Shaked Askayo, Kubiya.ai | DevOps Experience 2022

Hello everyone. My name is Shakira skile. I'm the co-founder and city of virtual assistant solution for devops and platform engineering team Today, I will talk with you about my personal Journey exploring different solutions for enabling a third service culture organization.

Did you think the bottom of the developed team and the powering body developers and operator to focus on walk the tournament? I start off with my personal experience Journey. So professionally I started the system administrator at HP or later stages.

I wanted a devops engineered Intel. And most recently before County could be I had a SRE team at the last winter company or you on your stages have been a devops engineer also a few and in all of these all that which frequent interruptions context which are and repetitive requests on a daily basis. Eventually, I built a solution that helped reduce my team workload by 70% But before we get to that, let's review some issues that my team and I dealt with I'm sure that some of these issues are very familiar to you.

So it's starting from stuff like planning building and managing multiple types of pipelines for the R&D team. It could go to type of pipelines such as cicd cell with automations all of the different tools and Technologies in order to accomplish this stuff. It could go to our managing all types of infrastructure score.

They see from development staging and production therapist of the application lifecycle. Dealing with permissions requests from various personas within the organization while keeping the security team in the loop. and taking the care of the entire onboarding and of boarding process of the engineers in the company, which could go very complex because of the fact that they could have different roles and different systems in the All of that while doing it also had to keep an eye on multiple select channels mostly to answer questions for the tech organization touch questions could go from well my love.

Why did my pipeline fail? Why did my service start on responding kubernetes questions Cloud questions, and so and so I assume that most of these questions are very familiar to you. So continuing with the journey while searching for the perfect solution.

We tried several different approaches. So I discovered a bunch of solutions and the projects to make my team and myself more productive by powering developers with self-service culture. As I went along I tried several different approaches and go over some of those with you today.

So I start off with internal developer platforms internal developer platforms basically empowers a combination of different tools and technology that might be using the organization and by those under the same umbrella or easy access and management. For the purpose of reducing the workload. We consider building one.

The two problem were that there were too much planning required. With so many different organizational tools that our developers were used to and the culture of the organization which was very Dynamic. We found ourselves thinking and struggling on how to create such platform and what are the features that will be included in.

we will also very worried about getting yourself a new kind of style so we could reduce the developer oil but embracing a new one which is managing the platform life cycle any new feature supporting the different users in front of next I'm going to talk about workflow automation another attempt to reduce the toilet so that the industry standard solution one of the solution that our Tech organization on Floridians of tense, but utilizing traditional tools that our developers are already familiar with natural Jenkins were created automation for nearly any issue you can think of Estimates an issue occurred we were able to create a dedicated pipeline called it. The main problem was that developers did not know where to find this workflow. Even if they knew were to find the workflow.

They did not know which parameter to fill in. So returning back to the same problem of them approaching to us for help and contact. Permissions was another issue.

It might be very risky for the organization if workflows can potentially perform chaotic actions. As well as who has the authority to one such workflow having to Grant permissions and approve those with the security team will also take a lot of time and effort. The life cycle of such permissions is also problematic.

Because how can you manage it in a proper way when a user release the organization for example, or in case of getting a new workflow? How would you define the different emissions that are required and for who? Another problem which was if their work release edited to include more actions.

What is the process to review it? To serve to solve this issue that we had I decided to create a chat one. In light of the fact that the request Came To Us by using a chat interface such a slack what if it's using the direct method in each Channel or the uncle Channel which added the collaborative element to it.

And because of the fact that jacket already had the vast amount of automation from the workflow automation parts that we experienced between it seems very logical to combine the two. In order to address the security concerns in terms of permissions management. I connected this chatbot to the organizational identity provider.

by doing so the chatbot will know which all the organization in the organization the Persona place with it and require this assistant and So this Empire with us to create different policies on who can do what in terms of consuming the operational walks of the pot can do for the organizational, you know with requests coming in from developerative. I developed redefined workflow that combined workflow automation systems with cloud provider. So in order to combine the content together in part you could ask the Bots together you context about resources in the cloud and this way you could know which little set to fill in as parameter for the different workflow.

That's basically existed on the workflow automation systems such as Genki. Besides solving the lack of a domain expertise and lack of interest in operations the checkpoint acted as epoxy with guidelines for developer. This allowed them to do what they needed without over permissioning.

most importantly introduced my team workload by 70% while eliminating long delays for end user preventing them for waiting in the virtual queue for a very long time to a complete the work that I wanted to do. However, chatbots are not exactly user friendly. User says to choose from a static list of workflows and use predefined words or less comments in order to accomplish the task that they want from the bonds.

In any case it will not be able to handle anything not included in its rule-based can interaction. Our Dev team will be left empty-handed as a result of death out of scope requests. Another important and very problematic Point were maintenance.

Maintenance, however were Star Wars. With so many walkthroughs to create and so many Devils Cooks in the kitchen. I could not enforce a standards programming language to define the different workload that about can do or not a stand up at all.

Well one walk from work, they going out all of the dependencies and how to fix those two way too much wet blood until If I wanted to add a brand new workflow, it will require very significant effort and domain expertity. So only the maintainer of the path will do it attending to the 10 problem of us managing now another toilet, which is managing this chatbots. The person cons of all of the different solutions leads to what I believe is the most ideal form of virtual assistant that uses conversational AI.

These benefits both the developers and the operator. Let me explain. For the end you though the Developers?

The system is very end user friendly combining all of the features of the chatbot with an element that understands the user intention. The workflow literature system offers can ask questions if contact is missing in other words it transforms the content hearing power, which is very technical into a conversation that the user has the future. So if a context is missing the virtual assistant is able to feel it for cloudy cell and assets or part organizational workflows.

Of course that the solution for the end user also boost the productivity and closes the gap between the endurance and the operating. using natural language to extract the user intention allows the virtual assistant to execute the correct workflow based on partial related and input while learning from the organizational behavior, which results in imposed service experience also the time With this information operators can identify which requests are coming in. Which workflows are walking accordingly and how to adjust the self-service culture in the organization?

For the operator on the other hand the cost of production and two-way feedback loop. For the second hand the several picture of the solution including command line interface for managing it. Which means that you could manage the virtual assistant the same way you would use for different systems such as kubernetes.

And also a domain specific language which allows to manage the virtual assistant workflow this to conversations that discussed about in a declarative way. Along with honoring detox concept so imagine this workflows can be stored in detail positive. So they could be part of the cicpd pipeline and so on.

Besides this fact the system Hollows serverless functions. So you could use all of the Power of several Health in terms of programming language users or library is external libraries of third party integration without losing the structure of the cultural system. The workloads are keeping on being declarative but is capable of involving such serverless action.

For Less technical user and no code interface to manage workflows in a visual way. All of that with the Deep connection to the organizational identity provider in order to Define access policies and overall policies. So this way even the security team is part of the loop.

So that's how I describe the best way to manage the self-service culture in the organization. I hope that you enjoyed people presentation today. Thank you so much for taking the time to watch my journey.

It's discovery. And of course feel free to visit our website that could be out of AI to learn more about the service experience insights and discoveries that there could be a platform also. Thank you so much.