Linear CEO Karri Saarinen declares 'issue tracking is dead' and reveals Linear's AI-era pivot

I'm a big fan. So, thanks for having me on.

Thanks so much. We'll talk to you soon, John.

Great hanging.

Have a good one. Let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest. Securely connecting bank accounts to move money, fight fraud, and improve lending now with AI. And without further ado, we have Ki from Linear in the Richream waiting room. Let's bring him into the TV. How you doing?

What's going on?

I'm doing great. How are you guys?

We're doing well.

You've been uh stirring up the internet this week.

Cooking. Cooking. It's It's my favorite hobby. It's like It's definitely like something I like to do.

Yeah.

Well, we're excited to we're excited to talk about it. So, yeah. Take us take us through kind of the the announcement this week, what what led to it,

all that stuff.

Yeah. So, so this week at Liner, we we kind of like wanted to share more about like what is our thinking and kind of read on the market and and in our category, which which kind of historically has been this issue tracking, project management. And like I I wrote this essay that like kind of like tries to like talk through some of these topics and and it starts with this like some of the like controversial statements saying like issue tracking is dead. And I think what I was trying to do with it is is to really like shake the industries and everyone's thinking around this. And I think what still is happening is that how companies and software companies operate, they still are operating this old assumptions and like kind of processes that are built around the old world like the world before AI agents and and and that kind of development workflow. So I think like historically is tracking has been this ticketing uh task system for engineering. So basically engineer doesn't do anything unless there's a ticket and then like you have to like create the ticket, someone else creates it, it goes goes to engineering, there's some negotiation happening. Eventually, the engineer does something with it and it's kind of like a it's almost like this line cook system where it's like someone yells like we need fries and and steak and then that then in the kitchen they start cooking it

and it's it's kind of dumb and like I I always thought it was dumb but it's just like how a lot of companies tend to operate. And the second problem I think is that these legacy platforms vendors have kind of like created this this kind of idea that it should be complex and like they kind of encourage the complexity and the overhead of this process. So we have customers coming for us to us and like hey we want to switch to linear but then they say like oh we have this like bug tracking process which is 40 different states and the buck report needs to have this 30 different custom fields and then they ask like can you do that and we our answer is like I don't think you we should or you shouldn't either like the point of moving to linear is actually rethink some of this stuff like is this still necessary

yeah and I think like this becomes even more clear now is that if the agent can fix the bug in 5 minutes, but your bug reporting process or the the process of running through that process takes a week, it starts to be look pretty silly that like the actual work takes a lot less time than actually running through the whole process.

And you're sitting here being like I don't want to be like part of the bottleneck.

Yeah. And so I think like my thinking here is like I think we can start kind of compressing and simplifying a lot of these workflows and even I know that companies do need some process where it's like compliance or security or some some other reasons that you just can't remove. Um I think like the agent can like really handle that. Um but like this this statement was really for us to say that we I think we should all like kind of look into mirror. It's like what are we doing here and like why and do we still need to do these things and and from our perspective our perspective we don't um I think we should start looking into different problems and the the different problems is like yeah like how do you work with the agents and I think the problems I noticed today is one is like the context problem like how does the agent know what to work on or like how do you set up the agent to work on with the right context and we think like linear can be that or is already that context system for humans so It can also be the contact system for uh the agents. So, linear collects c like user bugs. It collects like user feedback or customer requests. Um there's also kind of projects. You have a goals, initiatives. There's a lot of context of like what is your product? What is the state of your product and also like what is the future state of the product and then if you can bring that into the agendic like workflow, I think the agents can work a lot better. So I think at linear we want to like look into that category of like context problems and try to like find solutions there. And then the second thing second problem like I noticed is the the agency part that a lot of these systems historically has been quite static like it doesn't do anything unless you move the thing from one state to another. And we think like now the system should be with this right context and understanding of what you're doing and what your team or product is doing. it could start moving the work uh on its own like it should become more self-driving and so this for example this bug reporting example is like that if the bugs get reported in linear they should just like the linear agent should look at it like start debugging it connect to other tools make a fix then eventually at the end like tag in a human to like hey can you authorize or verify my fix and so then this like 40 state buck tracking process is now like one state because the human only has to like verify it and even if it's not correct solution, you still have the option to fix it and work with the agent. So I don't think we can onesot problems all the time. But that's just like the direction we want to go. It's like is there work that linear could do automatically that the humans don't even have to look at and in and at least at the very beginning of the state and linear could start doing it for you. Um and the last thing I'll say like I think that um we we are launching like several things over the weeks that the we launched this like linear agent which it's a a surface for linear. So you can ask things you can like how I use it is I get feature requests from people then I go there it's like I wonder what other customers are saying about this problem and like can I understand the underlying need because I think the problem today starts to become like if you can build anything fast the problem is not building things anymore. The problem is like deciding what to build and why and even understanding like what is the problem. So I think like linear can be a system like where you can understand the problem, shape the problems and then pick a direction or solution and then start working on the the the like have the agent work on it with that like correct context and direction. So um and then like in the in the future we'll also like we're working on our own coding agent which I think there was some like confusion about it and I think like the goal there is like it's it's it's more like a wrapper around this existing coding agents and what do we do there is like we can like dog food our own platform which we have this um like agent platform that third parties like Codex and and and Devon and cursor have uh participated in and like they have their agents in there but like because we are building our own, we can kind of like create the capabilities and the tooling and the hooks or whatever we need better because we also notice the opportunities as we build these things. And we also just generally see that there is probably like situations where the linear coding agent can be better than some of the like generalized coding agents. Yeah, I I think I think it's I think it's really smart to to you know, if if linear had been started, you know, uh 10 years prior, you would have been able to just like add like do the the legacy the legacy SAS playbook of like let's just try to add a lot of seats every month and compound forever and that was a beautiful existence and now

