AI Agent Builder Summit, Encore Presentation | Abdi Goodarzi, Deloitte

Clips
More from AI Agent Builder Summit, Encore Presentation

Abdi Goodarzi

Head of Gen AI Products, Innovations and New Businesses

Deloitte

play_circle_outline Building Trust through Successful Pilot Projects: Simple Use Cases for Tangible Results

play_circle_outline Shift from assistants to co-workers for better decision-making

play_circle_outline Early adopters will be winners in the AI revolution

play_circle_outline Role of IT evolving to a business partner and culture change

play_circle_outline The Crucial Role of Chief AI Officer: Empowering Strategic Thinking Without Job Replacement

play_circle_outline Maximizing AI Performance: Importance of Data Standardization, Executive Buy-In, and Starting Small

play_circle_outline Maximizing AI Solutions: Zora Methodology for Risk Management, Compliance, and New Technology Adoption

Info
Transcript

Abdi Goodarzi, Deloitte

Abdi Goodarzi

Head of Gen AI Products, Innovations and New Businesses Deloitte

Abdi Goodarzi, head of generative AI products, innovations and new businesses at Deloitte Touche Tohmatsu Ltd., joins theCUBE’s Scott Hebner at the AI Agent Builder Summit to explore the business strategy behind agentic AI. Drawing on 25 years of experience guiding Fortune 500 clients, Goodarzi shares a grounded perspective on how organizations can responsibly integrate AI into core operations.

The conversation highlights practical use cases as a launch point for AI adoption and outlines the evolution from digital assistants to AI-powered coworkers. He... Read more

explore Keep Exploring

What is the recommendation for organizations thinking about adopting AI technology in their operations? add

What was highlighted by everyone in the conversations at the summit? add

What should companies be considering when it comes to ROI on agentic investments? add

question: What impact has the changing role of IT had on organizations and their decision-making processes? add

What are some key considerations for companies looking to take advantage of AI technology in the workplace? add

What is a recommended approach for implementing AI within an organization, starting with small use cases and gaining executive buy-in before expanding and leveraging the creativity and innovation of late adopters? add

How did the idea of creating a platform with Zora come about and what are the main reasons for its implementation? add

bolt Powered by CUBE AI

Abdi Goodarzi, Deloitte

search

Scott Hebner

>> Hello, and thank you for tuning in to theCUBE's AI Agent Builder Summit. I am Scott Hebner, the Principal Analyst for AI at SiliconANGLE Media and theCUBE Research and your host for the summit. Welcome to this industry-unique digital summit, dedicated to exploring the rise of a Agentic AI and how you can speed your way to ROI and competitive differentiation. The summit consists of 12 sessions where you'll hear from over 20 industry pioneers sharing their real-world experiences in building AI agents and deploying agentic workflows. This includes vendors that provide software and platform solutions and end-user business leaders discussing their aspirations, needs and strategies. Today, however, we're going to discuss the more strategic aspects of Agentic AI, a business perspective, if you will. The why of Agentic AI. How to get started, the keys to success, the use cases that are delivering proven ROI already. So a big-time focus on business, strategy and how that relates to the technology of Agentic AI. And to help us with all this, I'm excited to have Abdi Goodarzi, the head of Generative AI Products and Innovations at Deloitte. He's here to join us today. And Abdi, thank you very, very much for being here.

Abdi Goodarzi

>> My pleasure, Scott. Great talking to you. Looking forward to the conversation.

Scott Hebner

>> Yeah, I think what we're looking for you to help provide is that holistic point of view here because you, vendor-agnostic, obviously you deal with a cross-section of businesses, and so, you all know Abdi's been a trusted advisor to the Fortune 500 for a long time. He's an industry veteran with over 25 years of experience in what? AI, cloud computing, data analytics, ERP, strategy, a whole bunch of things, automation, and I think you can help us put all the pieces together as we hear everything on the summit and what we should be looking for.

Abdi Goodarzi

>> Absolutely. Yeah.

