Episode 7: How Agentic AI is Accelerating Software Development and Support

In this insightful interview, with Sri Krishna Ravulapalli we expose the transformative potential of agentic AI in customer support and software development. 

Discover how automation, proactive support, and AI-driven processes are revolutionizing industries and enhancing customer satisfaction.

Key Topics:

• Definition and evolution of agentic AI
• Impact of automation on customer support
• Proactive vs reactive AI support
• The role of human in the loop in AI systems
• Future developments in agentic AI and enterprise applications
  
  Sound Bites
• "AI agents can support 24/7 without fatigue"
• "AI is already here, don't lag behind"

"Embrace AI and explore its potential in your daily activities"

Chapters
01:25 Krishna's Journey into AI
05:22 Understanding Agentic AI
08:00 Proactive vs Reactive Support
12:25 Customer Satisfaction and Agentic AI
15:13 The Future of AI and Agentic Solutions
20:18 Modernizing Mainframe Systems with AI
23:00 Final Thoughts on Embracing AI

Resources:

Krishna Ravupali on LinkedIn

Playbook AI Partners

Signature Playbook Series Course - Bonus 75% Off
Community: The Huddle Community - Live Training Sessions - Free March and April
AI Platforms - Making The Right Choice - Free

#podmatch

Transcript

SPEAKER_03 0:00

AI, Agentic AI, the acronyms just keep coming. As we work to navigate this fast-moving AI landscape, one question rises to the top. What is Agentic AI and how can it create meaningful impact in our businesses and everyday lives?

SPEAKER_02 0:22

Welcome to Playbook AI Partners Podcast, the show that turns AI from overwhelming into actionable.

SPEAKER_00 0:29

Your host, Daddy Kibling, is the chief playmaker, helping business owners and teams stop chasing shiny tools and start using AI in a way that actually moves the numbers, saving time, reducing busy work, and driving growth.

SPEAKER_02 0:42

Each episode features industry experts and real-world tactics, providing clear to the snack, plays, simple AI workflows, and practical guard reps. So you can use AI safely, execute with confidence, and get results. Let's run the playoffs.

SPEAKER_03 1:18

But what is agentic AI really and what does the future hold? To help us get into the topic, I have Krishna Ravupali on the show. Krishna is an AI and data enthusiast. Krishna is an engineering leader with over 18 years of experience in an Agentic AI, AI ops, data platforms, and enterprise cloud applications. He's also a certified Google Cloud Professional Architect and Data Engine with a Master of Technology from IIT Delhi. Krishna now advances agentic AI to transform software maintenance and support. Welcome to the show, Krishna.

SPEAKER_01 1:54

Thank you, Sandy. Thank you for having me.

SPEAKER_03 1:57

You're so welcome. And you have some vast experience. So I was hoping you could start out by telling us about your journey into AI and what you're doing today.

SPEAKER_01 2:06

Sure. Um I have around 18 years of product development experience. I started my career in India and then later moved to the United States. Okay. So during my initial days of my career, I have seen so much of manual work being done in terms of uh production support, especially. Let's say you have a product released into the market by a company. So and uh the users are using it, and if they find any bug or issue with the product, they call the customer care and they start complaining about what's the issue they have. And the person who is sitting and taking the call for that product, he will manually take all the details and he will be asking the users about what is the exact issue, when was it reproduced, and there are a lot of manual stuff over there. I have spent days working on uh such production support as well during my early career. Once we develop a product in software, it gets shipped to the customer end, and there they start using it, and if they find any issues, they call to the production support. And production support takes all these details manually and passes to us, and we look into it and fix it and then send it back to the customer. So there is a lot of manual uh intervention happening in terms of resolving an issue faced by the customer. So due to this manual intervention, there is a lot of time taking in terms of resolving the issue, and that in turn affects the customer satisfaction. So as we are moving towards this agent TKI, so initially there are some workflows developed in terms of collecting the data from the customer that about the reported issue and all. So everything started with uh some kind of automation, and this automation is eventually led to this agent TKI, what we are seeing right now. That's how the agent AKI has come into existence. I mean whenever there are any manual steps and those are repetitive and those keep on happening again and again, and the humans are spending a lot of time in doing that, we can automate it. First, you can automate it, and once uh that automation is going fine, you can make it as an agent. So instead of human triggering the automation task, the agent itself can trigger whenever it is needed and it can take decisions like humans and we can automate that entire uh flow. That way the customer gets the results in the and the fixes for the bugs whatever he is facing in a quicker and faster way than what it was earlier, and also he can have much customer satisfaction and he can stick to the product, and as long as the customer is stick to the product, the company seller.

