Refining ITOps with Generative AI – Muddu Sudhakar, Aisera

This is texturing TV. Hey everyone, welcome back to Tech strong TV. You know, you can't throw a stick without hitting something about AI in today's Tech World.

It actually not just in the tech world even in the mainstream media. So I'm really excited for our next guest here and we're gonna talk a little AI stuff AI Ops as well. Let me introduce you to modus pseudocar.

Moodoo is the CEO and co-founder of a company called Acera. That's a i s e r a voodoo. First of all welcome and thanks for coming on.

com. First of all, thank you for having me. It's pleasure to thank you.

It's my privilege. Thank you. It's our pleasure to have you on so Voodoo.

Let's start with you. Let's start with a little bit about your personal background. If you don't mind you are the CEO and co-founder, but give us a little bit of your history.

Yeah, I mean I can go quite bit but let's to keep it short we started isera around 2017-18. I left switched out to start isra and at service now I ran the atarman itsm for a while right before that. I was at Splunk running cyber security there.

So sure. I know a little bit of AI and little bit of Ops and it has some business. You're being humble.

But yes, those are and those are two Powerhouse companies right service now and spunk. What tell us about I Sarah what's the mission here? Yeah, so look, I'm Alan.

I'll tell you so back in 2016 15 17 when I was looking at this space, right? The reason we start Iceland fundamentally looking at this now space and looking at the customers the problem was and these before Chad GPT. But what about before gender to AI my vision back then was how can we automate most mundane task people do right across employees users customer service operations right back to the whole idea of how can we automate things so that people can do more enjoy their lives more enjoy their more important items that doing the repeater task.

That was my mission. I really wanted the contact center customer service agent. It administrators to do more high value items not focus on it.

Like world doesn't need million devops tuners. Like if you have a problem you shouldn't be watching a dashboard or tell you I don't believe in this anomaly detect Shan and trying to do it give you millions of alerts. Nobody have the time to go through that.

I don't need it. I don't think that's what I want you to be a predictive in general. Tell you Alan today.

You should do this with high accuracy. If I fail you next time, I'm not going to get you as a user right. So that was the mission for us and say look what can we shift in the entire Society so that we can move everybody else.

Up in the food chain. Absolutely. So what do you you know you zero right in on the bullseye?

Generative AI promises to be a disruptive Game Changer across multiple multiple Industries. I think an easy one is the the it up space right because this is an industry that's been waiting for this. For as long as I can remember, right we yeah service orientated architectures and every other ml lops.

We've had so many different. You know trying to break the pinata or pin the tail on the donkey, but it's meant but it's missed. What has generated AI going to really be the game changer?

We think it can be. A lot of look we're just feeling the onion right I call this is the AI Generation generation AI right. It's a decade.

We're just starting now Allen. So anything that you and me can think people will come even more better ideas in the next 10 years. So I think two things fundamentally is happening is what charge GPS has done is it's consumerized the whole AI experience.

It's also made it probabilistic. It made it random. It made it fun.

So even if it gives you some hallucinate answer people are not being put off on it, right like humans are going to be wrong. You're not going to fire an employee because you've made a mistake you don't have fire and devops person because he couldn't detect it. Right same thing is going to happen is with AI the bar was much higher but he has like, oh, I want you to be 100% Correct?

I have even customers ask me. Oh does your system will give you answers always correct? I'm not God.

Nobody's going to give you 100% answers. Nobody can tell you 100% what incidents are going to happen or what outages will happen performance issues with 100% accuracy. So given that what Attributing generative has done is made this such a pragmatic experience now where you can use this is to determine apply NLP and genitway and chat GPT Technologies in the IT world in the operations world because time series data alone is not important to your point.

Ml Ops aarops doing time series and ability detection is one approach companies have done it for last 10 years now with generative and charge GPT. Now people are going to apply like humans can look into your logs events traces to predict. When will you have an outage right?

Why will you have an outdate give your root cause right. Do you have a performance issue with a high degree of accuracy and give you also the root cause it's almost going to be like a copilot. So the airbs will become a copilot for your operations.

