Everything Is a Code Problem | Uri Eliabayev

Season 1 Episode 2 | Everything Is a Code Problem | Uri Eliabayev

Founder of Israel's Biggest AI Community, on Why AI Strategy Is Really an Engineering Problem

Uri Eliabayev, who founded Machine & Deep Learning Israel, now the country's largest AI community with tens of thousands of members, joins Sascha Poggemann to explain why he evaluates AI models on task-specific outcomes rather than benchmarks, and how reframing any problem as a code problem changes what teams believe is possible to automate. They also discuss why AI adoption succeeds fastest when it aligns with employees' own motivations rather than being mandated top-down, and why enterprises should embrace an iterative, patient approach instead of expecting one-shot results.

👉 NiCE:
 https://www.nice.com

🔗Connect with Uri:
https://www.linkedin.com/in/urieliabayev/

🔗Machine & Deep Learning Isreal
https://machinelearning.co.il/

🔗Connect with Neeraj:
https://www.linkedin.com/in/neerajsverma/

____________________

01:03 — Building Israel’s AI Community

05:33 — The Next AI Frontier

10:12 — Taking AI Into Production

14:01 — Making Enterprise AI Work

18:45 — AI Governance and Regulation

23:34 — Will AI Replace Jobs?

26:24 — Becoming Better With AI

29:32 — Getting Employees to Adopt AI

_______________________

About Speak to an Agent:

Speak to an Agent explores what it really takes to bring AI into large organizations, why AI can't fix a problem you haven't already solved, why real understanding matters more than a polished voice, and why the domain expert who knows how to use AI is becoming the most valuable person in the business. The show features the leaders, founders, and technologists building and deploying AI at scale, and they get candid about what's working, what isn't, and what changes now that machines can reason, plan, and act at a human level, and sometimes beyond it. If you're responsible for putting AI to work inside your organization, this is a conversation you won't want to miss.

• (01:03) - Building Israel’s AI Community
• (05:33) - The Next AI Frontier
• (10:12) - Taking AI Into Production
• (14:01) - Making Enterprise AI Work
• (18:45) - AI Governance and Regulation
• (23:34) - Will AI Replace Jobs?
• (26:24) - Becoming Better With AI
• (29:32) - Getting Employees to Adopt AI

Creators and Guests

Host

Sascha Poggemann

Guest

Uri Eliabayev

What is Speak to an Agent?

Speak to an Agent explores what it really takes to bring AI into large organizations, why AI can't fix a problem you haven't already solved, why real understanding matters more than a polished voice, and why the domain expert who knows how to use AI is becoming the most valuable person in the business.

Featuring the people and ideas shaping the future of CX AI, the innovators, the thought leaders, and those making the decisions. They get candid about what's working, what isn't, and what changes now that machines can reason, plan, and act at a human level, and sometimes beyond it. If you're responsible for putting AI to work inside your organization, this is a conversation you won't want to miss.

Uri: 00:00

There's a misconception that now everyone is going to lose their job. The most deadliest creature in the world is a domain expert who use AI. Yeah. You cannot beat that. You will be bulletproof.

Neeraj: 00:11

Welcome to Speak to an Agent. Those words used to mean one thing, get me a human. Now the agent people actually want is AI. It's faster, sharper, and never puts you on hold. If that surprises you, you're exactly who this show is for.

Neeraj: 00:27

The builders, the decision makers, the implementers, comparing notes while the next generation of technology gets built. You're not just listening. You're in the room. I'm Neeraj Verma. Today's episode is guest hosted by Sascha Poggemann.

Neeraj: 00:39

Joining Sascha is Uri Eliabayev, AI consultant and lecturer and founder of Machine and Deep Learning Israel. Let's get into it.

Sascha: 00:46

Tell me a little bit on how you built that community back then. I mean, it wasn't like all over the place yet, right? How did that get started, you being a founder there?

