Clean Data Comes Before an AI Service Desk
In this Techstrong TV interview, Mike Vizard talks with Priya Kumar, global head of product support for ServiceDesk Plus at ManageEngine, at the ManageEngine User Conference in Orlando. They discuss what it takes to build an AI service desk, why data quality is the real hurdle and how AI will change the work of IT support teams.
Best Practices Still Need Human Expertise
The most common questions Kumar hears from IT teams are about best practices for processes tied to compliance, audit and governance, risk and compliance (GRC). These range from user onboarding and offboarding to full asset lifecycle management, so teams have evidence ready for auditors. AI can offer some guidance. However, Kumar says it cannot replace the experience of seeing how many other organizations set up their processes.
Clean Data Is the Real Hurdle
According to Kumar, setting up the tools is not the hard part. The struggle is the data inside them. Records are often spread across several applications, conflicting or labeled inconsistently. When AI reads from that data, its answers are only as accurate as the records behind it. Her team works with customers to consolidate, archive and correct records based on their process and industry compliance requirements. That groundwork lets the AI service desk predict ticket surges during an outage and support proactive problem management.
Less Burnout, More Meaningful Work
Kumar expects AI to reduce turnover by automating monotonous tasks such as password resets and access requests. Technicians can then focus on strengthening cybersecurity and infrastructure. She does not expect tickets to disappear, since end users will always need help when a suggested fix does not work. Still, she believes AI will reduce the share of those tickets, and that service desks will need to adapt as AI agents become end users themselves.
Audits Get Easier, Not Fully Automated
Automation can gather and send much of the evidence auditors request. However, Kumar says a human will always need to verify that records are accurate and have not been manipulated. Her advice is to get processes such as asset inventory in place early, before a major incident shows the gap.
For more information, visit manageengine.com. Watch more conversations in Techstrong TV’s Interviews section.
Transcript
Hey, folks, we're back at the ManageEngine User Conference in Orlando, and we're having a little chat about how AI is changing the way we manage IT support functions with my good friend, Priya. Welcome to the show. Nice to meet you, Mike.
Nice to meet you. I am certain that you are probably on Zoom speed dial for thousands of IT professionals out there who are always looking for somebody to help them with something. Yes.
So before we get started with the AI, what is the most common thing you hear from IT folks lately when they're looking for help and support, and maybe what do you wish that they knew already? I would probably say most of the questions that we get from the IT presence is how do they set up the best practices for any sort of process, especially something related to compliance, audit, or GRC perspective. So they ask us those kinds of best practices.
They do have a process, but they do also like to vet it with us. Okay, is this the best way to set it up? That's the kind of the questions we usually get, or most of the time, those questions, the back-end process always ends up in an internal audit or compliance.
That's how we generally get a lot of questions. It could be as simple as how good we can do an user onboarding, offboarding, to how can I do a complete asset lifecycle management so that when there's an audit or a compliance where I need to show some evidence to my auditors, I have everything recorded in place. So as we move into the AI era, is a lot of that going to become more automated?
Is it going to be like a natural language interface that I go ask an AI agent for some insights there? What will be the line between what the AI does and what you might still need to do? Right.
So the AI can give me a bit of guidance on how I can do it, but not necessarily that guidance is what is the actual process they want to set in their organization. And that is where we come in and help them because we have seen other organizations, how they implement their processes. So we teach them the best practices on how to set up those processes, like I said, anything related to the audit perspective or GRC perspective.
AI can help a little bit, but not in the end, is what I would say. How proactive will the system ultimately get? Will it be able to see what people are doing and make suggestions, or is this more of something I have to call it first and then it will learn about my environment?
Or I guess, how dynamic is that going to get? So the major part what we actually see is some of the environments or organizations, they have a better process. So the data, what they have in their tools, which can be consumed by the AI, is perfect.
So in those cases, the AI accuracy or the results is much better. But most often, we say the underlying data, what they have in the application, which feeds the AI, is not really streamlined. So if you even ask a question to the AI, it basically reads from the dataset, which is already available, which might not be accurate.
All right? So that's something which we frequently see. And then we help them to correct some of the system data, which is available, so that in time they will have the AI to help them.
And how does that process work? Does the data have to go into some big giant cloud somewhere, or can you see the data where it resides and help them figure out what that workflow should be? It depends on the organization requirements.
Sometimes they would say, like, "Okay, I would just migrate some of the data to a new setup and have the process completely configured, and then I will have the AI consume that, where I'll have to archive some of the records, which I might not really need. " And in some cases, we usually don't send that huge bunch of data to another tool to do the cleaning part. We have certain steps.
We understand the process, what they wanted to do, what sort of compliance that industry would be put into, and then we streamline the data corrections for them. That's basically what we help. So as this continues to evolve, how will the role of the service desk or the help desk, whatever you want to call it, will that evolve in some different way than we perceive it today?
