Organizations are increasingly adopting modern application delivery models to support accelerated time to market, but it has tipped over the traditional IT operations execution methods. Today, IT infrastructure must be provisioned almost instantaneously with minimal downtime to keep pace with continuous code deployment requirements and achieving it in a typical enterprise cloud environment can be extremely complicated. Thus, many enterprise IT operations leaders are looking to invest in automation and AI/ML to help build these capabilities. What is the most efficient and effective way of moving from traditional IT Ops to automated AI Ops?

Contributor

• • Siddharth Dhar, Executive Vice President & Global Head – Infrastructure Management Services, Hexaware

Transcript

Sanjog Aul [00:00:00]:
Welcome listeners. This is Sanjog Aul, your host and the topic for conversation is From Traditional IT Ops To Automated AI Ops. So organizations are adopting modern application delivery models increasingly to support accelerated time to market for their products and service updates to their customers. However, it has tipped over the traditional IT operations execution methods. To keep pace with continuous code deployment requirements, we must provision IT infrastructure almost instantaneously and that too with minimal or no downtime. Trying to do the same with a typical enterprise cloud environment can get extremely complicated. As a result, many enterprise IT operations leaders are looking at investing in automation along with AI and ML to build these capabilities now. So how can they move from traditional IT ops to automated AI ops in the quickest way possible? To discuss this, I have with me Siddharth Dhar. Sid is the Executive Vice President and Global Head for Infrastructure Management Services at Hexaware, a consulting firm focused on transforming IT solutions and solving complex business problems using a combination of human creativity and intellectual.

Sanjog Aul [00:01:21]:
Their three pronged strategy of automate everything, cloudify everything and transform customer experiences enables enterprises fast track into the digital era. Hello Sid, thank you for joining us.

Siddharth Dhar [00:01:34]:
Thank you for having me, Sanjog. I’m looking forward to our discussion today.

Sanjog Aul [00:01:37]:
Great, so my first question for you Sid, is why should a company invest in automation and AI ops to optimize their IT infrastructure and operations management efforts, in short, what business problem will this solve?

Siddharth Dhar [00:01:55]:
That’s a great question to begin with Sanjog. All too often in IT infrastructure, given the nature of our job and operations, we tend to not focus on the bigger picture of what business benefits are we driving. Actually, AI ops and automation really are geared towards business objectives more than anything else. To my mind, there are really four business problems that we hope solve by deploying something like this. The first one is time to market. Delivering infrastructure these days has to be done almost instantaneously. Over the last five years or so, most enterprise organizations have chosen to change their software lifecycle to agile from waterfall and that modern application delivery really needs speed in terms of delivering infrastructure and that speed can only be guaranteed by extreme automation and AI ops. So that’s a really good reason to believe.

Siddharth Dhar [00:02:47]:
Second, I would say reasonable is as organizations adopt more modern application delivery, they are looking to deploy code almost as soon as it is committed. So that from the point the code is committed to the point the code is deployed is one automated cycle which we call CICD. That by definition means that the underlying infrastructure platforms must be available all the time in order to accept that code. If you’re not, then you’ll actually reducing or impacting your release cycles. So again that’s business problem that we need to solve and again it is a complicated problem to solve, especially if organizations like most do have a hybrid private and public cloud. So that’s another problem that automation and AI helps us to solve.

Siddharth Dhar [00:03:35]:
A third issue, again very related to the first two is as organizations and enterprises are making plans or already having footprints in private as well as public cloud infrastructures, our security organizations are extremely worried. They’re worried whether we’ll be able to maintain our security stance, whether we will be able to identify drifts from our security stance and be able to correct them in product and that’s what automation does. We have systems that can deploy security guardrails as soon as any workload is deployed in your public or private footprint. We have monitoring capability to identify drift from your security compliance status and actually fix those in an automated fashion. So again, great reason to do that because without really making your security team comfortable, I can tell you very few CIOs will have the ability to adopt cloud at the pace that they need to. So again, helps you resolve a business issue. Finally, just given the amount of automation and the amount of artificial intelligence and machine learning that we deploy, your cost per unit for operations delivery is also impacted down.

Siddharth Dhar [00:04:44]:
So which means you’re able to optimize your support cost. That’s a good reason as any as I can think of. That’s like almost a cherry on the top. So really those are four reasons why somebody should do it and quite frankly all four reasons can be directly you can draw a straight line to a business problem that it solves.

Sanjog Aul [00:05:03]:
That’s a lot of good promise from AI ops and automation. So next question is what all are the prerequisites for an organization to exploit this automation and AI ops in their environment and what are the pitfalls and gotchas?

Siddharth Dhar [00:05:19]:
Great question again Sanjog. I think any deployment of this nature is complex. This one especially so because the footprint of change is across almost every aspect of IT operations. Anytime you have a change that impact such a large footprint of your IT organization, you can bet you will have challenges of execution. You may have technical and technology challenges, you probably will have people challenges, and you will certainly have political challenges. The number one prerequisite in my mind is management, buy in and sponsorship because that is what you will count on when you hit those roadblocks and I promise you, you will hit those roadblocks. The second I think important prerequisite really is you have to have a mentality to embrace change.