Uri: 00:56

Yeah, actually it's a funny story. It was a coincidence. Back in the days, a decade ago, I was in a meetup in Tel Aviv about deep learning, and it was super fascinating. It was handled by someone who just sold his company to Facebook, an Israeli company called face.com, face recognition, they sold the company and they gave a lecture about deep learning and then my eye will open. I said, wow, super fascinating, super interesting, how can learn more about that and just meet the people who are doing that?

Uri: 01:28

So I'm still not sure why to this day. I thought to myself, okay, I will open a community, a Facebook community, and I will start to write more about that, and over the years it become much bigger and bigger, and now we are like the platform for everything in Israel, but it was totally coincidence.

Sascha: 01:49

And that's founder mode, like not knowing where exactly you're going, but it took a lot of time and investments in a conceptive way and constant way that you poured into. When looking back now at those ten years, were there certain themes that emerged that really sparked your mind?

Uri: 02:09

Yeah, it's just a good question. I think back in the days, it was silos for computer vision, for NLP that you are all familiar, maybe working a little bit about tabular data. So you will see everything as like a silo of research, so neural networks emerging and starting to become something much more mature, and there was the trends, and then three years ago, of course, Charge GPT moment totally changed the way we operate, work. We got many new people who got interested for the first time in AI. So the trends were very, very technical, very, very researched and now through this revolution, it's much more open to all, many developers, new programs are starting to blend in, so it was really nice to see the trends change throughout the years.

Sascha: 03:03

And that's exactly what I was imagining it to be, because in the early days it was very geeky, very technical, very niche to some degree, but you and others in this market have helped open up this type of community for a much broader audience in order to understand. Take us through what that looked like. What did you do in order to articulate very complex topics to audiences that are newly joining that community? How did you get them up to speed in learning AI?

Uri: 03:35

Yeah, this is a good question because even back in the days when it was super popular, it's still focusing on the science, and people sometimes struggle with that. Even though the majority of our people in the community have advanced degrees, PhDs, my main mission was to take all those advanced topic and allow people to learn and understand it even if they don't have like a PhD in data science. So we have created a lot of workshops. We have created a lot of conferences and meetups. When we invited people to do that and even so, I'm really proud of that, we took some of the famous courses by Stanford for computer science, deep learning for image processing, the classic one, and we opened it to the community.

Uri: 04:26

We had our own instructors and people who teach in this content and the most unique thing is that we took people who are from physics, from biology, for economy and managed to do the transformation with them. So now many of those people are now leading companies, leading AI in those specific companies, but back in the days, we worked really hard in order to let more people, even if they are not super technical, to join this entire community.

Sascha: 04:56

I still remember in the early days when we went on stage to talk about our stuff that we are doing at Cognigy, we are looking at faces like very weird. What is this technology? What is this going to do? Right? And we've come a long way from there.

Sascha: 05:11

What are you seeing like emerging now as the big trends in AI at the moment?

Uri: 05:17

So, first we have agents. Again, it's all over the place. I think this is for the first time we see them mature enough in order to do complex tasks, but if we are going to talk about the next big thing, I see that everything is related to physical world, physical models, world models, Now it's a huge trend in which models can generate world when you can interact, change or even move around the environment. So this is super fascinating. Also a very important trend for my behalf is multimodality.

Uri: 05:52

For the first time you can analyze not only text, but video, audio, and many elements of let's say inputs. And last but not least, this might sound really geeky, but this is my personal preference. For the first time you see models that are capable to create files. I work a lot with Claude about creating PowerPoint, Excel, Words, even much more complex formats, and this was possible not like six months ago, it wasn't possible at all.

Sascha: 06:26

Yeah, and for us, very interesting to see that you are also testing new frontier models and evaluating those. Something that we are seeing in our space, broken it down into conversational AI and agentic AI, those models have become so good, right? And many of them are good enough. Exactly. What's your take on that?

Sascha: 06:47

Where's that industry heading with the new models and the new releases and everything?