Scott Hebner

>> So let's start off with where I think most businesses should probably start, which is thinking about use cases. What are some of the specific use cases that companies should really be thinking about and what do you see in the most interested in terms of use cases?

Abdi Goodarzi

>> That's a great place to start, Scott, and definitely, my recommendation to every one of our clients that is thinking about where do I start with the agentic solutions? Number one, you have to take a step back and understand that this technology is like no other technologies that you have used in the past. And I'll give you an example. When client server concept became available and we passed beyond mainframe, it was just a level of integration and distribution of the data, how the systems would work and talk to each other. Later on, consolidation was a big trend in the industry and ultimately, we got to cloud, which basically made things simpler and faster and cheaper and better for us. Now, there are a lot of people out there that think about, well, AI is just another technology that is a partial statement. And to some extent, I agree with it. However, AI has some characteristics that no other technology of the past had. Was there a technology in the past that could actually make decisions and recommend ideas on its own? None of the technologies of the past could do that. So I think from the get-go, my recommendation to people is treat AI as technology as is, different than however you used it in the past and think about it as a digital companion. If you had humans with massive amount of learnings and capabilities and had access to every piece of information on planet earth, how would you use them in your organization? How would you train them on the data, the part of the business or the elements that makes your business successful? And if you treat the use of AI like that in your business, you can definitely come up with use cases that are valuable. Second thing is a lot of organizations are thinking about going after complex and sophisticated use cases of AI. Well, the more sophisticated, the more complex it is to bring it to life, and obviously, the trials and turbulences of doing that will also cause some delays and perhaps some disappointments. So the best is to use simple use cases that allows you to feel and touch the results and gain confidence and trust in the ability that AI brings to the organization. Second thing that that case does, I also go around and tell people, us humans, we've been trusting each other for thousands of years. Even when we made mistakes, we gave up the mistakes and figured out how to course correct. Humans haven't trusted the machines. We don't know what the machines are capable of. We're constantly questioning it. So if you start with small use cases, come up with ideas that can have tangible results and you scale that up, you gain trust to the product, you actually see it in action and it's easier to commit and adopt it. And that's my recommendation to majority of my clients.

Scott Hebner

>> Yeah. You hit on the trust index and the trust factor, which in every single one of the conversations I've had, and certainly, for the 20 or so people that participated in this summit, trust was one of the things everyone highlighted. Because like you said, we're evolving from tasks into coworkers that are designed to help you do things better, make better decisions, problem solve, solve goals that you have, and then ultimately, even further, to help an organization do all the same and trust becomes more and more important. It's like it's the currency of innovation, right? No trust, no ROI, and that's what I think makes it such a profound advancement. Because you're right. We went from interconnecting everything, to instrumenting everything, to changing the economics of everything with cloud. Now, we're into the world of intelligence and Gen AI was step one, predictive models and then Gen AI. But like you said, this is really where you start touching on co-workers, actually working with humans and almost being a co-worker versus an assistant.

Abdi Goodarzi

>> Yeah, absolutely. Absolutely. And a lot of organizations start with an assistant, which is a great place to start with. However, the ROI of assistants and co-pilots is limited because co-pilot is designed to actually be a simple companion versus if you leverage more sophisticated capabilities of AI, you could actually activate advanced reasoning. You can actually, beyond task execution and workflow processing, go after problem solving. And that to me, is the ultimate status where you can truly say, I have a digital companion in my business that can help me solve the problems. And again, remember this digital companion has a tremendous amount of capacity for learning. You can almost say it's limitless, assuming you have unlimited amount of storage and computing that you can throw at it. So the advancements and capabilities that you can bring to life are in a sense, almost limitless, right?

Scott Hebner

>> Yeah, it's incredible. And actually, I think you hit on something I'm just going to emphasize here because there's a lot of noise out in the marketplace. So 2025 is the rise of Agentic AI. Everyone's talking about it. And part of what we want to try to do here is identify the signals in the noise. And I think you just hit on something which is, there's a difference between assistance and collaboration. There's a difference between assisting a human being and doing something versus collaborating or co-working with them to get something done. And maybe that's the way we need to define what Agentic AI is all about. We don't have the masquerading assistance that... Because everyone's doing Agentic AI now, right?

