The Future of DevOps with Matt Schvimmer at Atlassian Team ’24

This is Techron tv. Hi everyone, it's Ashley. We are here at Team 24 Atlassian in, uh, Las Vegas, Nevada.

So having some great conversations. One of my favorite topics, talking about Agile and DevOps. Speaking up of which, Matt Swimmer, who is SVP of Agile and DevOps.

Correct? Yes. That's it.

Great to be here. Mitch. Mitch, we've been bumping into each other this whole, we finally get to talk, so That's right.

I've been looking forward to this. That's right. Not in a casino setting.

Not Nice, not at a casino or a Starbucks or whatever. So tell tell folks about what you do is, uh, I'm sure Agile and DevOps are covering a lot of people's passion. So what do you do at Atlassian?

Yeah, for sure. So I'm responsible for the Agile and DevOps portfolio that includes Jira Align, uh, JIRA. Like that's the combination of Jira software and Jira work management that I'm sure we'll talk about.

Yep, yep. Um, compass, JPD and Bitbucket. So all the tools that developer, oh, sorry.

Acronyms. Uh, jpro Discovery. Okay.

Thank you for that. It's, you're welcome. If there's one thing DevOps likes, it's a good three letter acronym.

TLAs, we were all about the acronyms. Right. Great.

We've talked about discovery quite a bit too at one of the folks on with that. So, um, so being, being in charge of the head of, what does that mean? I mean, it's gotta be a lot of responsibility here.

Yeah. Well, um, you know, it's really about driving the product strategy to make sure what we're building matches what the customers are asking for. So it's kind of meeting customers where they are, uh, making sure we have a prioritized roadmap, that our products fit together in a very coherent way and are differentiated in a compelling manner for customers to, customers to love.

Mm-Hmm. Right. And so a lot of what my team does is we're focused on the development team as job one, and Job two is kind of the extended organization of all the different players involved in the software development lifecycle from executive leaders, uh, to, you know, the product managers, to the operations folks, how that all comes together.

Mm-Hmm. So that's what we think about. I'm curious, um, 'cause I've worked in security base space too, and security to design a product to solve this problem, to look for this kind of attack or whatever it might be in DevOps.

It's, it's not a tool, it's, it's a way of creating software. Way of collaborating. Yes.

Uh, philosophy around smaller units of work, you know, automating feedback loops, all those kind of things. Similarly, a lot of it's lifted out of Agile as well. Yeah.

How we work together and certainly tools are part of doing that, helping us do CICD pipelines and all that kind of thing. They're, how, how do you support organizations where I might be at the beginning of what's this DevOps and how do we get started to, like, we're, we're, we're close to from mastery to unicorn status of knowing how to leverage da, DevOps and Agile. That that's a big spectrum of skills and knowledge.

Oh yeah. A hundred percent. A hundred percent.

I don't, I don't, I don't know if there's ever an end to that, that spectrum. Mm. Right.

It's continues to evolve in pretty interesting and amazing ways. Uh, security is a good example. Uh, uh, that you brought that Up.

I'm oversimplifying, but Yeah, just as an, Uh, no, it was a great representation of, um, like DevOps is not one thing. DevOps is a collection of processes. Right.

And the thing about we've been focused on Atlassian is like, how do we connect all those different pieces, right. That, that, that, and that's been like the lasts focus. When you think about, we launched, I think it was two years ago, open DevOps.

Mm-Hmm. I remember That as the initiative. Right.

And it really is that acknowledgement that, you know, you're always going to have best practice tools to meet certain jobs to be done. But it's about how we connect that graph of all those tools. There's the graph Where, again, we're talking about the graph.

We talk a lot about the Graph, which I think is great, by the way. It's uh, it's incredibly critical. It's fundamental to our strategy.

It Is. It's a differentiator for you. I think.

A big one. Yeah. It's, it's a hundred percent.

Like a lot of the investments you see here, uh, I'm, I'm talking about developer experience. A lot of the investments we're doing are delivering value from the graph to make the knowledge worker a lot smarter at the point of whatever part of the DevOps process they're responsible for. It's kind of the neural system behind all the work that's happening that knows who's doing what and work work's happening and helps you kind of plug in Jack in like, Neo and, and Neo.