Siddharth Dhar [00:06:02]:
Anything that has such a large footprint of change across the organization and in this case there are really three dimensions that we are looking to change. We are looking to change the underlying technology to make it more modern. We are looking to change the skills of the people. Most infra and operations organizations have very few or very little coding capability and this is all about coding. So you have to reskill and retool your people and finally the old organizational structures of technology based organizations or operations versus engineering, some of those will get collapsed when you deploy them. So you have to be ready to embrace significant change across your organization

Siddharth Dhar [00:06:40]:
and sometimes I’ve seen all of organizations underestimate the quantum of that change and then struggle with it. The third prerequisite really is you have to invest time upfront in defining what your goals from this is. You have to be very clear of why you’re doing it and if you can’t draw a straight line to a business benefit, then you’re probably getting mired in cost based or cost related goals which quite frankly are not the right goals. Cost is a consequence of doing this, not the reason to do this. Finally, if I have to give you word of caution on what would be a gotcha, really the one gotcha Sanjog that I want to talk about, we obviously don’t have time to go into a lot of detail but really the one gotcha is there is no magic bullet. The quantum of change, and I just described it to you, is pretty significant across the organization and across dimensions. There are companies out there that will say hey, this is a problem that my tool will solve or this is a problem I can solve,

Siddharth Dhar [00:07:36]:
outsource your operations to me and I will build this for you. There is no alternative but for you to work hard at it. A lot of the change you have to drive is internal. The technology and the deployment of technology is the easy bit that is not the hard bit. So please, if there is one thing to avoid, the one gotcha to avoid is don’t think of this as a magic bullet. You’re going to work hard to get there.

Sanjog Aul [00:08:01]:
Let’s take a quick break listeners. When we come back, Sid would be great to learn how successful these automation and AI ops investments have been for the organizations and what benefits did they realize? Please stay tuned listeners. We’ll be right back.

Sanjog Aul [00:09:09]:
Welcome back. So based on your experience Sid, how successful these investments related to automation and AI ops have been for the organizations who tried it and have they truly been able to realize the benefits it was intended for?

Siddharth Dhar [00:09:27]:
There are no guarantees in life Sanjog. You have to have commitment to an endeavor of this nature for it to deliver success. I will give you three examples of my recent past engagements where we’ve seen a great amount of success and that’s been something that has been recognized internally in those organizations as well as externally by the market as well. The first example I’ll give you is of a healthcare insurance provider. They’re one of the largest ones in our country. So this company had been grappling with should they move to public cloud, will they be able to deploy public cloud in a secure way while they’re grappling with that? One thing was clear to them, that their current internal in house data center had to be upgraded to behave more like a private cloud. More software defined if you may. So they did that process last year and in that process what we did was we deployed AI ops and automation inherently inbuilt into the new private cloud data center.

Siddharth Dhar [00:10:21]:
What we have been able to do as a consequence of that is pre deployment of this new private cloud infrastructure. They will take anywhere from 27 days to more than a month to deploy the environment that their application teams needed to code. Today the deployment of an entire environment takes less than a day and mind you, I’m not talking about just a VM or maybe just the network address. Entire dev environment including deploying the three tiered architecture of the application that they have to code on is done in less than a day. So that’s a great result. You’re right there, you’ve saved almost 26 days of time and you’re enabling your business to to make releases that much faster. The second example I’ll give you is of a large Europe based bank.

Siddharth Dhar [00:11:10]:
They are very committed to automation. They in fact have had automation program very successful one for the past three years or so. One of their most successful automation deployments has been to automate their end to end patching process. Now this is an organization which has more than 50,000 servers globally and they are now able to patch each and every one of them across the globe in less than 8 hours flat. Think about just over a year ago we had the heartbleed zero day vulnerability. I know of organizations that have taken days if not weeks to patch all of their estates to get rid of that vulnerability and an organization of their size and scale of 50,000 plus servers can now do it in 8 hours flat. Again, a great outcome for them.

Siddharth Dhar [00:11:58]:
So the third and the final example I’ll talk about is another healthcare company based here in Illinois. It’s a mid sized company. We did a full AI ops and automation platform deployment for them last year and one of the biggest benefits that they have seen, almost a 50% reduction in their meantime to detect or identify a seven issue today and seven issues as we know lead to business loss eventually. In a traditional setup, the way you identify the root cause of a seburn is to open a bridge, have multiple teams in there, and by a process of elimination you come down and zero in on what the actual root cause is. Today in this organization, even as the seven incident is happening, the system itself is intelligent enough to give us the root cause of visual so that we can then solely focus our time on fixing it. In fact, this year we are going to even attempt to automate the actual fixing as well. So in a way you get to auto healing from it.

Siddharth Dhar [00:12:54]:
These are all great results. These are results that allow deployment of infrastructure quickly, these are results that allow allow your security postures to be better and these are results that allow you to keep your downtime minimized and these are real life examples that I’ve seen that just in the past year.

Sanjog Aul [00:13:13]:
So what holds back companies from being able to fully exploit this automated AI ops and what are the challenges and the suggested remedies you would like to present to the audience?