Abdi Goodarzi

>> Yeah. And there are a lot of capabilities out there that are more mature than others. If you think about code generation companions, there are many of them out there. Well, code generation or code creation by humans is a major activity that happens in every single enterprise to support the business expansion, contraction, or even opening new market, opening new products and so on and so forth. Well, if you go after those use cases that are proven and the AI status is a lot more mature than others, you can immediately see benefits and the benefits will not come in terms of value gain or dollar sign that is massive, but you save time, you save effort, which gradually allows you to do what, in my opinion, is the ultimate thing to do with AI. Is reimagining what we do today. And I tell you, an example of that would be when was the time that humans thought, what if the cars drive on its own? That was more of a dream, but ultimately, we reached to a point that technology advanced so much that reimagining how the driving is done is autonomous driving. And the same concept can be executed around autonomous back office, autonomous front office, and also, whatever the operations of the businesses are, it's absolutely something that can be done. Because we went from Gen AI to Agentic AI, nexus physical AI, and ultimately, AGI when we get there, and whoever is... When I say whoever, whatever organization that methodically thinks about this and goes with small use cases, small value to scaling up those into greater and grander results will be readier to disrupt other competitors in the industry. Right?

Scott Hebner

>> Yeah. The autonomous car is actually a great example of trust.

Abdi Goodarzi

>> Exactly. Exactly.

Scott Hebner

>> You start getting into physical trust there, right?

Abdi Goodarzi

>> Yeah.

Scott Hebner

>> So how should companies be thinking about ROI on these agentic investments? As an executive, how should you look to measure the value?

Abdi Goodarzi

>> Yeah, measuring the value, it's actually easy because for the past two, three decades, what we have actually matured quite a bit is business process flows. We know the steps and executions and workflows and preceding and exceeding and post activities of any process within the enterprise. From A to Z, we know all the processes. Those are great. We can basically figure out what is the process automation that we can bring AI for and advance the processes. And again, you go with activities that are simpler, less steps, less complex tasks, and you bring those values to life and then you expand upon that. To me, the moment that you can take that and say, "I am ready to actually change the methodology of how the work is done and reimagine how the work is executed," that's when you get to a point that you can ultimately say, "I'm fully taking advantage of AI." The first, I would say, the pioneers in this model are people who are thinking about effort reduction. That is a good measure for you to say, I'm ready to actually... I can see results, it's tangible and I can rely and trust it. The second is like time reduction. Now, you get to a point that things take lesser time to execute, whether it's any back office processes or any research or operational activities that you do. Like someone says, "My customer service wait time is four minutes or five minutes per customer in peak times, what if you take that five minutes and drill it down to one minute?" Because you've been able to train an agent that has a lot more knowledge and a lot smarter about how to respond with humans, and you can accelerate that entire process. That's a ultimate benefit. And eventually, when the last step of this measurement is when humans are comfortable to commit to those measurements and say that I can commit to it to do things faster and cheaper and sooner, that's when all of a sudden, you will see bottom line impact. But not until the humans actually commit to that. I think we will be in this dilemma of, technology, great, but I don't trust it or I can't commit to it. And up until then, we would be in the dilemma of, oh, I don't know whether this thing is real or not.

Scott Hebner

>> So, as time goes by, would you look at a different set or an extended set of ROI in the longterm, let's say three, five years from now if this continues to progress?

Abdi Goodarzi

>> Absolutely. And we have come to the realization that the pace of technology change has dramatically improved or velocity has gone up significantly, which in translation, it means you have to have a more agile organization to respond to change. Think about it, the language models that we used six months ago are now rewritten or enhanced to the point that it has a whole lot more capabilities added to it and advancements in terms of thinking, in terms of validations, in terms of reduction of noise in the results that the models are creating. So how do you go from this static model of execution? I need to do something, test it and validate it before I can actually productionalize it to, I need to productionalize things much faster. And that requires you to have a lot of agile capabilities. You don't need to document anything the way we documented the last 20 years, which was use the Word document and put everything in it, and that's your reference. Well now, you can have a living operating procedure that is kept in the memory of your language model that you have trained, whether it's a closed language model or open sourced one, no longer you need to do documentation. But that doesn't mean that system of record or the procedure is lost. Just thinking about how we execute, thinking about how we change our methodologies will be a huge factor in measuring and gaining success. Right?