SPEAKER_03 5:06

So I love it. You had me with pain because I don't think there's anybody listening that hasn't been in that situation with the manual process, repeating the issue, the ticket number that's probably you know 26 letters long, right? That you have to write down and the time. And I'm not trying to be condescending of those that do try to help, but it is a very labor-intensive process. So I wanted to break down a few things that you said um before we get to that. I I do have a few questions, but how would you define a agentic AI? How would you define that for listeners that may not know what that is?

SPEAKER_01 5:42

Agentic AI is a simple term, like uh a single responsibility action which can be performed by a mission with no human intervention. That is called an agent, and agentic AI is making all these smaller agents working together is called agentic AI.

SPEAKER_03 6:00

Okay, well, thank you for that. So by using the agentic AI, then you're now taking that almost it sounds like, let me know if I'm wrong, like maybe 10 steps out of the manual process, consolidating that into fewer steps, getting a faster response, and going back to that customer satisfaction. So if I have that right.

SPEAKER_01 6:20

Yeah, you are right. I mean uh to add more each agent does one single action and what it knows. I mean, uh in terms of uh handling with customer support issues, there are th uh there can be multiple tasks. One task is taking the phone call that can be automated, and one agent is always responsible in taking the phone calls, and another agent is responsible in collecting the diagnostics information happening at the customer environment. And another agent is sitting at the product company which listens to the agent at the customer and passing the information. So each agent does one action and orchestrating all these uh agents working together for a meaningful output is called agentique.

SPEAKER_03 7:12

Okay, that's no that's awesome. So I don't know if you have a statistic on this. What would you say takes the time frame from the manual process versus the agentic process? Is it is it reducing getting to a resolution in days or in hours? What would you say that is if you're able to answer that?

SPEAKER_01 7:30

Yeah. Uh earlier in terms of manual intervention and assisting the customers and all, it was taking days. I mean uh it is definitely taking days because uh it's all uh reactive support. With agent reactive support in the sense when something issue happens and seen by the customer, then he reports and then the actual uh steps to fix that issue starts happening. Now with the agent TKI, it's all proactive. Proactive in the sense before customer sees it, agent can sense it based on the diagnostics information present at the customer environment. So now we are transitioning from reactive to proactive support where the agent can detect if there is any error or alert hap happens at the customer environment, it immediately informs the agent sitting at the sitting at the product. So they can talk and collect the diagnostics and fix the issue. Sometimes it can fix even before the customer realizes there is an issue.

SPEAKER_03 8:34

Oh, that's cool. So when you say agent, I guess I'm kind of getting confused. So are you talking about the agentic AI process is all the AI components collecting that data, or is it a combination of the agentic AI collecting the data and then also talking to a human in the loop to verify is it is it a combination of both?

SPEAKER_01 8:55

Yeah. Agentic AI, I'm talking about agentic AI. It's a fully agentic AI. There is no human in the loop. That is what we are envisioning right now. Currently, there are humans in the loop. I mean, to make sure the agents are doing the right actions or not.

SPEAKER_03 9:12

Okay. Oh, but that makes sense. So if someone puts it in place, you you so the agent, a agentic AI person has been set up and programmed to handle, you know, if this, then this scenario. And so it sounds like the human in the loop is just to kind of verify did it get it right, and if not, how do we fix it? So if I have that right. Okay.

SPEAKER_01 9:34

As the human sits in the loop and makes sure the agent is performing the right actions. That way he guides the agent. Yes, you are doing right this way. I mean, the agent keeps on learning the feedback given by the human. That way the agent can also improve. It's like you can treat an agent like a new junior engineer getting onboarded onto the system. And the junior engineer most of the times he takes the feedback from the senior engineers and keeps on improving by himself. And if he does a mistake, he learns from it and keeps on correcting those mistakes. You can see these agents similar to a junior engineer, and they are kept on evolving. And once they are matured enough and they're kept on doing the right things, they are good to go and perform with no human in the loop.