Uri: 06:51

Yeah, this is a good question because nowadays it's really difficult to evaluate the new capabilities of the models. We see big announcements, but at the end of the day, people are not sure always how they can understand what will the new changes, what is the delta. So what I say to majority of my clients, but also in my community, that it needs to be task specific. You couldn't care less about the model or the benchmark because we all know that benchmark is nice for PR, sometimes they are not telling the entire story. What aiming for is just to have a specific task that you know what are the real benchmark and moreover, as a domain expert, you know how it should be completed.

Uri: 07:34

So this is the majority of the thing that we are focusing. And also in Israel, we have many companies that are actually the one who evaluate the models for the big labs. We have several companies that before any release, they are working with Natropic, OpenAI, DeepMind, they are evaluating them. For example, one company evaluate the risk of cybersecurity, how the models are good on just hacking into systems. So we are trying to figure out a different way to evaluate the models and see how they can react in the real world.

Sascha: 08:10

Yeah, in line with that, this is what inspired us when we first talked to the NICE teams. It's like how innovative they are, how deep they are in technology and this whole Let's maybe talk a little bit about the AI ecosystem in Israel because there's some frontier stuff happening there for sure.

Uri: 08:28

Yeah, I think what is amazing in Israel that we have players in the majority of the stacks of AI. Unfortunately, we don't have in the LLM, in the foundation models, top players, but we do have for, let's say medical, we have a company that has a foundation model for medical. We have a company that is doing foundation modeling images and videos that can give the attribution of the people that was trained about in the data. So this is something that in the bottom. Israel, they are really good also about orchestrating tools for developers in order to better perform your LLM evaluation, everything around that.

Uri: 09:16

And also in the layer of hardware. We have many companies and also multinationals who design the chips that we all use if it's TPU by Google, Anapurna by AWS, and Vidy of course, has a huge R and D center. So Israel is also good at that level, and many more other layers that are part of the stack.

Sascha: 09:38

Right. Yeah. And now you have all these frontier models, the core capabilities, some of the base foundational technology, but if you put that into operations, if you put that into operations at scale, what has your community learned about that? Like taking a POC into production, for example, or

Uri: 09:56

even Yeah, this is a for good one, I think, because when you're starting to meet clients, to meet real data, you see that the thing that you have built is not always good as you thought. And so we're putting a lot of effort on several elements. The first one, as I said, eval. There's a lot of companies, but also a lot of best practices about how do you evaluate the process of agents. Moreover, a very important concept about memory.

Uri: 10:23

Memory of agents, we have a meetup about that next month. How do you train a model to become from a junior to senior? You want to create not only memory of the thing it does, but also to create experience. Know, how you become better of what you are doing. As human, we know how to do that, but how do you do the same for agents?

Uri: 10:44

So this is important layer, and also the ability to reinforce learning to do on the models in order to let them improve by itself. So this is something that also a lot of focus and just, you know, to learn as you go by, meet the clients, fill the production, and then you will be able to just get along.

Sascha: 11:07

Well, the conversational AI and contact center space, what is very important to some degree is the predictability of certain use cases, right? So, while you want to have the conversational capabilities, you at the same time also want to have the ability to stick to process when needed.

Uri: 11:25

Exactly, and this is something I see when I'm working with customers that we distinguish between the planning process and which you can show a deterministic flow. Okay, this is step one, step two, and this is helping a lot of client to be sure that they know what they are approving. And then later on, there is a code that has been executed that is follow-up by the plan. So this is something I see the majority of the companies are starting to adopt, and the most interesting part that you can have one model as the planner, it could be for example, Gemini, and the execution could be held by clone. So this is something really unique that allow you this way to maneuver between models and choose the best one for your task.

Sascha: 12:11

Yeah, and in line with that, when you take certain use cases into production, you need to give it much more of a monitoring, much more oversight, and sometimes even human in the loop abilities to really navigate that it does what it's supposed to do, right?

Uri: 12:28

Exactly, this is the hardest part I think nowadays because even sometimes it's hard for companies to evaluate the work of humans. When I come to companies, okay, you have this process, how do you measure it? What is the benchmark? And sometimes they tell me frankly, we don't have anything. We know the things are getting done, but if you are now building the entire process for agents, you have to have those sensors.

