Key Takeaways

  • Linear reached over $100M ARR while keeping customer support entirely in-house, refusing the standard enterprise playbook of offshore tier-one outsourcing.
  • Linear uses LLMs strictly to generate internal draft responses rather than deploying autonomous chatbots directly to customers.
  • Support specialists review and edit every AI-generated draft to eliminate verbose, unhelpful bloat before the customer sees it.
  • This drafting workflow decouples customer support volume from linear headcount expansion while protecting customer trust.

The Default Playbook Destroys Customer Goodwill

When Cristina Cordova interviewed for the COO role at Linear, the founders drew a hard line immediately. “When I was interviewing, the founders were very clear that like they did not want to outsource support,” (play from 10:59) Cordova recalled. “And I've never worked at a company that didn't outsource support. So that was like okay.” (play from 11:05)

Most enterprise SaaS companies view tier-one customer service as an operational tax. They outsource tickets overseas or plug an autonomous LLM bot straight into the front door. When the bot hallucinates or misinterprets the ticket, the customer has to fight through canned text before getting help.

Cordova criticized that common approach directly: “A lot of other companies will say, hey like 100% of questions that come into our support team are going to be like answered by an LLM first. And only if that customer writes back angrily, right? Will a human come in and be like, oh sorry about that. Like let me give you a real answer, right? Um and we just don't believe in that model, right?” (play from 37:36)

Use AI for the First Draft, Not the Final Send

Linear resolved the scaling problem by changing where the model sits in the communication loop. Instead of placing the LLM in front of the customer, they placed it behind their own staff.

When an issue arrives, internal software drafts a suggested response based on Linear's documentation and ticket history. An internal specialist reviews the draft, checks its technical logic, trims unnecessary filler, and sends the final note.

The biggest flaw in autonomous AI replies is verbosity. LLMs default to pleasantries and generic explanations that waste time. “Very often it's like gosh this would have taken like three sentences and instead I have like three paragraphs of, you know, a reply that ultimately is like not the most helpful, right?” (play from 38:10) Cordova explained. Ambitious engineers and designers using Linear do not want three paragraphs of conversational filler. They want a bug confirmed, a feature request filed, or an exact setting toggled in two clean sentences.

Scale Efficiency Without Sacrificing Craft

Keeping support internal does not mean hiring an army of reps. Linear grew past $100M ARR with an unusually lean headcount by making individual staff members faster rather than relying on sheer numbers.

“I don't think we've ramped it in terms of headcount as we would have without it, right?” (play from 39:31) Cordova said. “So, I think some of these tools are are very helpful for enabling us to like not always have to grow teams linearly with like how how fast the business is growing, right?” (play from 39:36)

By turning AI into an internal drafting copilot, Linear stripped out the manual labor of looking up documentation and typing boilerplate answers. The human provides the final judgment, voice, and quality bar.

What to Do With This

Audit your customer support setup tomorrow. If you have an AI agent replying directly to incoming tickets, change its permission settings so it only saves responses as internal drafts for your team to review. Have your support leads measure how many sentences they delete from each AI draft over the next 48 hours to find where your prompts are generating unhelpful bloat.