Neo. I sometimes call our founder Neo in the Mac Neo. Oh, okay.

Oh, well. Yes. So hopefully that's A compli.

I don't know if he appreciates that. Hopefully he doesn't hear this. Well, you're still working here, so I guess it's okay.

So Far. So far. So, um, So many things I'd love to talk to you about.

Um, it's a lot of discussion about platforms. Yep. Not just here, but also in the industry of moving to more platforms, things rather than, um, siloed, integrated tools, you know?

Yeah. It gets in the business of integrating tools rather than creating software. 'cause it just gets so more complex.

What, what does the platform strategy look like for open DevOps and Agile and you're, you're part of that platform. Yeah. So, uh, at the core of the platform is work.

So you, you hear a lot in the show about the system of work. System of work and, and Jira is really that kind of, if you think about DevOps as a hub and spoke system and will go with me on this riff a bit Here. I follow you wherever you go.

Okay. Alright. So I'll remember that.

So, uh, JIRA is really the hub, right? We manage all of that work. And the spokes are, we understand the connectivity to all the other information across that DevOps lifecycle, right?

We understand the signals and feedback and ideas that come in on the front end. We understand from product, from designers, uh, from maybe in Figma or Adobe, what they're designing, how they're iterating on it, right? We understand from observability tools, once we throw things over the wall, hopefully not throw a all or release it to production.

DevOps is the Virtual wall. The virtual wall, um, that we're gonna tear down. We're gonna tear down this wall.

So when we turn it over to operations, uh, you know, we understand the impact to all the different services, what we've deployed, and we can provide from that connectivity from Jira, that hub, that amount of intelligence. 'cause we could connect all those different pieces and make it kind of a coherent value stream, if you will. How it all fits together.

There Is an in value stream. That's the idea of flow, right? Mm-Hmm, yep.

Helping create and, and, uh, maximize, remove barriers to make things happen more fluidly. That's right. It sounds like that's kind of the thing, one of the thinking behind having the graph, having the data, having the tools, but it's all about making that work together.

Yeah. And you know, um, a lot of times people will talk about it as if it's about productivity. Mm.

Um, I'm not sure that's the right word. Uh, I think, uh, like a lot of what I, what I talk about is about dev experience. Mm-Hmm.

And it's about reducing friction and cognitive load so that things flow easier. Yep. Right?

And so no matter where you are in that stage, you have the information to, uh, to make that process more informed, smoother, faster. So not necessarily productivity. Co Cognitive load's a huge, it's a massive issue, especially in context switching.

Not to make us all computers, but you know, it's the, I'm doing five things and every one of those takes a certain amount of context switch to get my myself back in that head space to work on that problem. Where was I? Okay, now I can, alright, I can start to work again.

Yeah. So how, how do you, how do you help with that? I mean, I think you're right.

It's not just Protectivity. I wish it was that simple. You know, uh, this is definitely not one of those Lord of the Rings one ring to rule them all kind of things.

There's a lot of different ways Mm-Hmm. I think to do that. Um, one is a, uh, I'll try to not say AI too many times, sorry, but It's a, it's an okay word to use that it's, We're good.

So we're all among friends. Yeah. So, in ai, um, one of the ways we're using that is to, uh, when you come back, 'cause you context shift a lot, you might jump in time, slice on different things when you come back.

How can we use AI to surface just immediately? What's changed or what has happened since you last took action on some of this? So giving you up to date information, like, one of the things in je line we're doing for this, by the way, if you think about, uh, an executive leader who hates to read as, you know, five minute information.

I heard Just a few of those are in, there's a Lot, there's a lot of those. Uh, we won't name names. No, no, no judging, no Judgment, no judgment.

But you have to have you just, like, you need to get in there, you need to be brief, you need to be bright, you need to be gone. I love to be brief, be bright, Be gone, be valuable during your valuable Time. AI gives us that.

