Agentic AI: Unlocking the Future of Enterprise Automation

TLDR:

AI agents are changing business, enhancing productivity and efficiency across various sectors. In this talk we explore:

• Key trends driving AI adoption in different industries 
• Difference between agent frameworks and platforms 
• Use case: automating HR tasks with AI agents 
• Challenges in integrating AI with existing systems 
• Essential components for a successful agent strategy 
• Future predictions for AI adoption in business 
• Getting started with building your own AI agents 

Podcast:

Ever wondered how AI could revolutionize your business operations?

Meet Siva, the founder and CEO of Lyzr.ai, who shares insights on the impact of AI agents in today's business landscape. Discover how Lyzr.ai's platform empowers both tech-savvy developers and everyday users to craft AI agents that tackle mundane tasks, boosting productivity and cutting down on distractions.

Siva dives into the trends pushing AI adoption to new heights, especially in areas like customer service, sales, marketing, and HR, and unveils fascinating applications in banking and airlines. 

Gain clarity on the difference between agent platforms and frameworks, and learn how AI agents can integrate effortlessly with existing systems to deliver personalized customer experiences.

Join us as we explore the strategic deployment of AI agents in complex scenarios, like merger and acquisition analysis for large conglomerates, offering a cost-effective edge over traditional methods. 

Hear how small and medium businesses are leading the charge in AI integration, driven by their quest for affordable efficiency. 

By 2027, anticipate a landscape where AI manages a significant slice of organizational functions, transforming sectors like HR, sales, and customer support. 

Siva also shares a compelling narrative of AI agents conducting unbiased exit interviews, revealing their potential to enhance business processes without replacing existing systems. 

Tune in for an eye-opening discussion on the future of enterprise automation with AI.

Visit lyzerai.com for more information on agentic AI solutions.

Chapters

• 0:00 The Rise of AI Agents
• 10:19 Advancing Enterprise Automation With AI Agents

Transcript

Speaker 1: 0:00

Have you ever imagined a world where a powerful, scalable digital workforce who speak your language, understand your industry and business and work to deliver 24-7 business value are available? Well, this isn't science fiction. This is happening today, powered by the rise of agentic AI. Welcome to AI Unscripted, where I talk to the creme de la creme of the world's best business and technology companies. Today, I have Siva, founder and CEO of Lyserai, a local platform that allows you to build scalable, generative AI applications, on to talk about building AI agents in a matter of minutes. Welcome, Siva. For those who don't know you, would you mind giving them a brief introduction about you and Lyser? What is it that you do? How do you add value to people's businesses?

Speaker 2: 0:46

Sure sure. Thanks, kieran. Thanks for having me here. Happy to be here. I'm Siva, founder of Lyser. So at Lyser we built an agent platform that both the developers and the business users could use. So the billing the agent is the same process. You could build an agent to automate any repetitive tasks that you do and if you're a developer, you could access the APA endpoint and continue and integrate that within your software. But if you're a business user, you can build the agent and launch it as an app instantly without having to depend on any developer. So you automate a lot of your work on a daily basis so that these co-pilot agents run around you and helping you to whatever handle your productivity better or help you with less distraction, and so on. So that's what Lyser does. It's a platform for building agents and it caters to both developers and enterprise and business users I.

Speaker 1: 1:49

I love that you're aiming at both, by the way, because I look statistically and there's far more managers in the world than there are developers and therefore, the more we can open up you know great technology to allow people who aren't you know traditionally software engineers, to allow them to do amazing things, the better. So what, sibba, are the key trends driving the adoption of AI agents in industry and business today?

Speaker 2: 2:14

Yes, customer service, sales, marketing, HR. So if you look at the functions, the horizontal functions in an organization, these are the four areas where the agents are able to show immediate value. And then obviously you have finance, accounting and procurement as well, who are also seeing some use cases. So these are all, I would say, horizontal departments. But what's very interesting is the vertical functions. For example, banks automating customer refund process or an airline automating ticket booking. These are all the vertical functions, or rather business functions, where we are seeing agent-powered automation happening. So that's even more exciting. I mean, we are solving some problems. I'll walk you through a few more use cases during this. I would say call, but again, you're able to see some very strong value, direct value that customers are realizing in the business functions.