Scott Hebner

>> Yeah. I would imagine if you're able to scale better decisions across an enterprise, an organization, making a better decision is probably a very difficult thing to measure the return on.

Abdi Goodarzi

>> Absolutely.

Scott Hebner

>> But an organization is a collection of decisions, and if you can just make better decisions because you have one of these agents helping you make that decision and helping you play out countless scenarios and giving you advice on why one action might be better than another, I don't know how you measure that, but I think that's maybe where it's coming in the years ahead, right? Beyond speed and adaptability and some of the things you pointed out, would you...

Abdi Goodarzi

>> Yeah, absolutely. And this elevates the role of IT to a whole new level, in my opinion, in the organization because up until now, IT was a supporting cast for any organization. They enabled systems, they activated or orchestrated capabilities for the business to take advantage of. Now, IT is actually to a sense, becoming a business partner for every aspect of what you're doing. Think about the arrival of digital workers. All of a sudden, you had human resources department. Now, you have IT kind of carrying task for digital workers. You orchestrate them, you train them, you educate them, and you make them more powerful or capable. IT will be at the seat of the table to determine how a business decision needs to be executed in order to accelerate that decision so you can take benefits faster. So, the role of IT in the organization has completely changed. And you see that chief AI officers, CTOs and all of that, they actually sit in the executive boards and executive rooms all the time. And that to me, is actually the arrival of digital workforce into the C-suite, right? So in a sense, which makes us to think about work and think about how we can take advantage of AI completely different.

Scott Hebner

>> That's another very interesting observation or insight there. Your chief AI officer is actually not a technologist about skills, but their role, it's actually an AI agent, HR person. That's an interesting angle on it, right?

Abdi Goodarzi

>> Yeah. Well, some of this sounds futuristic, sounds like far out, but in my opinion, the advancements we've seen in the AI space in the last two, three years, it allows us to be open-minded about how fast and how far we can go. And again, that imagination and how you bring that imagination into your culture. Because remember, we're talking about technology being super powerful. You still have the culture. Humans are the ones behind AI, humans are behind the business processes, humans are behind the change management. So, the sooner you actually activate an AI enabled culture, the sooner you can take advantage of these capabilities, which translates to my chief AI officer should be in the boardroom deciding where we're going and what we are doing. The use of AI in every aspect of the business, front office operations, everywhere, is a must-do. And as these ideas come to life, people think like, oh my god, AI is going to replace jobs and it's going to replace human... It's not. AI actually frees humans to think about more strategic activities and more critical things that matters to the business. And like every other industry revolution, the people who actually see the light at the end of the tunnel sooner, they're the frontiers in that they will take advantage of it. I'll give you an example without referencing any particular organization, but Amazon is the organization that took e-commerce far more serious than anybody else And that had its own results and remarks for them. It's the same concept. The organizations that see this and take advantage of it and experiment sooner, they have more proof of concept, adopt the use cases sooner. They are some of the winners of the future. And the future, by the way, is not 10 years away from now. It's like a couple of years, two, three years from now.

Scott Hebner

>> Yeah. And that's how the organizations have to get their workforce to think about this, right? It's about giving them superpowers to be more strategic, to do things they like to do better, where they can add more value. And the chief AI officer is going to help pair you with digital co-workers. It's a cool way of looking at it. And also, Abdi, I'm glad you brought up the culture point because actually, in our analyst pre-game show, before we kicked off all the sessions, one of the five success factors we talked about quoting Lou Gerstner from IBM is culture is culture's not part of the game. It's the entire game.