SPEAKER_03 10:21

Wow, that's that's pretty that's pretty awesome. I recently called to uh get my car, had a recall notice, so I had to get it into the dealership. I think I dealt with an a agentic agent in the first maybe three steps, and then I was transferred. So it's also, you know, use the customer service scenario, but I'm beginning to see more businesses kind of use it in small steps, like taking that first call, routing that call, and then and then maybe speaking to a light person. So it seems like there's various versions of that out there.

SPEAKER_01 10:52

Right. You got it right. You have experience as one of the agents.

SPEAKER_03 10:55

What do you think the accuracy rate is? Is there just a lot of you mentioned customer satisfaction. Is there a good satisfaction rate? Are people more frustrated, or are people embracing agentic AI and saying, hey, I'm getting my problem solved quicker, less less time on the phone, better resolution? I was just curious about that.

SPEAKER_01 11:14

Yeah, right. Everything needs to be measured to know if it is working good or is it beneficial to us? Like the same way for the agents in terms of production support, the uh the key metric is customer satisfaction. So how how much the customer is satisfied, that much our agents are playing a key role in terms of uh resolving the customer issues and getting good customer satisfaction survey. So when does the customer get happy when he when all of his issues are resolved quickly, and uh then then he'll get happy. So the key metric is um how fast we are resolving the issues by empowering the agents to take a couple of manual actions. So that's the key metric we had to look at.

SPEAKER_03 12:02

We like happy customers for sure. But do you know today, I mean, of the of the agentic AI solutions that you've maybe worked on or put in place, are you seeing there's a a better success rate where the customer's happy on the ad with the agentic a agentic AI versus the manual process? Or is it still even? Or I was just curious how that played out.

SPEAKER_01 12:24

Yeah, um whenever I see the customers are seeing good satisfaction survey. The primary reason is uh if it's a human taking all those actions, there is a limit for the human human. He can work for eight hours per day. And uh after hours, there is no one to handle the customer issues. If it's an agent, it can handle throughout uh the day and night times. That is one key responsibility of an agent to take care of the actions throughout uh both day and night times.

SPEAKER_03 12:56

Yeah, no, that's a good point too, so you can get that help whenever you need. Because that is a frustration, especially with time zones, and you're like, okay, I have to wait until tomorrow. So that's really a good point. So tell us more about you. What is your motivation behind exploring agentic AI solutions?

SPEAKER_01 13:13

Yeah, I started looking into it when I exposed it to Google Cloud environment. And I have been going through all the trainings and uh listening to the latest developments happening in the Google Cloud in my previous organizations and the current organizations. So when I I see the new releases happening every six months by the Google Cloud, I I was so excited because uh the amount the agents are taking and evolving, and the amount uh the contribution of agents we see in the current environments. Right now, if you see the news, the agents are itself uh writing the code if you as a problem scenario. So they are evolving there and they will eventually mature. But right now it can code and help the engineers in terms of uh developing the code as well. So a lot of development is happening in terms of uh coding assistance, assisting the software developers in terms of writing code and developing the products. Now the future is about uh the production support as well. So once the product now the rate of uh developing the products is increased with the coding assistance. So the number of products come into the market will also increase. So to support all these products, we need uh more a agents as well in terms of supporting the products, not humans can support all the products always.

SPEAKER_03 14:40

Yeah. So it's such a so excited, at least I get excited about AI because when I look at what tools can do and what whether it's agentic AI or new tools that are out there, I mean things have just changed. You know, the the time to create an app or the time to create a website. You know, it used to be uh I I was talking to a friend and he was talking about building at like an app. And it would, oh, I it took him months and it but working through bugs, but AI has just really changed the time even to production for building an app, building a website, even taking out some of the coding and that too. It's just so exciting of the things that you can do. So, what do you tell people? Just, you know, there are those of us that are excited about it, and there are those, you know, that are saying, I I'm not not so much, I'm not excited. What would you if you had to tell people, you know, to be hopeful or excited about the future of what AI could bring, what would you say?

SPEAKER_01 15:34

I would say you should be excited to the the current developments happening in the AI. If you are not, then you are lagging behind in the market. So if you don't upgrade, you will lag behind the engineers or the people who are getting acquittained with the AI agents or the AI models. So the message I want to convey is it's not something new that is coming up. It's already existing and you it's the time for you to learn and uh try to make use of it in your day-to-day work. If you don't use, you will lag behind.