Speaker 1: 3:22

So, for those who are maybe not familiar with agent frameworks like, what is an agent framework? How complex is it to build an AI agent to grab value in these horizontals or verticals?

Speaker 2: 3:35

Good question. There is a difference between an agent platform and agent framework. Agent framework is the underlying blueprint of how the agent actually works. It's very important that you get the framework right. The popular open source framework is LanChain, and Lyser was built ground up as a framework from scratch to cater to enterprises. Now the framework is something that defines what kind of platform you could build on top and what kind of agents you could build on top of the platform. So, again, like I said, uh framework is where the nuances of how an agent handles a message, what kind of ai, safety, uh and responsible ai standards are that. I mean that you could include. All of this happens at the framework level, even the context. How will the agent understand context? How will it process the context? All of that goes at the framework level. So there is this fundamental difference and happy that you brought this question because a lot of times we end up confusing these two aspects.

Speaker 1: 4:40

So if we've got agents and we've got frameworks, can we then, or can businesses then, integrate agents into CRMs, back end databases, call logging systems? Because the real value is actually getting the agents up to intellectual speed with everything that relates to a particular customer or prospect so that they get a human-like, personalized experience that feels right.

Speaker 2: 5:08

Exactly. I mean, that is the version one of agents. I would say use case, the very popular use case, the world that we are in currently. For example, we had a customer who automated exit interviews HR exit interviews with a voice agent. Now the voice agent and the brains for the voice agent is built on Lyser, but it integrates into their existing HRMS platform to understand who are in the pipeline from an exit point of view and then the agent dials them up, introduce to them that it's an AI exit interview agent and it asks a lot of questions to the employee and collects feedback.

Speaker 2: 5:51

The best part is, since an AI is calling the employee lets go of the guard, they are able to give a lot more feedback and the agent is not biased, obviously, so they share a lot of inputs. The agent is able to understand a lot of it and then the agent analyzes, it, sends a report to the leadership team and the organization was able to immediately see the impact. They were able to get a lot more critical feedback which they could work on, compared to the previous model where a human called and asked these questions Because all of us we don't want to obviously hurt the feelings of the other person, so not everyone gives those critical feedbacks. This is a very good example of how an agent is necessarily not replacing a system. It actually comes in as a very good bolt-on feature, but it makes a huge impact already for the organization I suppose there's a question.

Speaker 1: 6:46

There isn't there, because unless an agent is authentic enough, then are you willing to engage it, but if it's too authentic, do you not trust it. You know, I run that intellectual argument in my head all the time, siva. You know, when I see some AI that looks too human, it's off-putting, but when I when it's not realistic enough, it's more comical. So I adore the fact that people are actually using the agent to answer questions to get some real exit interview material, because that alone can make the life of everybody else who's staying behind hopefully that little bit better.

Speaker 2: 7:24

I mean you're right, but that's the world that we are moving into, whether we like it or not. This is how the future looks like. That's the world that we are moving into, whether we like it or not. This is how the future looks like Now. For example, in customer service, we use AI chatbots. We use AI voice agents to answer queries, ai email agents to handle service tickets.

Speaker 2: 7:41

Now, if you're a customer, you are already frustrated with the fact that some software or device is not working. All you need is support and you don't mind coming in from an AI which is slightly dispassionate, which is obviously looking at just solving the problem for you. Now compare that with a HR exit interview. It could be a bit more emotional compared to customer service. So organizations take this call, like, for example, organizations with 10,000 plus employees. The exit interview agent makes a lot of sense for them because they're not able to effectively employ humans to do that. They miss out on a lot of critical feedback. But when it comes to customer service, this is becoming the norm now. An ai customer service is something even customers are liking, because they're able to ask questions in natural language and get the support almost instantly. So, yeah, it's kind of a balancing act to an extent, but organizations are actually moving forward really, really fast in terms of adoption.

Speaker 1: 8:43

Yeah, I'm seeing that myself all the time, which is why this is such a hot topic. But we always hear a lot of hype in industries when technology comes into play for the first time, and I've no doubt we will see the same thing when it comes to agentic AI. But agentic AI a bit like robotic process automation beforehand or intelligent automation or advanced process automation. It's not entirely just out of the box and simple simple if you don't know what you're doing. So, Siva, what are some of the common pitfalls organizations encounter when they're trying to build agents or they're trying to put agentics or visual AI solutions into their actual business workflows? Yeah, good question.