It allows us to summarize information for a leader to say, here's what you missed and here's what we recommend. So just, that's a one way we're doing it. The other way is from the system of work, by taking advantage of all these products being on Atlassian platform, uh, we're surfacing information, uh, within applications more coherently so that you don't have to jump around.

One example you'll see here is, uh, I'm gonna talk about Compass, that's our developer experience platform. And it allows you, just one quick aside here. It it allows you to track all of your different microservices, all your components across your ecosystem, right?

The health of those components, the dependencies, there's a bunch I can talk about there, but, uh, what we've done is we've surfaced that inside of Jira so that when you're working on an issue, I can associate the right components, the right services, the right APIs to the issues. So developers don't have to go and search for what the heck am I supposed to be working on? Right?

Mm-Hmm. Uh, it's a much, and it's also easy to get insight when we talked about, uh, some of the metrics that you want to get out of. You wanna see what I'm investing in each of these different services by connecting that at the, at the unit of work, again, the hub, you know, we have more granular insight to understand how much we're investing in, how much of this unit is changing, that kind of thing.

How does it, it's a great topic and I appreciate you explaining it that way. You know, back in the day, I remember, um, crucible and Bamboo and some of the tools, I dunno if they're still around or not, but doing code reviews and, um, developers we live in, in the ID so much Yeah. And having to bring up a, a bug tracking system and, you know, whatever documentation and I repository and yeah, I can have a window or a tab in my browser for all those things.

But that's just, that's for one problem set. Then you need to go work on a different, which may be a whole different set of things. How, how do you incorporate that this, so you think about developers 'cause they also have their IDE that they're working in in addition to the interface.

Oh, yeah. Oh yeah. Uh, the ID is not going away.

Yeah. Uh, Well, matter of fact, just getting more complicated, I think It's, uh, what, what's, um, what isn't happening is that we're not gonna double the number of developers in the next few years. Right.

So the goal is we have to get, uh, we have to empower developers to do a lot more. Mm-Hmm. Right.

And so some of that is in the ID like, one of the things that we're gonna talk about is how we're making other copilots smarter. So inside the IDE, anything except that we have access to, within the teamwork graph, you can ask questions to within, if you're using tab nine, using GitHub, copilot code, any of those things, um, we put the power of Atlassian within that copilot, so make 'em more productive in the editor. But a lot of what we're doing, because we're not adding them any more developers Mm-Hmm.

To my point before, uh, we're finding ways to, uh, make it less necessary to get into the editor in all situations. So for people that come to the super session, they'll see how we're doing something we call auto depth, which is taking a unit of work. And when we feel confident, like unit work being a description of an epic or an issue and being able to automatically code that and be able to say so from an issue, develop it if we have a high degree of confidence to get it to a pr to, to a developer to review it and then just make minor changes to it.

So low code, you know, smaller items we can get you outta the editor. Get those out faster so that, you know, the more complex things where we have to be in the editor, we can be a partner with you to help give you advice and give you context. But both in the context of, you know, kind of expanding the aperture of what an individual developer can handle.

Well, And that's, you know, it's, it is interesting me 'cause fixing a bug is much different than creating a new capability. Right? Sure.

Bug fixing, I've gotta look at the logic and I don't remember how I did that. Or, you know, it's a different math, you know, there's an algebra to that and a calculus to how you create things. So it, it seems like one of the areas we can help developers a lot with is let's get rid of the algebra kinds of activities.

Just not to pick on algebra, but give you more options, more time to work on the things. Really take the, the cognitive ability. You've got A hundred percent.

We always, uh, talk about reducing cognitive load and I didn't think that was a good way to describe things. I'm like, who talks about cognitive load? Turns out everybody talks about Everybody does.

It's a, it is a, it is a topic to Be seen. It it's a topic. And so yeah.

That I, I'll give you an example of one of the ways we're trying to do that. Um, so I, developers right now, they spend maybe 20, 30% of their time actually writing code. They spend the other, Sometimes less, But yes, sometimes less.