SPEAKER_03 16:10

No, that is so true. And I will say there was a time when I was one of those people, so I just want to admit it, right? That I was just thinking, oh my goodness, and just waiting, it's another news story, it's just hype. But once I really sat down and learned and, you know, worked with a community and a network of people and saw the power in my own life, I give an example. I by using um I use Chad GPT a lot, perplexity, but I created my own navigator, some call it a clone, but to really help AI understand my business. I have several businesses, but so that when I put you know a prompt in in the in a particular model, I've taught it about my business by putting in files and whatnot, so that it it is more authentic to my voice. But all that said, um, I really think that's really helped me to get a more refined prompt. But it my point in embracing AI is it has really helped like shave off honestly 15 to 20 hours a week in terms of my newsletters, my research, my follow-up with clients. So, you know, never to sound preachy, but I I love what you said. Just embrace it and just try it out and see how it can work in your own business. Now, Krishna, what are you doing today to kind of help people with with AI if they're on this journey? What what are you doing to to help people and and the world of AI?

SPEAKER_01 17:35

I work uh in a software organization in Plano and we are make we are modernizing the traditional mainframe systems currently. You know about mainframe, it's there in the market for decades, and it's still continuing to be there because right now we are trying to modernize it and uh making sure it we have AI capabilities enabled in the mainframe as well. Right now AI is available in most of the cloud environments and the distributed space. You want to bring the same expertise and exposure to our customers who are still on the mainframe system?

SPEAKER_03 18:12

Wow. Now, mainframe, now that is old. Let's just say that. I mean, older than probably you and I, but but mainframes, is it uh but a lot of data is stored on there. And I mean, are mainframes, is it taking AI to remove that data off the mainframe onto maybe the cloud, or is it keeping the mainframes but just modernizing it so that the data or anything that's stored on there can still be used?

SPEAKER_01 18:37

Yeah, it it's not possible to move everything out from mainframe and keep it somewhere.

SPEAKER_03 18:43

Okay.

SPEAKER_01 18:44

Though mainframe is old, it has its own uh positives and it can still perform a good transaction rate compared to the existing uh distributed environments as well. So we want to we don't want mainframe to be isolated. We want mainframe to coexist along with the modern landscaping uh in terms of uh uh cloud environments. So we want both mainframe and cloud exist together, and whatever the strengths of mainframe, we can make use of it along with the existing modern cloud environments.

SPEAKER_03 19:18

Well, that is really interesting, and I didn't mean to be insulting the mainframes. I don't know anything about them other than I, you know, I haven't heard that term in a while. But uh what I love about that is to your point, I mean, not everything can be transferred to the cloud like the photos on your phone, right? It's so interesting to take, you know, a mainframe and be able to use AI to still make that work and make that connection. And again, that's where I get excited about AI because there's so many cool things that can be done. So I appreciate all that you're doing. This has been such an enlightening conversation. But as we draw to a close, I mean what are some kind of final tips or thoughts that you would give people listening to about AI?

SPEAKER_01 19:57

Yeah, I I would suggest to all the listeners, don't think uh that AI is something that will come later. AI is already in place and it is around all of you. So embrace AI and get used to AI. Keep uh exploring where you can make use of the AI in your day-to-day activities and can save time and can deliver your outputs in a much faster way.

SPEAKER_03 20:21

Amen. I agree with you there. I've lived it. So that's awesome. Well, I want to say thank you for all that you do. I mean, the work that you're doing is is really very, very interesting and certainly very important. So thank you for all that. And I appreciate you taking the time out to provide your perspective and your expertise today. It's been very enlightening. So thank you for that.

SPEAKER_01 20:40

Thank you, Sandy.

SPEAKER_03 20:42

I enjoyed having Krishna on the show. What a wealth of information. I really appreciated the examples he provided to help simplify agentic AI. I hope it was helpful for you. In our next episode, I have Ben Taylor on the show. Are you still experimenting with those AI tools, just trying to figure it all out? You hear about workflows, automations, and prompt engineering, but how does it all fit together and why does it matter? Well, Ben is the founder of Nexaserve and AI systems company that builds custom AI agents and large-scale workflow automations designed around how businesses actually operate. We're going to talk with Ben about those automations and how to prevent automations from creating more complexity instead of less. Until next time, take action, execute, and let's run the play.