Because I think a lot of what those folks do today tends to be reactive. Mm-hmm. So can it become more proactive?
How will that manifest? Of course. They can do proactive IT management when they have the datasets properly.
And with AI in place, it can proactively understand what kind of ticket volume would come up because there is a big outage. So it can predict and give you an information where you can be prepared for that situation, or you can mitigate it even before it becomes a problem in the environment. How hard is it to set all this up, and where are we on this journey right now?
I wouldn't say it's actually hard in setting up the tool for this process. The actual struggle starts with the actual data that they have in the applications. It could be one or two or more applications.
Sometimes the dataset is not properly streamlined is what we generally see. Or they have data in multiple places, which we can consolidate and have the AI consume those data. But the actual problem lies on the accuracy or the right dataset is what I actually see.
And sometimes the datasets are conflicting because different people put different things into different items and call different things the same thing, and it's got two different names? Right. All right.
And then we check with them, which is the right thing to do, and then we have to clean up the records which they don't want, and then correct the records which they really need to be consumed by the AI or for them to do a proactive help desk or proactive problem management, the ideal best practices, and that's how we generally help. There's no secret that there's usually a lot of turnover in help desks and service desks. Do you think in the age of AI, that might be reduced because the amount of manual effort and all the things that burn us out might go away?
I would definitely bet on that because, for me, if I'm an IT technician, if I just keep doing a password reset, and I don't think so that's really productive for me. I'm not learning anything new on that perspective. If AI could help me to do those monotonous or automate those monotonous tasks, then I would concentrate really on how can I strengthen my cybersecurity, how can I strengthen my infrastructure, which is what is currently needed, right?
Mm-hmm. As you look at all this, will the roles of IT teams evolve or converge? People will be doing more things, maybe.
Everybody I talk to who has AI is probably working a little bit harder, not less, but it's because they're just able to take on more. Mm-hmm. So if the IT folks are not spending all their time chasing down tickets, what can they be doing or what things will they be doing that might be more interesting?
If an IT technician would be just chasing on the regular tickets, like I said, the password automation or giving me access to an application, that's really not very productive for me, like I said. But if you give me a challenge to say, okay, there are certain set of network achievements which needs to be really looked into to see where it could potentially cause some of the loopholes in my network, and that's where I need to concentrate and see how I can remove them or upgrade them and ensure my infra is actually strengthened. The huge volume of tickets, I would say, it can come in, but that's where you put in an AI chatbot, which can help the end user, or which can also help the IT technician to give more insights of there were incidents like this previously, and this is how it was solved.
Those kind of stuff will be really helpful. What is your sense of can we ever maybe get rid of the ticket altogether because we are so proactive about it that there won't be some ticket generated because we'll be able to be in front of this the whole time? I would definitely not say that.
Okay, if I say that, it means I'm lying. Mm-hmm. See, at the end, help desk is also not just meant for getting the alerts for monitoring tools or cybersecurity tools.
It is meant for your end users, right? And there's always a chance where the end user might not land up in the right set of questions, or they might not get the right set of solutions, or they might have tried some solution which has not worked for them. So obviously, they will end up raising a ticket in help desk, right?
But what I would say is the percentage of such tickets with the help of AI might get reduced, is what I perceive. Of course, our current systems were set up for end users that are humans. Mm-hmm.
But as we go forward, it's pretty clear there's also going to be end users that are AI agents that are performing any number of tasks, and they'll need support at some point. So, are we going to have to kind of re-engineer our service desk to support these, I guess I'll call them non-human end users, of which there might be tens of thousands of them? Yes.
You need to definitely remodule the way how you do the help desk part in your organization if you really wanted to be productive and also adopt the latest technology, which would be beneficial for your organization. I wouldn't say every part of it, of the help desk, needs to be completely automated. You definitely need some part of it where the human have to verify or make use of AI to get those documentation pre-built to help the end users.
Right. " Certain times, I would probably see on small things that they might have missed. Basically, the process is what I've seen in most cases.
Mm-hmm. Like I said, like, "Oh, I have missed to do a proper inventory of my asset records. " That's where I see most of the times they think, "Oh my god, I missed this.
" Do you think that we're getting to the point soon where the audits that are required for a lot of regulated industries for compliance, that too is going to get automated and maybe make the whole thing a lot less painful? Because nobody I know enjoys an audit. Oh, yes.
But again, I don't think any audit would be completely automated. They, of course, would like to see certain informations, meaning an auditor can get some of the proofs which I can automate, I can send, or they have another tool which can fetch some informations. But of course, there would be a human intervention to verify the records, whether it's right or wrong or whether it's manipulated.
So that will always be there, however your organization is. All right, cool. Well, folks, you heard it here.
You know what? We're on the cusp of trying to make IT fun again, and we're getting close. Hey, Priya, thanks for being on the show.
Thank you, Mike. It was pleasure meeting you and talking with you. All right.
And we'll be back in a minute.