Speaker 2: 9:28

So, the way we look at it as you, I mean great that you brought the robotic process automation into picture, because agentic automation, as we call it, is the, is the next, uh, evolution, with generative ai, I mean, and then ai powered agents that have come into picture so how is it fundamentally different now?

Speaker 2: 9:48

when it comes to robotic process automation or any other form of automation that we've had so far, they're all rule-based. To some extent, we used predictive AI so that the outcome could be measured, but with generative AI, you have a new logic engine that is being built. So, like Satya Nadella, microsoft CEO, recently mentioned the future of, I would say, knowledge will not be in the databases, it will be in this logic layer which these agents actually form. So a good example is a large conglomerate of a customer of ours built a merger and acquisition analysis agent on our platform. So this organization grows inorganically. Their main focus is to acquire companies who are in their industries, so they, in an year, they actually analyze at least 5 000 companies to acquire. They have several evaluation criteria, and then they had to narrow down to the set of companies where they want to do a due diligence, and that itself is a huge number. By the way, they at least do due diligence for hundreds of companies.

Speaker 2: 10:57

So they had this agent, but they could not do this with rpa, unfortunately, or even predictive ai or writing hard code. So what they did? They built agents that could evaluate an organization based on multiple parameters and with reasoning and given output. So the reasoning part is the new, I would say, capability that we never had as a software industry. So why enterprises are excited about this new technology? Because they now have a low cost reasoning engine. That cost just sends it to even reason and take a call, whether they should acquire a company or not. So that is the logic layer that I'm talking about, that these agents are powering today.

Speaker 1: 11:42

Because that's the interesting piece, isn't it? Because you know, I adore RPA. There's absolutely nothing wrong with it and I've seen many a time individuals talk about the death of RPA, but it's one of those evolving automation technologies that's now a platform and a of it being rather brittle, rather, you know, predetermined or deterministic in path, they can start to work out and question and understand context and everything else. So it is exciting. What are the essential components of a successful agentic strategy to get a return on investment then? Good question a return on investment then?

Speaker 2: 12:26

Good question. Offlit because we have generative AI, because chat GPT is so popular. All of us have tasted chat GPT to a large extent, but agentic is slightly different. You could do certain things with chat GPT but with an agent AI agent you could do a lot more because with the API, you could have your own version of like gen AI use case. You could do a lot more because with the API, you could have your own version of like Gen AI use case that you could automate.

Speaker 2: 12:52

Now coming back to how do we approach? I mean, how do we approach with customers? When customers have an idea that they would want to automate something, we first try and break it down step by step. That's the most important aspect of getting started. So once we define the entire process, we then figure out which of these processes could be automated with agents. A simple way to do is just go and ask ChatGPT again, just highlight all these tasks, ask it to break it down first of all, whatever process that you want to automate. Then take an individual task and ask ChaiGPT if generative AI could solve it. That is a very easier way to do it.

Speaker 2: 13:32

At Leiser we have AI agent consultants. Obviously, we know a lot more because having built 500 plus agents. But if you are planning to automate something with agents, that's the best thing to get started. Break it down, analyze each task and figure out what all could be automated. Tasks that cannot be automated will have to be solved by coding. We need to have hard code, like, I would say, functions that we would need to write to bridge the gap. So that is how we see the agent automation happening. It's a combination of these generative, ai powered agents and the combination of some hard-coded logic, which is a function or a code that sits and the whole workflow gets orchestrated.

Speaker 1: 14:15

Yeah, so what's the future of AI and the GenTech AI that you're seeing what's going to happen in the next 12, 24, 36 months?

Speaker 2: 14:24

Well, actually in the AI agent space or generative AI. There's a massive timeline. To be honest, we've been hearing this hype around AGI and super AI et cetera, so let me share my view of it. 2025 will be still the penetration of AI agents into enterprises probably will still be 2% to 3% max. I don't think we're going to see a lot more adoption. I mean to give you an example.