That's where the energy is going to be. I I agree with you a hundred percent on this. And I I you know what?

You said something that it's kind of like you you put your finger on it. That's bothered me. So many of my friends in Tech.

It's just their nature right? I don't hold it against a personally, but it's their nature like they're so quick to point out. Oh, look it got my biography wrong.

I asked them who I was and it said I worked over here. I never worked there. I just wrote for them.

But I never worked there right come on, you know what? I mean? Realistically speaking.

We are just scratching the surface. This is like a big baby that still needs to go to school in many ways. Listen, but but you could see where the potential is you see where the future is, right?

And and I think people who try to deny that by pointing out that it's not a hundred percent, right? You know, they're they're the people said I'm not sure the Earth is round or I don't know if we'll ever get to the moon. That's right or stuff like that.

So it's I I think it's ridiculous but good. No, I think you're you nearly trades in you and me know this but most people that I know of cannot do what you're saying, right? They expect the systems to be hundred percent, correct?

So somewhere sometimes when we buy a tool and a solution for it and the body is high then a human. So what I end up trading guys, it's going to make mistakes like like a kid, right if you have a son or a daughter they're going to math class. They're going to make mistakes, but they're going to learn give a chance don't throw the product or Air Technologies too soon and people want the answers like well, I will do a proof of concept in two months and three months I want to do you think you're high and apply for three months if he doesn't do a good job, you're letting go I mean why why this concept so I think all the people I see in it.

They're like I want to be using for one month two months. And if you don't predict it, I'm gonna really I mean you're setting yourself for a failure and you're justifying yourself why you don't want her to so that's the challenge that Alan that you and me should Industries. Give it some time the invest with it stay with it be persistent if when there is a will there's a way you if you believe in it stay with it that which are wonder you pick be consistent stay with that person invest in a Ops and go along with because you need a co-pilot for your operations because this is not a problem.

You can solve by throwing humans at the problem because it's a non-linear data is nonlinear data is going to grow so big you can't derive Trends over the data with throwing humans of the problem. I I agree with you. You know what?

We saw this in the devops journey. right It's not that. It's not that humans are bad.

It's just that the speed we're working at at the scales. We're working at. This she's not enough humans to do that.

We don't you don't have that and and we just don't work that fast as automation does so if we truly want to Go with the speed of business. There is no choice now. Let me just play Devil's Advocate a little bit with you module if it's okay.

there are some there are some Mission critical. Mission critical shins that way we can't afford a mistake right nuclear power plants or you know, something like this, right? When when do you in your mind for see us being able to?

Turn those kinds of things over to AI or maybe there's things that we do and things that we never turn over to AI. No, very good question. Look even for Mission critical system, right?

Like we work with a lot of Department of Defense in il5 and il-6 environments. Right? So what's happening is more than we think average person the government agencies already head of the curve, right?

See they're being attacked by Russians and Chinese and like, you know, the whole world is going after our Department of Defense. So these guys who have to protect their environment. So what they're doing with the cyber security in the tools is they're already looking at using AI to make a prediction see I would rather have a remember the world of the poor before you'll have true positives false positives through negatives false negatives.

I would rather the cool copilot. Tell me the inside saying you know, what there is some attack is going on this environment. I see a performance issue here are there's a low grade instant happening.

Even if there's a probability. Let me be the judge. Of the team Valencia me.

I am the devops administrator. I am the AFS admission. I'm looking at the data you tell me handful of key observability points that I can play with it.

So I think the even the highest order sense to infrastructure will start deploying this both in the it systems and OT systems. Right. That's what I'm talking about iot system because we thought that you just to your point you for the speed of business.

You really need humans cannot throwing humans around watching the dashboard will not help you you need to do non-linear analysis. You have to do NLP analysis. You have to do the derive the information out of the locks to predict what might happen because of the relationship graph analysis.

So good news is you have deep learning. You have Knowledge Graph use this in the a Apps World. So for a house, well is only time series analysis, right?

What should is like I call it like 80 versus yeah, we see right. It's it's before AI after a charge GPD and 2023 is like after AI right anything that happened before that you pretty much through it and start all over again. I love it.

