Amit Eyal Gorvin, Kubiya AI | KubeCon + CloudNativeCon Europe 2023

This is texturing TV. Hello and welcome back to kubecon plus Cloud nativecon Europe. I'm Mike Vizard.

We're with Amit government. Who's with kubia AI and we're talking about how AI will be applied to devops and welcome to the show. Thank you very much.

Thanks for having you. Everybody is Brothers talking about chat GPT, but you guys have been ahead of this game for a little while. You've been putting conversational AI in front of devops platforms, and now you're figuring out how to help people build workflows.

How automated can automate it get I'll tell you the true answer. It's not about automation aspect of it. It's a human and a loop element of it that actually is what extends most of these processes and makes him a lot more cumbersome than typically a machine to machine.

It's very easy to go into automate some with apis. It's a whole different process. If you actually have to go and get human feedback get people to approval flows get them to do SUB transactions that may not be part of that workflow.

And that's actually where a lot of operations extend itself and getting cues and so forth. So this is really a way to to consolidate much of that. A lot of people are always suspicious of things that are magic and things they don't understand so walk us through how this actually works because that will demystify it for people and they'll be less afraid of it.

So Chad GPT for devops is really more of writing the buzzwords, but let's actually consolidate what that actually means. There's two sides of the coin that we're solving for. The first one as you correctly mentioned is a conversation AI for the end user experience, you know, when we launched last year we talked about our service here for devops, but that was pretty Chad GPT.

People didn't really know how to contribute or attribute conversational AI to operations. So we said look think about it as a way to prompt using natural language having the end users really be able to consume operational workflows devops function. So think about Knowledge Management, how do I do this in let's go one step back think about the concept of data catalogs.

Okay you have Can Confluence your technical docs people come and prompt the on-call engineer? How do you do how do you roll back a kubernetes environment? How do provision anyone cluster and so forth.

These are things that typically will require a human and a loop an operator to go into Health provision or you'll throw you at a few sets of docs go and read it for yourself. Think about the large language models now training on the customer data. Okay, using a vector database.

It's all their own namespace. So it's all secure and now we're using structure and unstructured data to embed that into human conversation. So now you can go and ask the machine.

How do do XYZ it will go and not just throw the Links at you. But also the precise answer with actual actionable prompts into it. So that's end user experience ability to go into interact with the machine as if you're talking to human on that other side.

That's one side. So it doesn't sound like AI is gonna eliminate the need for devops not if there's per se as much as it may give them or augment them. A hundred percent.

I think everybody asks you think about Ai and its respect. It's not here to replace you. It's here to empower you and that's actually where to empowerment on the other side of the Corn exists.

When we're talking more on the operator side the platform Engineers devops SRE teams people who are in charge with really standing up on an internal platform to serve their their customer their organization needs and to maintain those that's worked. So creation that's data management. That's all the various.

It's it's a lot of work a lot of heavy lift oftentimes. If you're having to replace a system. This could be a project and that factors into a lot of buying Decisions, by the way for some of these organizations.

So what we've done in this is what we're now announcing coupon is allowing generative AI for workflow creation and automation ability to essentially prompt using natural language. The system will build out with all the business logic. Attributes access controls all the various components that the operator would necessary have to have and build it out in low code and allow you to fine-tune that workflow and then commit that debug it and committed to your organization.

So things that would take otherwise hours would take now minutes to As we kind of sort that through if I go into an organization, I might see hundreds of developers and maybe four software Engineers on the devops side. Sure. So we change in the math so that those four people can actually support a hundred devops developers type folks or because devops in some places as a bottleneck, so bottleneck going to go away That's the idea especially in today's economic climate people may not have the privilege of hiring more head count in order to support organizational needs but the organizational needs are not going to wait and this is a way to really augment that and really help get a lot more.

I don't want to say 10x productivity. We've seen that happen before but have a lot more productivity out of the existing head count. And so think about this for budgetary reasons.

We're also helping optimize licenses seats of different SAS providers. We're helping people and that budget controls within their infrastructure provisioning very easily say within slack or teams. So that's also a way to really rein in on some of the economic factors while also optimizing the efficiency of your organization.

So what's next what should people expect we have conversational we have the workflow is created. So where do we go from here? Do you mind if I give you more of a statement?

I think one of the reasons why large language models has been the iPhone moment and people are getting infatuated with the magic of how it looks like and how easy it is to create very nice demos, but the end of the day what it actually did for us it changed the way humans and machines interact with one another. Okay. It's a conversation up until now humans needed to be us, you know, apt enough to code or low code with all the business logic to interact with the machine.

So it understand now for the first time machines can speak human language and convert that into code so that change the dynamic of how people interact so think about this as the gateway to the next level how to interact with the machine as if it's a human operator and really help streamline a lot of operational knowledge and data silos that exist in humans that may leave the organization and leave the organization exposed or otherwise, you know inefficient. This is a way to really maintain your tribal knowledge within organization in-house and to to have a continuity player for the organization that also skills with its size. So how smart will all this get am I gonna be able to go into the office one day and just verbally talk to the machine and say tell me the three things that might get me fired today and then it'll come back with these are the three things you should be working on.

We're not an HR, but people to tell you that it's actually very controlled and rained in and politically correct and so forth. It's very fun interaction. By the way, it can tell you jokes.

You know, what I'm to my favorite jokes that tells you where can I date take my date tonight? It will say the kubernetes club it's so there's all sorts of fun prom so you can have and it tells jokes but at the end of day it's conversational. It's supposed to be non-threatening.

It's supposed to be easy to use as if you're having a chat with Jason or Bob or Frank and other side only that here it's available 24/7 doesn't take time off. So really gives you that kind of end-to-end day to day experience. We've seen the emergence of prompt Engineers as a job doing is that a skill for real or am I just gonna have natural language?

What does it prompt engineer going to do? Well prompt engineering is how do you interact with the machine to get your intents across in a optimal manner? You need to know what it is you're looking to prom but the machine the more specific you give the fine tuning if you may you give it the more precise of an answer will give and that's precisely how the the new methodology today prompt engineering is going to replace, you know citizen development and coding if you make so often till now.

May have been low code you can go and express yourself relatively easy with drag and drop you still have to know the business logic behind it. In order to prompt a local here. It's a way to it use your words in order to prompt it.

So what's your best advice to folks as we start this Brave New World of AI that we're looking at? What should they do to get started? It is a Brave New World, right?

So Embrace change and embrace AI If I can give you the best comparison I can give it's where you know first Industrial Revolution, you know, the Ford Motor Company the first smart assembly lines. So initially it was humans Manning every single p*** of it screwing everything about you know, the the first factories were getting dozens at most per month, you know of inventory out the door, then the smart assembly lines came on board and the humans became operators of machinery and could get a lot more done. It didn't replace humans all it is allowed the humans to be more domain focused into what they're doing in optimize their work.

I think that's very much how we should look at AI it's supposed to help Echo our best attributes. And as long as you can can find yourself within that realm and know how to reinvent yourself a bit in terms of your day today. You can be that much more valuable to your organization.

All right, folks wise one said the Futures here. It's just unevenly distributed time to get started. I meant thanks for being on the show.

Thanks for coming back my friend. And we'll be back in a minute.