Abdi Goodarzi

>> Yeah, absolutely. Absolutely. I totally agree with that. And again, culture is the one that drives all these adoptions and changes because remember with AI, what you're doing, you're accelerating change. That's what you're doing. Some of these changes without AI would've supposed to happen 10 years from now as technology would've advanced without AI. Whatever it was after cloud, let's say AI wasn't, and whatever it was, it would've been 10 years of adoption curve. Now with AI, you have accelerated that adoption curve, therefore, your ability to observe change faster would be a huge success factor. And the only place that the roots of that exists is your culture. Culture of the organization.

Scott Hebner

>> All right, let's move into getting your advice, particularly for the parts of the audience out there. And I know there's many of them that are just getting started here, thinking about getting started on their Agentic AI journey here. And so, talk to us a little bit about what would be a successful first pilot? What would that look like? What should they think in terms of expectations around ROI? Especially around this notion of deploying digital workers.

Abdi Goodarzi

>> Yeah. So again, start small and try to grow your confidence into the use cases and the ability of the technology to give you what you're expecting that's number one. Second thing is focus on simpler processes. In general, in every organization there are many back office functions that have simple processes that are easier to take advantage of AI. And to me, those are finance space, for example, there are lots of repetitive tasks and repetitive activities that exist in that area that could be automated significantly and eventually,, converted into a digital worker concept. The third thing is if you haven't looked at your data standards and standards for a lot of business processes and technologies, thinking about implementing serious standards across the organization will be a huge factor. And I give you an example with the ERPs. When ERPs came along, this is 30 years ago, everybody thought, okay, I'm going to have one system of record. Well, that was just the story that people put one system of record in that ERP. Then they expanded beyond the ERP and build boundary systems and bespoke, sophisticated and specialized systems like one finance, one supply chain and so on and so forth. The system of record converted to systems of records. And with that, you will have hallucination, you will have errors that AI cannot handle because AI is powered by data. So thinking about how you standardize your data and don't go after, I need to standardize my structured data and unstructured data separately and so on and so forth. There are technologies that allow you to get rid of that noise and simplify it. But think about standards far more than you've done before will help you. And ultimately, again, go with the small use case, go with a small area and think about things that are easy to observe and easy to digest by the organization and then start expanding beyond that. Right?

Scott Hebner

>> And would you say the biggest potentially could be biting off something too big? Start with score points, put points on the board, play the game in the field, not the dug out, start doing things, learn and adapt, versus trying to transform an entire...

Abdi Goodarzi

>> Exactly....

Scott Hebner

>> Process or something. Yeah.

Abdi Goodarzi

>> So when you think about these use cases and how you would start, you always should start from the smallest unit of measure for the organization to think about applying the use case and growing and scaling that. I've mentioned this before, there are a lot of back office functions that is easy to do, easy to implement, but also, inside the IT organization, there are a lot of processes and procedures that are great use cases and great candidates for AI digitization. And typically, an IT organization is a lot more tech-savvy and their ability to adopt AI is better and faster. So I always recommend going after those areas. And again, don't make this a giant effort. Start small, do the proof of concept, believe it, get the executives behind it. Some of the things hasn't changed in the way we execute things. You got to have executive buy-ins that AI will make the change, will bring in benefits. And then after you truly prove that proof of concept idea and you can truly commit people to apply it and implement it within the organization and you can keep expanding. And I can guarantee you, I've seen this multiple times, even inside our own organizations, the people who were late adopters, once they felt comfortable with it, they're coming up with ideas of how to create more benefits with AI. And that's what you really want. You want a lot more believers and people who can feel comfortable with the technology. Once that happens, believe me, creativity will kick in, innovation, mindset because everybody knows what they're doing and where they can actually create relief or benefits. And that would be the beginning of this snowball effect that your values and benefits of AI keeps growing.

Scott Hebner

>> Yeah. Once again, back to the trust factor. Well, you've hit on trust, obviously trustworthiness, you've hit on quality, you've hit on transparency in terms of all the outputs that come out to gain all that trust. How are you architecting Zora AI, your methodology, to help ensure that all happens?