I love this. So. If we say this is year zero or year one year one, right?

In the in the you know, post chat. GTO a post AI era right P AI When do you think we'll actually? Like when is it going to be built into a Sarah?

As like the main, you know as the engine if you will. Yeah, good questions either. Look, we're already I'll give you some proof points, right?

I started the company around 2018. Like we've been off five years into the journey. We have customers and a hops with 70% to 80% in six to nine months.

We can predict both performance issues incidents before they happen and give you an early warning earlier. We're talking about hurricane Stoners. Our early warning is 24 to 48 hours.

I'm not giving you two minute warning like earthquakes that I get in California right now. People want to know can you predict how much every morning are you giving me? And what is your accuracy is Right 70 80% is pretty good accuracy today before they happen.

I'm still miss some earthquakes. I may miss some insurance. But whatever I pretty I'm here pretty good accuracy.

I may have still false positives, right? So this whole concept I think it's there but it still will take to your point. Can you do this for services?

Can you do it for infrastructure? Can you do for apps can you do it for like the question? What data supposed to do?

It how long will the data sources are and can you take into account seasonality? Right things will be different on a weekend versus weekday. What's a night time holiday schedules like you got to take into all those new things and new application comes in.

How do you a new services not a false positive existing. So there are always a what I call the Corner cases. You have to work through it.

But I think it's already there to me. I would rather deploy a option 2023 today now then keep waiting for it because the longer you wait guess what happens you still need intubation period of a year or two. So there's if you're a good citizen of your company instead of investing that in and their employee you you should buy like you're buying a flood insurance.

You should buy an air Ops insurance. I call this an insurance. What is the tournament in insurance policy actually?

Actually lights are running should a great way of looking at it, right? You tell the lights of London to sell this. What do you think?

All right, I wouldn't be surprised if they start selling it because I think look I I think this like I said, This iterative AI what you do today? Iops is low hanging fruit. But I think it's gonna disrupt the insurance industry.

It's good. It has to it has to right. I mean, they you've got an industry there that runs an actual aerial tables, right?

Right. You start applying Ai, and that that's Child's Play for it, right but technology I was there in 2014-15 at Splunk when we were early days of cyber security Insurance. Nobody would sell a cyber security engine back then and I would go to all this insurance companies here guys cyber security is a big problem.

We did use a beer and it takes people now everybody wants to sell cyber security engine. So same thing will happen with a hops. People say look, let us tell you the copilot operation insurance.

If you don't have it how you your insurance policy will be high. It's like the good driver discount right if you don't have yeah, but that's right insurance, but that's what cyber insurance is doing right? It's forcing companies to raise their bar to to have ai apps to have good, you know cyber hygiene and so forth and in many cases the insurance companies almost becoming the security provider because they're the ones with the stick and sometimes you need to stick right?

Buddha we jumped into this. com is the website people can get more information there. How do they engage with the company though?

Like what what's on ramp? Yeah. I don't know.

I look first of all but being on the show Allen, we actually offer preposis tell I want your audience to look we give you a free assessment free Pilots. So we proof of concept we want to do that right as much as possible particularly with Chad GPT generator depending on the environment and the business case we offer that what's your qualified customer we want to do it. com get if the demo is also available on our web page.

You can experience out demos. You can feel it. Our customer case studies around our page.

Yeah customers like all the way from so to Grand thought and to Gap and work that we have a whole bunch of customers Quizlet, right? So I would love to have people to come and experience what we can offer look at other customers in case studies and happy to provide a proof of concept to prove our value to you guys. Absolutely, but do I want to thank you for coming on Textron TV today?

This has been great. That's the luck with that Sarah. Keep coming back.

We're good. Look AI Ops is not Ai and iterative Ai and AI Ops it's not going away. Let's keep the dialogue running come back and talk to us soon.

Okay. Thank you. And thanks for having me today my proud my pleasure Voodoo sooner car CEO co-founder Sarah here on Tech strong TV.

We're gonna take a break on text strong. We'll be back in a moment with some more guests and interviews news standby.