Linear scrapped its enterprise roadmap to build agents—and won enterprise deals because of it
In 2025, Linear shelved its planned enterprise feature roadmap to focus entirely on MCP servers and agent platforms. This pivot won them enterprise customers the original plan never would have reached.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
To move upmarket, first build success selling to mid-market companies
Three weeks after joining, Cristina crossed out "move into enterprise customers" from the company's three main goals at an all-hands meeting. The company had only one salesperson and lacked capacity to execute this goal. Instead, she reset the target to first land mid-sized companies already using competitors, using them as reference customers for enterprise deals. This stepping-stone, defined in 2023, only formally upgraded to "move into enterprise this year" in 2025. The core logic: don't set goals you can't achieve at your current stage; first prove you can execute what's within your capacity.
— CristinaMore than half of new issues come from agents; human use grows too
Linear launched MCP servers in 2025, among the first in its category. Now more than 50% of issues created in Linear are created by agents, not humans. Cristina emphasized another concurrent trend: the volume of humans entering the app to create and interact with issues is also rising. This indicates the product works well on both the human interface and agent interface, not that agents replaced humans. She opposes approaches that "just turn your product into a database plus an MCP server," since companies then give users no reason to stay in their tool.
— CristinaDeclaring your product category dead let Linear rebuild its entire business
When the CEO wrote that issue tracking is dead, the sales team was negotiating contracts with customers. Cristina's explanation: at Linear's current scale, the company can afford to abandon old business definitions to pursue new directions—unlike Google, which would struggle to disrupt its own search business. Customers responded with "right, tracking work is dead—so what are you now, and what am I buying?" This forced Linear to reposition as a "product development system," not an issue tracker.
— CristinaCustomers hate qualification calls; skip SDRs by making first contact valuable
Linear has never had an SDR role. Cristina's reasoning: a typical SDR's first call just asks "how big is your company, what's your budget?" Customers hate wasting 30 minutes on that before seeing someone who can do a demo. Her alternative isn't eliminating this stage but using tools and automation so AEs can run outbound themselves, making the first customer touch valuable and customized, not a procedural qualification gate.
— CristinaAI-drafted support needs human judgment; automating responses breaks trust
Many companies have LLMs auto-answer all support questions, with humans stepping in only when customers complain. Cristina rejects this model: AI drafts responses, but humans must review for quality, since AI often turns a three-sentence explanation into three paragraphs of filler. She notes customers sometimes ask "is this AI, because it's so fast?"—the speed actually comes from team responsiveness. Using AI saves on hiring headcount, not on dropping human judgment.
— CristinaAI startups at $100M ARR mostly sold high-concentration deals, not growth
Asked why some AI companies hit $100M ARR in months, Cristina pointed not to growth speed but customer concentration. These companies sell to defense or large law firms, signing a few $30-40M contracts to hit a billion in revenue. Linear is 40,000 paying customers, historically all seat-based pricing—there's no path to hitting $100M overnight this way. This reveals how different customer structures hide behind the same "ARR growth speed" headline.
— CristinaAbandoning the enterprise roadmap for agents opened the enterprise market
The 2025 roadmap was originally all enterprise-market features. But as the industry shifted, the team realized they had to build MCP servers, new agent platform APIs, and integrations with Codex, Cursor, and GitHub Copilot—none originally planned. Cristina faced a choice: fall behind the new trend or run ahead of her own plan. She chose the latter. Most planned enterprise features went undone, but by shipping agent features early and well, enterprise customers who weren't originally candidates asked to be in scope.
— CristinaBuild company culture by setting engineering headcount ratio before hiring sales
Stripe once had sales hiring move faster than engineering and product. The founder halted it and set a rule: at least 40% of the company must be engineers, product, design, and data roles. If sales wanted to hire more, they first had to help hire engineers. Cristina notes this ratio is arbitrary, but what matters is clarifying "what kind of company do we want to be," then building headcount backward from that. She brought this lesson to Linear's hiring plans.
— CristinaIn their own words · checked verbatim
The great thing about our business is that we're not like so big to like kill ourselves, right?
Cristina0:00
So you have to think about, like, what are you capable of doing? Uh, what do you need to, like, learn in order to, to get there? And then, do you actually have the resources to, like, accomplish that goal, right?
Cristina13:06
now, um, over 50% of issues created on Linear are created by agents, um, in just, like, a year and a half.
Cristina20:35
but we probably shouldn't have a world where people are, like, manually tracking work.
Cristina24:50
we don't have that function because I don't think it works well for our customers, but also because you can create systems and tooling and automations that make it easier for an AE to do outbound to a customer, right?
Cristina36:20
I would much rather have AI, like, draft an answer, um, to the specific question based on the knowledge that, you know, it has coming in, and then a human can decide, like, is that a good answer or not?
Cristina37:56
if we want this company to be like an engineering and product-driven company, probably at least like 40% of the company should be like engineering, product design, data, et cetera.
Cristina1:09:24
Figures
| Largest customer before Cristina joined | ~500 seats | 3:11 |
| Revenue growth during Cristina's tenure | At least 10x | 4:18 |
| Share of new issues created by agents post-MCP launch | Over 50% | 20:35 |
| Linear's current headcount | ~170 people | 1:22:28 |
Glossary
- MCP / MCP server
- Protocol interface that lets external agents read and write Linear data.
- SDR
- Sales Development Representative; sales role focused on outbound prospecting and initial qualification.
- AE
- Account Executive; front-line sales rep responsible for closing deals.
- Forward Deployed Engineer
- Engineer embedded in a customer's systems to implement the product; Linear doesn't use this model.
- Loops
- Linear functionality that auto-assigns issues to agents and lets them begin coding.
How to listen
Startup COOs building sales organizations from scratch; operations leaders deciding whether to hire SDRs or outsource support; investors studying go-to-market for AI-native products.