Uri: 12:52

Sometimes I come to companies who are like 20, 30, 50 years old and they are used to do something that is aimed or it was built for human, but now that you need to build for agents, they need to break everything and build it from scratch again, the processes.

Sascha: 13:08

Yeah, well, and staying with those enterprises and those customers, what we've seen is that it is not enough to just digitalize their traditional processes, right? But it's a real transformation journey where you need to change the processes underneath the backend systems, connectivity, in order to really gain the value that you would want to see from adopting Frontier AI technology. Maybe we can speak about that. What's your learning there with the clients that you

Uri: 13:38

Some of the mature process or the thing that we understand that it's not easy to implement AI. There was a promise that it's so good, it will happen by itself, just plug the technology that you bought and it will work, but for what I see, customers still need the guidance, still need the understanding how they can get ROI out of this, how they can implement that in the right way. And I think that now people are starting to understand it, we will have much mature approach for the entire ecosystem, because AI is still not in a place where you just do it or implement it and it's working. You still need to work really close with the customer to be a domain expert as you are and understand their needs and help them hand by hand to craft the solution for their own needs.

Sascha: 14:34

Yeah, and in line with that, your community probably moves at lightning speed, whereas enterprises cannot, right? And I can just see cases where they are being surprised on what is all possible or the next demo case that is shopped around somewhere. How you match and help match between the art of the possible and reality?

Uri: 14:58

Yeah, first of all, I tell everyone that it's okay to have a FOMO. It's legit, we all Everyone have feels like they are left behind, they are not up to date, but I tell them that technology takes time to evaluate and evolve and to become something that you can rely on, and this is the most important for enterprise. You need something that you can rely on, and if you are chasing the most advanced model, you might feel that you are up to date, but sometimes this is not the right thing for you. For example, I just spoke with Benny about that here. Maybe the price is too high for you, you don't need to pay that much for a single task.

Uri: 15:38

Maybe the latency, it's too much. Maybe your user want fast answers. So sometimes enterprise need to understand that they need to find the right solution for them and not the last one. The last model is the best for them. Majority of the time is not like that.

Sascha: 15:55

Yeah, and staying on those enterprises that are having the appetite to learn, to understand, what would you say from the Israeli AI community? What do you want to bring to the enterprises here in Mainland Europe and in The UK?

Uri: 16:12

Wow, this is a good one. I think in Israel, the best thing is that even enterprise that are traditionals are early adopters. I'm speaking with banks, insurance companies, all the high regulated companies in the world or specifically in Israel, industries, and they are so open to technology. They want to test, they want to implement, they want to take it to production. Of course they are doing it in a careful way, but I think this is the main message.

Uri: 16:37

Don't be afraid to try, don't be afraid on integrating the core technologies to AI. And I know because in Europe people are a little bit more conservative about using technology or how they can adopt that. In Israel, we are moving really fast on that, and this is something I would suggest more others to follow.

Sascha: 16:59

Yeah, 100%. And so you're saying on the one hand that they need more risk appetite, also appetite to explore new things, potentially also to fail?

Uri: 17:09

Yeah, basically if you are doing the right preparation work, and this is what I do with the majority of my clients, I'm coming to the company and learning about what they are doing, what are the processes, what they're trying to improve, and the most fascinating thing that I'm not talking about them in the beginning about technology or AI, because sometimes they are so eager to use AI that they are missing the main element because you want to improve some business metrics. This is the purpose of what we are doing. The purpose is not to use AI for the sake of using AI. It doesn't make any sense. So when I come and work with them on strategic plan, then we can later on see how can AI, let's say improve what we are trying to do, but it still needs to be stick to a KPI, it still needs to be relevant, and sometimes I tell them, you know what, AI is not relevant here.

Uri: 18:04

You have a much simpler solution, you don't have to do that on purpose, and they really appreciate that.