Siddharth Dhar [00:13:29]:
Sanjog, I’m actually going to hark back to the previous question around what the prerequisites are, because to me it is not getting those prerequisites right that eventually holds back organizations from getting this right. The number one thing is most organizations that I have seen struggle with this, they underestimate the quantum of change, the footprint of the change and the dimensions of the change across technology, people and our structure and that is something I would say really hold people back from realizing the true benefits of this program. The second reason I see all the time, and I spoke to it a little earlier as well, is not getting your success criteria right. If you’re overly focused on the cost benefits of doing something like this. Cost of missing out the actual business benefits that the business is looking from you from something like this, then you’re likely to fail or even if you’re likely to succeed, you’re likely to succeed only in that dimension, not truly in impacting what the business needs from you. The last thing I will say is organizations have to take a pragmatic approach.

Siddharth Dhar [00:14:33]:
The quantum of change is significant. Therefore, you need to give yourself time and you need to give yourself investment and to be sure those investments will pay you back fairly quickly, but you have to give an upfront time and investment allocation for program like this in order to succeed. So I would say those are things that I generally see and things that can potentially be avoided early on in the program.

Sanjog Aul [00:14:58]:
So Sid, it looks like that automation and AI ops can be a very involved effort and very possible. An organization may require a partner. So as part of doing due diligence and asking the tough but fair questions in order to select a partner, what would you say would be those questions that one must ask and if Hexaware, which is your organization, ends up being the selected one, the selected partner to achieve this extreme automation objective, why should they do so?

Siddharth Dhar [00:15:36]:
Sanjog I’m going to wear two hats for the two parts of this question. Let me wear my subject matter expert hat for the first part. Obviously there are many questions that you can ask a potential partner as you are going through evaluation of potential partners but to me really the most important question if I were doing it for myself, the one question that I would ask the potential partner is are you willing to get into an output based compensation agreement which means we are not going to pay you for deploying an automation platform or for deploying an AI ops platform. We are not going to pay you for coding any use cases that enable some of the promises to be made good. We are only going to pay you when the actual promises that a use case has, are delivered in product. I can tell you only organizations that have done this before that have the experience and the confidence that comes from having delivered those results. Only those organizations will be willing to say yes to something like this.

Siddharth Dhar [00:16:37]:
So that to me is really the most important question that I were doing it personally. I would ask my potential partners. Coming back to the second half of your question which is why Hexaware and for a minute wearing my Hexaware cap on Sanjog if you think of hexaware over the past six years we have transformed ourselves into an organization that has an automation first mindset. It has not been an easy journey. The cultural change has been rough, but over the past five years we have managed to actually transform every Hexaware to be able to automate themselves out of a job. We have provided them the mindset and the cultural alignment first and then we have combined that with retraining our organization, especially on the operation side, providing the coding skill sets required to enable themselves to automate themselves out of very few organizations have made that level of cultural and training investments to be able to do. The second thing I will tell you is especially for organizations that are combining a program like this with operations and execution, you have to realize that all of the account teams that your partners have, they are goaled on customer satisfaction, they’re goaled on revenue growth, and they’re goaled on profitability. Anytime you take undertaker program like this, you are going to negatively impact the revenue for your part there it’s in that itself there is clearly a disincentive for your partners account teams to attempt to do this because they are cannibalizing their revenues. Come the end of the year they are going to meet their numbers.

Siddharth Dhar [00:18:22]:
What we have done at Hexaware is we have actually reversed the balance. We have actually incentivized our teams to cannibalize their own revenues. If any account team can show me that the reduction in revenue is directly as a result of an automation use case that they deployed, we simply add that back to their achievement. It’s a simple solution, but not a lot of organizations have gone down that path. Finally, I will tell you, experience matters. Today we are the automation partners of choice for over five Fortune 100 organizations and each of these organizations are large enough to have other large IT service providers in the system but despite that, they have elected to work with Hexaware as an automation only partner for them. That speaks to itself. Anytime you can say you’ve been chosen by a Fortune 50 or a Fortune 100 organization and you’ve been part of their program in sometime for over three years.

Siddharth Dhar [00:19:22]:
That experience is hard to get out there. So I would say those are three reasons why you should look out at Hexaware as your partner of choice.

Sanjog Aul [00:19:30]:
Once again, thank you Sid for sharing your thoughts and insights about how an organization can quickly move from traditional IT ops to automated AI ops.

Siddharth Dhar [00:19:41]:
Thank you so much Sanjog.

Sanjog Aul [00:19:43]:
And listeners, I invite you to find related conversations on our website at ciotalknetwork.com.

Download Podcast

Explore More

• • Transforming Traditional IT Into Cloud IT
  • Steps to Data Literacy and DataOps
  • Creating a DevOps Vision and Roadmap
  • IT-OT Integration
  • Taking advantage of digital in our companies
  • Are You a Rockstar IT I&O Leader?
  • Building Teamwork Across the Enterprise
  • Enabling Secured Innovation in Your Enterprise
  • Enabling an Effective Virtual Enterprise
  • Rewiring Your Enterprise Architecture With Social Media