Um, uh, they spend 70% doing a bunch of other things. I'll try to keep it clean a bunch of other things that it Is a family show it Okay. Family environment.

Um, and, you know, some of that is just looking for content. Some of that is bug fixing. Right.

Uh, some of that is how do I clean up my environment a lot. A lot of it is actually supporting what's already in production today. Yeah.

So, uh, one of the ways we're gonna talk about Encompass, uh, that I mentioned before is, um, how we can use AI to automatically identify things that we can fix for you before you as a developer have to manually go out. Case in point, uh, and Mike talked about in the keynote was cleaning up stale feature flags or being able to address security vulnerabilities that have been fed to us from third party security tools. Mm-Hmm.

Bringing it back to your security experience like that. But like, we call this thing auto fixx because we haven't got marketing involved yet. So, okay.

Auto Fixx allows you to say, Hey, I've recognized these things about your code and production that are providing an unhealthy environment for the components that you have under manage. Uh, the Product name should be toil, toil thing fixes, toil. Oh, I like Toil.

There you go. That's a, that's a good one. You have everybody On side better hurry Up and copyright that thing before I Get outta it.

I think you put that website running your Yeah, But that's a, that's a, that's a good example. The other way we're trying to reduce toil, uh, you're gonna hear an announcement. We just acquired a company called Optic for API Documentation and Management.

Mm-Hmm. One, one of the things we've found is that, um, uh, people don't like writing documentation. I've heard that.

Yeah. I've heard that. I, I hate, I hate writing.

Um, and so Optic allows us to automatically identify APIs, be able to create the documentation and identify if the API and the documentation start to drift. And so we can flag those things before they come problematic to the feature. And that's gonna be a big load off.

Developers. One of the challenges with a, with APIs we don't always talk about is sort of the lifecycle management of APIs, right? Because they have versions, generations of them, you know, backwards compatibility, sunset deprecate, that kind of thing.

Is that something that's gonna be part of this? Oh, I'll let you today. Or Could I'll let you track the whole life cycle.

Okay. From to tech manage to retire, Use that, that's been replaced by this and use version two of that. Well, what you'll see is what, what versions are in production?

What dependencies of other services depend on that? API. So like if you're gonna upgrade a version, you can identify where all the different upstream, downstream relationships that you have to also upgrade at the same time.

Right. You can also build workflow to say maybe if you wanna update to a certain version, I can create an automation to identify all those different points and kick that off. Is there, I'm, I'm curious 'cause I'm not familiar with that, that particular technology, API first is the strategy, you know, around building your software so that it uses its own APIs to do things.

It's not sort of that outer inner shell Access, it's just kind of a policy, like it any like cloud-based software fundamentals like Microsoft APIs, right? Yep. Exactly.

Is it, is there anything about this, this technology that you've acquired that will help you understand where you're using a, a large number of APIs, not just here's the 12 calls to do this and the 12 calls to do that? I think the nice thing about it is, um, uh, it's often hard to find like what, where the problems are. So you have tons of APIs, you're talking about let's go thousands of APIs in your infrastructure, right?

How do you know which ones are, are broken at any one time? Right? Instead of just trying to surf through and, and figure things out, we can surface those up based on scorecards and health metrics to be able to point to which ones are broken, which ones need attention, which ones are outta compliance.

So that's, uh, that you're Gonna sunset this in three and a half months. So yeah. Baby toil steps in and fixes it for you.

Exactly. Baby toil, is that what you call it? Baby toil?

Yeah. Auto toil. Auto want toil.

Auto Toil. Yeah. No Toil de toil.

We'll, we'll work on that. I'll get the marketing people. Yeah.

Clearly I'm not marketing involved. Yeah. Your talk around, um, developer experience.

Yeah, it's a great topic. It's, it's one of the aspects of platform engineering, right? We talk about, did we forget the developer, let's make sure that we're helping them.

Again, product productivity is a part of it, but it's a lot more than that. It's a lot more than productivity. Right.

It's, uh, and by the way, developers are pretty wary whenever you talk about productivity. Yeah. I'm here to help.

