← All episodes

Most SaaS Companies Got AI Wrong. Linear Waited.

| 10 products mentioned
Every Every host
Watch on YouTube product strategy ai integration saas business models workflow automation coding agents product development methodology organizational decision-making

Karri Saarinen, CEO of Linear, explains why the company waited years before integrating AI features—rejecting the rush to add chatbots that most SaaS competitors pursued. Rather than chasing token usage, Linear built agent-native workflows that position the product as the control center for AI-powered product development, creating a uniquely defensible business model where Linear owns the sticky interface and organizational context without bearing token costs.

Key takeaways
  • Most SaaS companies failed at AI by adding chatbots reflexively; Linear's advantage came from understanding which workflows actually needed AI before building features, then creating a platform that lets multiple agents integrate rather than owning the AI layer itself.
  • Linear's business model improves under AI adoption because it stays in the "decision-making upstream" position—collecting customer requests, bugs, and feature proposals—while agents execute the work, meaning Linear captures margin without paying LLM token costs.
  • The wrong metrics for AI productivity are token usage, PR count, and percentage of agent-generated code; the right ones are bug count, product quality, user love, and whether the output actually generates value.
  • Separate the problem-finding phase from the execution phase: spend time understanding the right problem and approach before committing resources, then use AI tools to accelerate execution once you've committed.
  • Linear is building a coding agent tightly integrated into the issue workflow so that bugs and feature requests automatically get spawned as agent tasks, with shared visibility for teams to collaborate on fixes in real-time.
  • Product development will shift toward "self-driving" workflows where agents follow explicit project guidance and rules (like a feature's memory and strategy), but humans must still do the thinking work—writing documents, setting strategy, and making judgment calls that no dataset can replace.

Recommendations (5)

Linear
Linear uses

"We were using it in 2020 at the very beginning of Every to act as our content management system for the newsletter. And at the time, it was like this very kind of like hush-hush, you couldn't get a..."

Every · ▶ 0:48

Claude
Claude uses

"I think we'll eventually have a lot of multiple models but now we use Claude for this."

Karri Saarinen · ▶ 39:22

Figma
Figma uses

"I still have Figma open and then when I have a problem or idea I just draw it in there. And it to me it's like I'm often like my work is often more like that kind of like exploring things."

Karri Saarinen · ▶ 19:30

Slack
Slack uses

"We use Slack a lot and like with our Slack agent like we have discussions and then we eventually decide like yeah, we should do this and then we just tackle in there."

Karri Saarinen · ▶ 21:03

ChatGPT
ChatGPT uses

"When I use this like Claude or ChatGPT or some of these tools or Codex is that like I have to really explicitly tell the agent or like the tool always like what context to bring."

Karri Saarinen · ▶ 33:10

Mentioned (5)

OpenAI Symphony "OpenAI came out with Symphony the other day, and the main thing that it hooks into is Linear." ▶ 2:11
GPT-3
GPT-3 "But like when GPT-3 first came out, I didn't see anything about that on Linear." ▶ 2:23
Codex
Codex "OpenAI brought their Codex cloud agent in there because we just had this available." ▶ 5:20
Coinbase
Coinbase "Companies like Coinbase and Ramp who are our customers. And they built their own homegrown coding..." ▶ 5:39
Ramp
Ramp "Companies like Coinbase and Ramp who are our customers. And they built their own homegrown coding..." ▶ 5:39