Sascha: 18:11

Yeah, and from your vantage point at the Rice Institute, how do you help resolve some of that tension between rapid pace of innovation, the open community, and where there are these strict needs for enterprises that need to adhere to policies, procedures, governance, safety, and guardrails?

Uri: 18:29

So this is a good one because some of my work is to work with government offices and some of regulations, some of them are designing, let's say the policies, and I always tell them that a policy without a tool doesn't move much because let's say you want to enforce something, fairness, bias, all those guidelines, if you don't have the tools to do that, so it's meaningless. So I'm working very close on see how we can take policies, and policies are important, regulation is important. This help us to have those rules and to protect the people, but you need tools that will allow you to enforce that, because let's say you want models without bias. The developer or the data scientist who is responsible for that, he also, he or she, they don't want the bias, but sometimes they don't have the tools, the technical tools to know if the data is unbalanced, if they are less representative of some people or minorities or etcetera. And this is the most interesting element because I'm telling that to enterprise.

Uri: 19:40

You can move fast, you can have all the galleries, you can have all the policy, you have tools that will allow you to do that, use them.

Sascha: 19:48

Yeah, and I mean, this is where all the new frameworks on ISO 42,001, AIC4, these frameworks of audits are emerging for a reason, right? Exactly. Is that something that you're seeing in Israel as well, or is that just happening in Mainland Europe these days?

Uri: 20:05

In Israel, they like to move fast, and after it's working to think backwards of how we can adjust that. So in Israel people like to build fast and see if it's working, get some traction, and then later on they add those layers because it's important for them to move fast and fail fast and you can adapt because if you are like a young startup and you put too much effort on policy and regulation before you have product market fit, so it doesn't make sense. You need first to understand the market. You need to see that if the technology is ready and then you will be able to, let's say adapt and make the changes. And also, as I said before, there are a lot of technologies that help you to do that relatively easily.

Uri: 20:53

Yeah,

Sascha: 20:54

mean, for us, these policies and audit frameworks are really big because they become a competitive advantage. We adhere to these policies, procedures, we have outside counsel to test against them, so that creates a lot of credibility. While I'm fully on board with you, in the early days you need to explore things without boundaries, you just need to think big, think outside of the box, think outside of your business, but when you think about production systems, need to take it back into this is a governed framework so that you can operate within the regulations that you have.

Uri: 21:29

Exactly, I couldn't agree more. I think once you understand what is the solution and what you are aiming for, then you should start to think about regulation and how to build it in the right way, because again, people are sometimes cynical about regulation because this is not succeed and you don't want it, but I truly believe this is important. Mainly if you don't understand the full outcomes of AI, sometimes we cannot understand what is the data that's been used or what the models learn about you. So it's okay to understand what are the most, let's say relevant, let's say customers for you, what is the product market fit, and then do that, but you need to have in mind that regulation is coming. I can give you an example, very good example, I have a friend of mine who developed AI models for medical and they moved really fast, managed to build a model that is like 100% accurate and then they realized that in order to get the FDA approval, they need to work in a different way, the way they label the data and then they realized that everything that they have done, they need to remove it and not use it because the FDA will never approve that.

Uri: 22:45

So this is again, something that's really important, and this is my advice to entrepreneurs, move fast, it's okay, but have in the back of your mind that someday you need to adjust and have the pipeline for this adjustment.

Sascha: 22:58

Also in order to be able to sell your technology into clients that are regulated, that are not getting out of these frameworks, but Now you sit outside the vendor world, so you can say things that others cannot. What do you think? What is the most overhyped idea in AI these days and what is most underrated?

Uri: 23:18

Wow. Overhyped? I think that people will get fired because of AI. There's a misconception that now everyone is going to lose their job, we will not do nothing and AI will do everything, and if you look into the reports, there was last week a report from The US that software development jobs are in the most high rates ever. People are still looking for developers, people are still looking for, let's say, knowledge and domain expertise, and the most, let's say, deadliest creature in the world is a domain expert who use AI.

