AI Ops has garnered a ton of hype, but are organizations actually prepared to implement it and get value?

Artificial Intelligence & Machine Learning (AI/ML)

1.6k views1 Upvote12 Comments

Sr Vice President/CIO, Self-employed

AI Ops is really just collecting everything from your incident command center and all these other tools. It’s the aggregator for all the tools we have in our arsenal. But my organization is not a development shop (we're not the Uber, the Netflix or the Yahoos) so our needs are very different. I buy commercial products more than I develop products. It's a very different model in the health care provider space, so AI Ops is a very different use case for me.

1 2 Replies

CIO in Software, 5,001 - 10,000 employees

I think there's some misconceptions. Even with Uber, it's not that we have all the talent we need. The question is, what do we want to focus it on? We have a buy versus build framework and it is tilted towards buy. We build what we can't buy, but we cannot build everything that we want to build, so it’s smarter for us to take on some platforms and build on top of that. But data is everywhere right now. It is extremely hard to make sure you're getting the right data so that you’re getting the right categorizations. I think that's a huge problem. I'm really struggling with it. We don't really have a great answer for it right now

Senior Director, Defense Programs in Software, 5,001 - 10,000 employees

We have clients that are buy shops that just want the slick configuration, but almost everyone has legacy operations and a contract that comes along with it. Usually the discussion starts with “I have 70 legacy applications that I'm not touching but I have issues with.” They don't want to modernize the cloud. They don't want to replace it. I find very few shops that actually have the architecture or enablements to do AI Ops from the get go. It's always “I got applicants I know nothing about.”

CTO in Software, 11 - 50 employees

Overall, IT teams are not equipped with data scientists that understand how to take data in, build proper models, continuously train them and apply AI techniques, potentially leveraging natural language processing to power chat-ops. That's a massive chasm and I feel like this ‘hashtag AI-ops’ is so much more aspirational than Dev –Sec-Ops, or anything else we've seen in the past few years. I'd love to hear from practitioners confirm if that’s true, or am I just bashing on the hype cycle?

1 Reply

Senior Director, Defense Programs in Software, 5,001 - 10,000 employees

I think you're on point. It's people who want to jump into AI-Ops because it's a Nirvana and they're not doing that groundwork. You need to have that data, that model, and have clean Help Desk data to actually know what solved the problem...not just ‘ticket came in, ticket got closed’. For example, you’ll see a ticket from me about Office 365 locking me out of all my accounts and then magically it's not... what’s it even supposed to glean from that?

CTO in Software, 11 - 50 employees

AI Ops takes in all of these data sources, but it doesn't really think about the context of the data sources. Just because something can be measured or monitored doesn't mean it should be. In my review of the top 15 vendors (at least), none of them take that into account. They don't ask the leaders, lieutenants, or directors, “Does this data source really matter to the business? Will it understand context from this, and be able to move the needle in improving something?”

5 Replies

CIO in Software, 5,001 - 10,000 employees

Fair point. What we measure is really important. But I also think sometimes we don't do a very good job in overall KPI's. We want to measure a lot of things but our focus is split for no very good reason. We need to apply discipline and ask, “What is it that we want to measure, what is it that we want to bring in, and what does it mean?” Some amount comes from iteration; you bring in what you have right now, you discuss it, you figure out that this is the subset. Iteration helps you get smarter.

CTO in Software, 11 - 50 employees

And why are we measuring it? Why is this important to the business versus how does this make me look better? It's the badge of business that all of us seem to want to wear, especially in Silicon Valley.

Senior Director, Defense Programs in Software, 5,001 - 10,000 employees

I think you’re totally on point. We start with collecting what we have, which was usually designed for a totally different purpose. It was perfect for what it was designed for, but you're using it for something completely different now. You really have to take that time to figure out what it is meant for and what it really means. You can look at data, see that your API's are getting hammered, and decide to scale up. But that has an impact. It has a budgetary impact, a fiscal impact. You need context to manage that tradeoff.