If SaaS Is Dead, Linear Didn't Get the Memo

Welcome to the show. Quick scene-setter: the world’s tilting toward many agents per person and per company, and the sticky place to steer them is where work already starts. I first used Linear back in 2020 as a stealthy, beautifully crafted tool, and I’ve admired how patient and taste-driven you’ve been. Now you’ve made Linear feel agent‑native. What was that emotional and strategic shift like as AI took off?

Our mission did not change—we’re here to move work forward and help teams build. AI actually lets us shoulder more of the drudgery so people can apply judgment and taste. We skipped the rush-to-chatbot phase to really understand workflows, published an open agent platform, and now many external and homegrown agents plug into Linear. We’re the context and decision layer, not the whole stack, and we’re adding a chat interface where it’s genuinely useful, like synthesizing customer feedback into clear priorities.

Public markets keep declaring SaaS dead, and a lot of teams feel pressure to ship anything with AI on it. Did you wait for a fat pitch, and how do you see this playing out for big public SaaS companies?

We didn’t have investor pressure, but the market’s speed creates noise. We treat new AI trends as signals, not commandments. The broad SaaS obituary is simplistic; what’s true is that moats are shifting and incumbents will struggle to adapt. We’re operating with a day‑one mindset, re‑examining assumptions as agents enter product development. We’re about one hundred twenty people, with roughly half on product.

Inside your team, how did you bring AI coding into the workflow without turning it into hype or glorified auto‑complete?

We nudged adoption but avoided vanity metrics like percent AI‑generated code or PR counts. Output volume without value is noise, and token‑spend incentives can push the wrong behavior. We track activity as signals, then balance it with quality and impact.

If tokens and PRs are the wrong dashboard, what’s a better way to gauge real progress?

The timeless ones still matter—revenue, retention, and product quality. We also run a real bug process. Internally we keep a zero‑bug posture with a one‑week SLA; agents do the first pass on fixes, engineers review in Linear, and we ship cleaner builds faster. Quality over sheer motion.

What changed most in your personal and team workflows, and what surprised you?

I use agents to synthesize feature requests into root problems so we can decide what’s worth doing. For design, I still sketch manually in Figma to think; speed doesn’t help if you’re exploring. Across the org we prototype more, preview builds quickly, and use our Slack agent to turn decisions into actionable issues on the spot. Net effect: loops shrink, and action happens right away instead of next week.

You often argue for slower upfront thinking. How do faster tools fit that philosophy?

Go slow to choose the right problem; go fast once you commit. Speed‑running decisions yields prototypes nobody asked for. We want tight execution loops after we’ve agreed on the why.

I sometimes need to build five wrong things to understand the right one. Is that at odds with your approach?

Exploratory building is fine when it’s framed as concept work. Think of it like a concept car—meant to spark direction, not ship tomorrow. We’ll completely reimagine a surface to test merit, separate from implementation risk, then decide what to carry forward.

How has AI changed your product strategy? Should you rely on third‑party agents, or bake your own into the core?

We’re doing both. Linear now has an agent that understands your org and product context for PM and design workflows, plus a coding agent with a guided cloud environment where you can inspect diffs and steer changes. We’ve always seen Linear not as kitchen tickets, but as the backbone that captures signals and decisions. Agents can triage and complete tasks, but context and prioritization still live with us. Relying only on external agents limited what we could shape, so we’re building native paths where context flows end to end, including running safe automations in Linear while you work locally.

From a business angle, I thought you could avoid token costs by staying the hub. Now that you’re adding your own agents, how do you think about margins and pricing?

Core assistant features will be bundled, but coding will be usage‑based since workloads vary. We’re not becoming a generic agent platform; we’re a focused product‑memory hub that other tools can tap and that injects the right context so users do less orchestration. For example, I can ask the coding agent to create a new app theme; it spins a sandbox, opens an issue, and we can co‑review diffs in one place, with the session visible to everyone for quick collaboration.

This expands your surface area into spaces others already chase hard. How do you avoid becoming a kitchen sink?

We stay upstream where leverage is highest—where work and bugs originate—and automate the routine. We won’t replace every coding tool or build net‑new products on command. Our rule is to follow the natural next step of the workflow: from issue to fix to diff to build, integrating just enough to make teams faster without bloating the product.

Five‑year view: how does product development change, and what remains constant?

Expect self‑driving elements guided by rules and project memory, where a feature behaves like an agent—spotting patterns, proposing changes, testing them with users, and looping humans in where judgment matters. Roles will shift, but craft, intuition, and clear strategy stay central. I don’t see thinking outsourced end to end; the best products still come from taste and focused decisions, not pure A/B autopilot.

Loved this. Thanks for coming on.

Before you go, hit like and subscribe to AI and I. Strap in for more sharp, high‑energy dives with Dan at the helm.