Uri: 23:59

You cannot beat that, and this is what I'm trying to preach to people. If you are good at what you are doing, and you know exactly, and this is something that you will be bulletproof. No one will be able to fire you if you are really good of what you are doing, so I think the other hype is people believing that AI is so powerful that you will lose your job for that. So, and less hype, this is a good one, I think that people, they are not understanding that you can solve everything that is code oriented. I was in a lecture by Anthropic several months ago in December, and they told that they are doing a paradigm shift, that they are looking for every problem as a cold problem.

Uri: 24:49

No matter what you're trying to solve, no matter what you're trying, this is a cold problem, and once this told us, this opened my mind, and I think that when you realize that everything is a code problem, you look on things in different way.

Sascha: 25:05

That's interesting.

Uri: 25:05

Because the Anthropics are this, it's not mine. They say, for example, we want to create files, we want to run some experiment, everything that they are doing, they're trying to make it a code problem, no matter what, and then it's opened my eye, I think once people understand that, every process, every procedure, everything you are doing, mainly in a digital world, it's a code problem, things will change.

Sascha: 25:32

And this is where AI then comes in to help build or out that whatever that future of code is gonna look like, right?

Uri: 25:39

Exactly, exactly. I think that it will be hand in hand with AI that will help improve AI. We see also this kind of content, and also companies who long to live with AI and adapt really fast using these technologies.

Sascha: 25:56

Yeah, and in line with that, if you could get every builder out there in this room to adopt one habit to stay relevant as AI agents take over, what would that be?

Uri: 26:08

I think the main element is what worked for me, is work on your memory muscle of how to tackle task in a way that AI can help you. Because the majority of the time, people don't think about AI as enabler. They know to use AI in a specific task, but they are not, let's say, evolving in a way that new task that they won't show that it was possible with AI, they don't even think about that sometimes. I can give you an example. I have my own podcast and I wanted to edit my videos and create reels and shorts out of it.

Uri: 26:49

And I thought to myself, okay, cloud cannot do that. But then I thought about, again, this is a code problem, right? When you are using Adobe or Premiere, they're actually writing code in order to make those cuts. So I sat on this task, tried to solve it in forty eight hours, I download all the packages that are actually doing video editing, and adding subtitles, and everything, and then I have a tool that can take a podcast, understand where are the good parts, cut it into use, add subtitles, and create Excel file with captions to social media.

Sascha: 27:28

Right.

Uri: 27:28

So why am I telling all this long story? Because I'm working really hard with myself to think about each and every test that I have, even if it's not straightforward in using AI, to see how AI can help me and how I can make it into a core problem, which I can solve by software.

Sascha: 27:47

Yeah, and what you're describing is a transformation on your own role, right? On your own skills and capabilities and how you leverage AI in order to do so. And that's something very exciting to see. Across our customer base, we're also seeing that there's not enough humans that want to do this contact center agent, human agent job at the time. At the same time, the contact center volumes, customer interaction volumes are rising many fold.

Sascha: 28:16

There's new use cases that you could spin. So totally on board with that story that you're saying that right now we are not in a place where everyone is losing their job because of AI, but what enterprises need to do is they need to elevate their productivity gains. What we are seeing is that there is a gap emerging from those that really embrace AI in a structured way, maybe even breaking it down. I like that analogy there, breaking it down into a cold problem, but also those that are lost in the bubble of everything is AI, right? So that navigation in this time, in this complex time where the world is moving so fast, where new capabilities are being announced on a daily basis, new things are emerging in terms of sovereign cloud and independence and things like that.

Sascha: 29:10

That is a complex ecosystem to navigate. How would you be able to help with that?

Uri: 29:16

This is a good question. I think that enterprise are struggling with AI, but sometimes we need to look on a lower level that enterprise are employees. At the end of the day, enterprise is like non defined thing, well, what is the enterprise? Basically it's the people, and sometimes you need to understand that you don't have always alignment of interest with the humans when being forced to use AI. And this is something many companies miss because when you want people to adapt, you need to understand their motivation.

