So it's built on a foundation of governance, right? Uh does Shimon have access to this purchase order? Jill, do you have access to this customer contract? And if

13:14 so, what elements should be redacted? What files should be accessible, not accessible? Is is central to content management? Uh which is what we have been serving for the last three decades. What we're doing now is extending these governance policies to the uh we're standing extending these governance policies to apply to the AI related insights and actions that come from those insights. >> Yeah. So, Chesh, when we think about 2026 and beyond, what does success look like when enterprises get their AI implementation right? >> Yeah. I think it starts with deep understanding of your business processes. deep understanding of the data that you have to build context on how your organization is operating.

14:02 um through that enabling multiple agent architectures, multi- aent arch architectures, agent meshes to be able to make contextual decisions and from those decisions automate business processes, take actions to get patients the care they need faster to enable students to enroll more productively into the into their universities of choice and higher education. to administer benefits at the state and federal level. Um to to approve loans, all of those we are at the precipice. We are actually solving 26 is really about driving scale enterprisewide and

14:47 globally. >> Yeah. And and to wrap up here, Shimona, I'm ask the same question um from for for you as well. When you think about implementation in AI infrastructure, what's the definition of efficiency and success? So I I I like to say um the nobody wakes up in the morning and is looking for AI. Everybody's looking for an outcome. So the success of AI is there is no AI. There's only an outcome that's achieved through AI, right? And it's not a silver bullet. Uh obviously success is measured by uh modernizing your environment so you can get an optimized utilization of your investment. um modernizing the storage networking and obviously the compute making it in a cost effective fashion because eventually you need to to be now a value profit and not a cost center as

15:33 as we mentioned and we're already seeing that and I expect that in 2026 we'll see more unlocking innovation there's there's replacing um some activities with AI agents and activities but there then there's more things that you couldn't have done had you not used AI can accelerate timelines uh predictions, unlock new research, new cancer environ new cancer solutions. Um autonomous environments. We're seeing a lot of autonomous environments that are built on these foundations and actually are changing our life. And my my my most favorite is when I can talk with my daughter about AI and she tells me how she's just interacting with AI on a day-to-day b and it's so natural for her. That that's success to me. >> All right. I appreciate both of your insights. Thanks for joining us on Trade

16:18 Talks. I'm Jill Melandrino, global market reporter at NASDAQ. >> Thanks for having us, Jill.