Yeah. I'm from this department. I'm here to help.

Yeah. Yeah. Thank you very much.

What, when I've seen, and it's funny, I would say nine out of 10 customers I talk to, uh, when I ask them for their, I, us, I always, always ask, you know, what are your big priorities? What are you working on developer experience? I would say nine outta 10 is the top three initiative.

AI is always in there, developer experience, and then there's a wild card one mm-Hmm. Um, but, uh, in developer experience, it's been a lot more about, um, uh, uh, enabling developers, like giving them the tools. So platform engineering is usually the buyer and platform engineering is, uh, chartered with unblocking developers giving 'em the tools they need to do all that toil.

Mm-Hmm. Less painfully. Yep.

Right. And so that's why, so we, we position Compass for that to the platform engineering team, but we don't position it as a, a stick that they point to developers and beat them over the head with do Shal Comply. Mm-Hmm.

Even though it does compliance, which is what engineer platform engineering likes, it also gives the developers reasons to come back. So that API documentation is one, understanding software health is another understanding. Like for all these microservices that I gotta go work on, your, your bug fix is an example.

Something comes up, I'm on call, I have to go work on a component who knows anything about this component? Who wrote this component? What are the documentation details about it?

It gives me all that information to reduce the toil. Mm-Hmm. It all comes back to your branding.

Mitch, you were, uh, boy, you're, you're like the soccer team. Stop right Here. I gotta go register that mean for you.

Does That answer your question? Yes, It does. Yeah.

Well, you know, you think about, you know, today folks use copilot and to actually be, to explain code Mm-Hmm. They think about now how can you productively both make that into documentation? Then again, I don't necessarily wanna read all that documentation.

I just wanna know the few essential things that this does make it easier to understand. It seems like that sort of balance is a place that AI could help us maybe a lot of other places Too. ai, uh, yeah, we're laying it all over the portfolio.

Um, some of that is about taking a massive amount of text and boiling it down to that B brief be gone that I'd mentioned before. Mm-Hmm. You know, we're doing that in JPD or Jira Product Discovery acronyms, uh, JIRA Product Discovery in, in terms of taking all this like, content that a product manager may write and succinctly write that down as something a human can consume.

Uh, we're also doing that, um, in Jira Align as I mentioned before. And then, um, some of the other things actually this funny, my favorite feature is just this thing called AI definitions. So you asked me, I you corrected me when I said JPD Mm.

Holy cow. We have so many acronyms. Yeah.

And just the ability to click on an acronym and understand a definition, uh, that AI can generate for me and all the related information across our teamwork graph that's relevant to that of like, who works on that thing? Mm-Hmm. What else has been written about it?

What are the, what are the, um, artifacts associated with it? In my day-to-Day job, it's been one of the most incredibly helpful things, uh, especially when I came on new to Atlassian. Yeah.

It was just an unbelievable time saver. So just Some of those acronyms, there's two or three definitions. I mean, you Go down rabbit Hole Exactly.

Over in finance. They have another definition for Yeah. And that's as long as we can connect to it through the graph, we can define it.

Which is, which is just amazingly Powerful. So while you're here, I'm gonna steal a little bit of your thinking. Um, I'm working on an analyst report, it's called, it's called DevOps Next.

Like where are we going next with DevOps? What, what's your outlook for DevOps, say in the next 1224? Maybe it's hard to predict fast a week sometimes, but what, what's your view on kind of where we're going with DevOps?

Where am I going? This is a, so it's an open, this is a 35 minute, uh, answer. You Can't be wrong.

'cause I don't have the right answer to tell you. You're wrong, so you're, you're good. Yeah, yeah, yeah.

You know, I, I really think what we're gonna see more and more is a movement left of code to be able to have a, um, less skilled knowledge worker be able to create powerful, powerful code. We're gonna see a massive amount of automation across the pipeline. Uh, and it's not about, and I think a lot of this is not about replacing people.

That'll be a huge fear about this. But it it'll be about, um, how do we expand the amount of velocity we can deliver Mm-Hmm. With the same team without having to try to double this, like high skilled, you know, high, high touch development, uh, for organization.