our entire industry is like disrupting itself and so you can choose I think Toby

Liy on David Sra like every category is up for grabs. There's going to be an AI native company in every single category. You can wish that this wasn't happening or you can decide to be that company and the mature hard thing to do is like admit that things are changing, embrace it, rally the team around it, rally your customers around it. I think it makes a lot of sense. Um, how do you think taking a step back like how do you how what's your view on how like enterprise software categories are evolving because I think a lot of people are are kind of like waking up to this idea of like everybody wants to be at least in software development in the in the pipeline of tools everybody now wants to be the same kind of like the same thing. you want to be the place where people go to uh you know encourage the machines to do things right and kind of do that and so um and how do you see linear's role evolving you know you've been very open to different agents leveraging the platform I'm ass sure you know you you'll continue to be but um but yeah how how are like software categories evolving in this era of like you kind of need to be selling everyone's trying to sell perk.

Yeah, I don't I don't think I have the complete answer there. I think there's a lot of nuance and it's like depends on the markets. I I do think like that's that feeling of everyone's building the same thing. It's I think partially comes from a lot of people are noticing the same primitives and and from like a same kind of blocks or pieces of this puzzle and then they're trying to like figure out like what is the right setup for that like what is the right order of those things and like who is actually the best thing best one to like own part of that puzzle and I think that's like the the feeling probably there but I think like my feeling my my thinking is that it's similar to earlier software days too is Like it's like when people invented databases now everyone's like why is everyone making databases like well

everyone wants to be a system of record you know.

Yeah. So I think there is like a lot of like there can be a lot of shared primitives across products. It doesn't mean that every product still like has the same outcome or like the same advantage or something. And I think like we we we don't know like how this will all play out. Mh.

And I think the other sentiment I'm seeing is that there's this like idea of like let's throw everything out like no SAS at all. And I think that's an option, but I I think like you can do that and then you start like inventing stuff back from the first principles and you maybe end up like in the same spot again. And I think like that's what we see in our space. It's like everyone's running their agent and now they have 10 agents running. So then they put them on a canban board which linear has and then it's like oh now I invented the agent orchestration like no you invented the canban board which has been like around for like 30 years but you just put the agents on a canvan board. So like kudos to like connecting those two topics but like it's not like that some things don't need to be reinvented like some things can still work but like if you just throw everything out you kind of like start from the beginning and then I think it's a long journey to like figure everything out again.

Yeah. I got a I got an investor update last year that was just like workflows have been rebranded in the product as agents and that was a funny moment because I was like okay some of this stuff is is you know very new approach to software but some of it is just kind of re some of it is just repositioning.

Uh I have one more question but first we got to give a shout out to Russell Smith from Y Combinator summer 2012. We're not beating the YCS12 uh mafia allegations on this show uh since I believe we all went through that one. But uh my last question is um uh you know in in we we we we talk about linear every day on the show 75% of uh enterprise work enterprise companies are using agents. Have you talked to the other 25%? Like what is unique about those businesses that they haven't jumped in? Are they just like asleep at the wheel or is there something unique about their business where they're sticky and for security reasons or some other reason

where they're using agents but haven't integrated them?

Maybe I'm just because yeah it feels like you know this is such everyone understands how powerful agents are. Why is it not 100%. That's always an interesting question.

Yeah. And to clear the stat is like like 75 77% of like linear workspaces have cloud agents.

Okay. So, so there's a difference between cloud agents and the local agents. So, we don't track the local agents, but okay.

So, I think

the couple things like the cloud agents are still somewhat like less developed today. There's like environment problems where like if you have a massive system or some kind of large company system, the environment starts to become kind of complicated and the industry hasn't fully solved it yet. I think it's solvable problem and many companies are kind of like solving it with the sandboxes of like you start a sandbox the agent has the right environment and can do the things. So there's like just like a timing piece a similar timing piece not everyone has like decided to do this and like go into the cloud agents.

The third piece which is it's is is kind of the stupid reason is is the uh policy. Um so

companies are still

procurement IT and and securities still run a lot of like decision making when it comes to tooling in companies and there is a lot of companies that like the employees ask can I use the agent and they say no you can't and they don't have like a reason for it or like we have to evaluate it and that evaluation takes months and I think like what the companies need to look into here is like what this has to change like you can't like like the industry is moving so fast so you cannot have like a month-long like many monthlong processes to evaluate every single new tool like I think like the the security standpoint is like valid but like you kind of have to like figure out a new way to think about the security and not like block new tools. So I like one hack to this I've seen in some companies is like they have this friction form

so engineers

that want to use like feel of friction

Yeah. and that they want to use new tools or AI, they form a fiction form like file a fiction form and then that escalates it to the higher leadership and then it makes a SLA that has to be resolved in a week or in 10 days or something

and so you force the hand of this like procurement IT and security people to like do something about it because the incentive a lot of times for this jobs is like you don't want to rock the boat so you don't want to change things if you don't have to. So, so then like the companies do I think that leadership should recognize that like we do have to like allow people to to use new tools otherwise we will kind of fall behind here.

Well, thank you so much for taking the time to come give us the update. Always good to hear from you and uh we'll talk