Abdi Goodarzi

>> Yeah. And with Zora, we actually apply all these ideas forward. If you look at Zora products, first of all, we created a platform with Zora. Why did we create the platform? Because there are a lot of solutions and capabilities and technologies around AI and organizations could actually at some point, get lost with which one actually gives benefit? Which one is the right one to use? And for an average organization that has multiple platforms, multiple systems for the enterprise, those are complex questions to answer. So we felt like as the experts in implementing and transforming organizations over the decades, we could actually solve that by having a platform. Then on top of the platform, we felt like if we are preaching this kind of a concept of start small, start with back office, start with simple business processes, why don't we activate those on top of the platform? So day one, when our clients are actually adopting the platform, they have a few use cases that they can very quickly deploy and start seeing results and being able to measure those results. And then this opens up the creativity and allows them to come up with bespoke ideas. Like, okay, I got this portion of finance done, but for me, the problem area is supply chain. I want to go fix that. Or I think about use cases inside my customer or service or sales, let me go do that. How about human capital and human resources? We got that. The other thing that we felt like it's important for Zora to have is handling risk and compliance. At the end of the day, Deloitte got an exceptional brand in doing risk management and compliance for many organizations. We felt like, well, if you want to create digital workers, we have the best qualitative knowledge to basically train those digital workers with these capabilities that organizations have to do for regulatory reasons and other purposes. Why not bring those things to life first and then allow the adoption curve to be manageable and acceptable by organizations? The rest of it would be history. And as we build more agents on Zora, we want to make sure the platform is robust. And by the way, we are applying exactly the same concept of, have agile organizations can accept the change. If a new language model comes in and it can activate Zora with more capabilities, that should not be our clients' and people or the user's problem, it should be ours. How do we make it happen quickly? So orchestration cycles are faster as well. So those are the attributes and characteristics we wanted Zora platform and Zora agents to have.

Scott Hebner

>> Well unfortunately, we're running out of time here, but before we go, I would love to hear, don't name the client obviously, but maybe an example of a real world example of Zora where the ROI surprised the client.

Abdi Goodarzi

>> Yeah. So we have a client that is a manufacturing organization and working capitals were a major problem for them. Thinking about cash management and how you can basically reduce day sales out and create a better profitability engine for the organization. And this was a multi-million dollar problem for them. And by using activating, actually, Zora working capital management agent, we were able to actually come up with foresights and information that allowed us to solve the problem faster because Zora agent dug deeper in the data and used different algorithms to bring out more insights to the executives and thinkers. Not only created more efficiency and productivity for them because think about how things are done today, an executive has an idea that something is a problem that's given as a task to middle management. Middle management gives to the analysts to go to the crunching of the data, querying the data and making sense out of the data and then bringing back to management layer to say, "Okay, here's what we found, what do we do with it?" So that what do we do with it is the problem solving. So we basically trained Zora agent with all that knowledge, years and years of industry, all the client issues that we had solved for that particular industry, and we made that problem solving happen in a more rapid pace. That allowed the executives to actually start thinking about, aha, I know where the problem is, let's go solve it, versus where do I get the data to tell me where the problem is? And that's majority of the time spent with the organization. We are able to actually, within the next few months, reduce that multi-million dollar number by 30% and the projection is to get to 50% within the next 12 months. So, I think that's a great use case where we started small and we're planning to grow it faster, better.

Scott Hebner

>> It's actually a great example too, of starting small, solve a problem, build on it, so, awesome. Well, Abdi, thank you once again. I encourage you all to make sure you also watch the pre-summit discussion that we had. That was a really great discussion also. So Abdi, I appreciate you being here today. Another very insightful conversation. For you all, please make sure you visit deloitte.com to learn more about Zora AI. Subscribe to the Deloitte AI Institute Insights and Perspectives newsletter. It's a great way to stay current in what's happening in the marketplace. And then finally, you can watch all the sessions from the summit, including this one by visiting the summit portal on thecube.net or checking it out on YouTube, on our YouTube channel. And also, you can get access to the articles and the clips and you can share those with your colleagues and on social media. So make sure you visit the portal. Thank you once again for tuning in. We are the leader in enterprise tech news and analysis. Bye for now.