Mm-Hmm. So I think you're gonna see massive amounts of automation, uh, crazy innovations around AI that just get us from the feedback signals that we hear from customers that we can listen to all these different channels to get that all the way to a PR in, uh, you know, a fraction of the time that it takes today. We were talking earlier, some other folks about citizen developers kind of AI is the new low code no code.

Yeah. In a way, I mean, in terms of enabling more people to contribute. Yeah.

I, you know, we've talked about citizen developers for a long time, right. And these low code platforms, people didn't have a patience for them. It's still, it's still quite a bit of work to, to do.

Um, Spreadsheets were the first low code no code, and they are, they still gotta learn a lot to know how to leverage that, right? A hundred percent. A hundred percent.

But I do think, like, one of the ways that I, I would take that would be, you know, we, we talk a lot about Jira here and the coming together of Jira software and Jira work management into a single product, right? Uh, I would absolutely see a marketing team, for example, being able to initiate the code creation process now. Mm-Hmm.

So, uh, a different take on citizen developer, it still starts with an issue or a request, like a business need that they have build Be the MVT or Something. Build be an MVP. I'm gonna get to that so much faster Yeah.

Than I did before. I think the Other piece, think of the time we would save not going down 20 paths, but get there faster. Right.

Fail fast. Yeah. So the, the, the, just the velocity of innovation of what we can create, iterate on and maybe fail fast.

We've had this mentality of, you know, agile is about, you know, iterating, failing, fast learning from that. Yeah. That cycle is gonna get exceptionally faster.

So if I was, and you don't know about me as your customer, but if I was one of your enterprise customers and I had a chance to sit down with you, Matt, what, what, what would you say, this is what I think you should be preparing for, preparing your organization for. Let's say we're, you know, if we're moderately good, we've been doing DevOps for a while. It's not a new thing here, but we, it's coming faster and more for more direction of all the stuff we gotta do.

What, what kind of things are would you advise your customers to kind of get yourself prepared for this is what you can do to help get to that kind of a future? Uh, first of all, I would be clear on what are the key platforms you're gonna depend on the organization. I would, I would try to simplify the ecosystem across development teams.

Some organizations are still so exceptionally pragmatic And it's hard. Yeah, It's super hard. But consolidation and standardization have real value.

Uh, that's one. Um, I would get clear on what you need to own as a development team and what you can source out. So understand what's your core differentiator and focus your high, most high value developers on that.

Uh, and then I would get, I would get super comfortable on leveraging ai. I talk to a lot of, uh, that sounds kind of like, no kidding. Well, so many companies are risk averse about, they're worried about, you know, what's happening with their data.

Uh, you've gotta lead into it because just the rate of change around ai, uh, it's not a fad. This is not, um, this is unlike anything I've seen and oh, it, I had a lot of hair when I started using technology years ago, right. It's, uh, it's, it's exceptionally transformative and, uh, it's gonna disrupt every industry.

So you gotta get in front of it. Now, I have a deal for you. Um, you are welcome to use toil or however you wanna brand it.

If I can use your extending the Harborview left of kind of into the product organization, how things get created. As part of that, I'll have my lawyers call your lawyers and we'll negotiate the Whole, I'll tell you what, just we'll write it on the back of this. All right.

This, uh, card and we're all right man. It was great talking to you Matt. Appreciate It.

Alright, thanks again. Matt Swimmer who is, uh, head of SVP of, uh, agile and DevOps here at Atlassian. So we just plugged in some free advice for you there.

I was trying to do my best to channel, uh, what a customer might, might ask for. So I'm sure there's a lot of great information that you could get. Good luck.

Um, I'm glad you're doing the developer, the developer experience session. Thank you so much. Folks will get a lot out.

I'm in Thursday. Okay, alrightyy MasterFiber. We will be back with another great interview.

Uh, the talent here is just phenomenal at no end. Every person we talk to, we just bring something to the table you can't get anywhere else. So thanks to Matt and others, we'll be back in a minute.