Speaker 2: 14:57

We were speaking with one of the world's largest sport drink company energy drink sport drink. They have a company-wide ban for using AI. They have some instance of chat GPD internally, but otherwise they don't use AI today because they're still figuring out with the European laws around AI, with the latest US laws as well on AI, so they're still figuring out. How do we get started? They don't even use it for internal purposes. So hence, from an enterprise adoption, 2025 will still be very low, single digit numbers. But the SMB adoption is skyrocketing. Already we have customers using AI SDRs to automate sales outreach. They're using AI marketeers to automate blog writing and social media writing. So the smb adoption is very high because they feel the pinch of cost and they need high productivity at low cost.

Speaker 2: 15:55

So this is how we see 2025 is going to be but 2026 and 2027, I think the organizations will start shifting towards a 20 30 percent% of their workforce powered by agents. For example, in my scenario, I mean, within our company, we have at least 10 agents doing their job. We have one HR and she is supported by HR agents actually, and same goes with sales and marketing teams as well very lean teams, but there are a lot of agents doing their job and individually as a person, I have about 12 to 14 agents. That starts the day with me every single day, including email writing, proposal building, nda, I would say, reviews, et cetera. So I think that is how 2025 and 2026 are going to look like the early adoption for simpler use cases. But when we go to 2027, I can easily say that 30 percentage of an organization will be run by AI, the customer support function and many other HR functions and so on.

Speaker 1: 16:58

They probably, as SMBs, are discovering they'll need it. I mean, I see, you know, particularly in Western Europe you know lower productivity figures, lower workforce participation, you know an aging workforce we'll never be able to get all the things done that we need to get done inside of a business. So without AI or some help, I reckon we're in a little bit of trouble in the next lot of years as a global economy. And it's always interesting because I see a lot of companies dare I use the word panicking or over worrying about EU legislation when you go. But we've had GDPR for years and this is an adjunct onto it, so you should have been aware of it and ready for it.

Speaker 1: 17:33

The same you know you can't deny a really great data strategy or actually fixing your work processes, because a genetic, a bit like RPA, won't work in a vacuum either. So I'm always surprised at how companies are surprised with where the world is actually moving. But for those people who don't know your company, where could they find you? Where could they find out a little bit more about it? To learn more about agentics and maybe just have a conversation as to how they could put agentic AI into the business using a platform?

Speaker 2: 18:02

OK, we are almost two years old now as an organization, having served 500 plus customers. Several large fortune 100 companies are our customers now you could visit lyzerai l-y-z-r dot a-i or z-r dot a-i, um and uh, and we also have the studio. It's studio dot lyzer dot a-i. So visit and it's very easy to build your first agent. You would end up building it under a minute or max couple of minutes. We always advise to get started with something very simple. For example, email formatting agent.

Speaker 2: 18:40

What if you wanted to write an email to your client or to your coworker or your organization? Just write an email and ask build an agent that could review it and rewrite it in a better format. These are all simpler. I mean, like I said, chat GPT, a popular question that we get to say can't you do this with chat GPT? Absolutely yes, but chat GPT doesn't. I mean you need to give the context every time you go to chat GPT. But with agent they have them in memory, they remember and they're designed to do that one particular task.

Speaker 2: 19:11

So the best way to get started is something that you will use ChatGPT today. Start building agents for them and you will start liking the concept that you don't have to necessarily provide context every time. You're a ChatGPT engineer, you have these agents, which are readily launchable as apps available for you running with you. So email writing, blog writing, linkedin post writing these are all some basic uses to get started, and then you will start seeing patterns on okay, where all could I apply this? Can I build a reasoning engine, can I build a customer analysis engine, and so on. So that's how you can go to then level two and level three, and so on Fantastic.

Speaker 1: 19:53

I think people need to do that, they need to dip their toe in the water, because if they don't, we really are in a position of whatever you want to call it AI or digital Darwinism, where competitors, companies, everyone will start to use AI, and agentic AI in particular, because the benefits are too hard to argue with. Sibba thank you so much indeed for joining me today, and thanks to the audience for listening in. Until next time, I wish you every success.

Speaker 2: 20:17

Thanks, Kieran. Thanks for having me. It was fun talking to you.