Uri: 29:55

Some of them are afraid to lose control. Some of them are afraid to lose their jobs. Some of them really love what they are doing. It sounds bizarre, but there are plenty of people who love their work, and they want to do They're

Sascha: 30:09

getting more creative with AI, Exactly.

Uri: 30:12

People, sometimes I need lawyers or accountants that they love what they are doing. They love to read contracts. They love to do this accounting stuff, and some, or for my exam, personal note, my wife is a programmer, and she told me I really like to code, I don't want AI to do that for me. So sometimes you see that when you are trying to implement AI in a way that you just tell people use AI because we need, or I know some companies that measure tokens consumptions, what does it mean? Why to measure how many tokens?

Uri: 30:46

You should measure the outcome. And when you are trying to do this mind shift, you need to be aligned with the motivation, the interest, even the feelings of the people. And I see that sometimes when you are doing that, the adoption rate is much higher because people want to use AI. The majority of them, and you see that in Israel, everyone is early adopters and use AI, everything, but when it comes to your personal work, things a little bit change. You can use in your openness a lot of chat, GPT, and cloud, everything, but in my work, I know that I still need to be responsible of the outcome, I don't want to be replaced, sometimes the tools are not good, so you need to have all these mechanisms together in order to understand how to do it in the right way.

Sascha: 31:33

Yeah, absolutely. And we are taking organizations now through different stages of AI adoption. And one motion is bottom up, really, the employees and availing them the tools, exactly what you just described. But then also, from the top down, you need to navigate. You need to give it structure, you need to give it guidance.

Sascha: 31:54

Exactly. And back to the discussion that we had earlier today, around you cannot just digitalise broken processes. Sometimes you really need to rethink how we operate, pull it all together, maybe into a single platform, maybe into some very verticalised solutions where needed, and then you take it forward into something that operates much better. And that transformation is a complex task, and you need a partner in order to go through that motion.

Uri: 32:25

I totally agree. This is what I said before. AI still is not in a level where you can just plug it and it will work. You know what? Even much simple stuff, when you see companies that are moving to CRM, having a new task management platform, even much basic stuff, still need configuration, still need to have this tailor made work for you.

Sascha: 32:50

Yeah, connectivity to And other systems, much more

Uri: 32:53

I see that in AI, when as you say, it comes from the bottom and from the top. On the bottom, I see that companies have like champion program in which they found several people who are really eager to learn more about AI and they love it, and also from the management, they know that they need to have a policy about what you can do, what you cannot do. They know they need to have some, let's say workshops for the employees to learn how to do that that way. So I truly believe that both ways are good. They could be in parallel.

Uri: 33:24

And the main element is to let the employees feel comfortable about that.

Sascha: 33:29

Yeah, we could continue to chat about this topic for much longer, but I think we somehow need to wrap it up shortly. What's on your mind? What would you give the audience as additional advice? What do you want to say?

Uri: 33:45

I want to say that don't be afraid to try. Sometimes people think that in one prompt, in one line, zero short learning, they need to get the best results. It doesn't make any sense. As human, we are not operate like that. Even you need to explain something to your colleagues, if it's really smart, know, so majority of the time you need more than one.

Uri: 34:08

More one explanation, one process, it's okay to have a process, it's okay to take the time for the implementation, it's okay to craft your AI journey and don't try to do that in a one shot really fast and enjoy the process.

Sascha: 34:25

That's a very good one because things that have not been possible three months ago are now possible, right? So we need to stay curious. We need to stay humble and bring it down to really, like, the roots of it and and understand the the underlying problems and then try to solve it with AI or sometimes even without AI.

Neeraj: 34:43

And that's a wrap on this episode of Speak to an Agent. A huge thank you to Uri. If you found the conversation valuable, subscribe wherever you listen and leave us a rating or review. It helps more listeners discover the show. We'll be back soon with more of the people making the AI agent the one you actually wanna talk to.

Neeraj: 34:58

Thanks for listening.