Enrique Gonzalez
Head of Support Engineering at Granola
Sourcegraph
AI-powered code intelligence platform helping developers search, understand, and automate across large codebases.
Industry
Developer Tools
Support team
20
Monthly Volume
4000
website
www.sourcegraph.com
MIGRATED FROM
Zendesk
Engineering teams working across codebases with billions of lines of code use Sourcegraph to search, write, and understand what's there. When their Head of Support needed to unify roughly 100 enterprise Slack channels, replace a three-tool support stack, and bring AI into the workflow, he migrated 26,000 tickets to Plain and built the integration with his support team. No dedicated engineering resources required.
Three Tools, No Source of Truth
Enrique was running support across three disconnected systems: Zendesk for ticketing, Foqal as the bridge to Slack, and [NAME] for AI categorization. None of them talked to each other cleanly.
Enterprise customers could post in Slack and wait hours before a ticket was created. Someone had to spot the message, click an emoji reaction or fill out a form, and wait for the automation to catch up. When agents did reply from Zendesk, responses rendered poorly back in Slack, so engineers gave up on the official flow and started replying directly in threads, getting pinged on personal aliases all day.
Reporting was no better. Years of freeform tagging had made Zendesk data unreliable, and the team had fallen back to quarterly spreadsheets maintained by hand. Enrique had an AI strategy for categorization, routing, and assisted responses, but every new capability required another tool, another integration, another person holding it together:
He needed one system, and he had one more requirement: make it accessible enough that new engineers could come in, read a thread, and respond without having to ramp up on the tool.
“Plain gave us the tools to handle high volumes of support during a huge growth period. It’s transformed how we manage support and set us up to scale effectively as we grow.”
Jo Barrow
Chief of Staff at Granola
One Platform, Built on Sourcegraph's Terms
Sourcegraph switched to Plain in December 2025. The team migrated 26,000 tickets from Zendesk, brought roughly 100 enterprise Slack channels into Plain, and had email and Slack running within days.
Every Slack message, automatically a thread
Enrique wanted to get closer to the customer and, with Plain, every Slack message became a thread automatically. That shift brought risk with it: higher volume, more noise, more false positives. But four months in, customers are telling Enrique their issues were getting answered faster, and messages that started as FYIs were turned into support requests before they fell through.
Support Engineers no longer jump between systems to piece a conversation together. Handling a thread means getting the full conversation in one place: customer messages, agent responses, internal notes, and engineering escalations, with a customer card attached showing account data like subscription status and product tier.
For managers, it's a window into how their team thinks through problems. For anyone coming into a ticket late, the full story is already there.
Clean data from day one
Enrique rolled out AI in two phases:
Migrate and Stabilize. Get channels live, move the team onto Plain, and build a clean label taxonomy by product area and issue type.
Automate. Turn on AI labeling once the data underneath it was solid.
Plain's AI labeling replaced the external categorization tool Sourcegraph had been paying for, bringing three systems down to one. Because the labels were clean from the start, reporting worked too. The quarterly spreadsheets maintained by hand were replaced by labels that fed directly into Plain's reporting, giving Enrique ticket trends, workload distribution, and resolution data without exporting anything.
An integration layer without a dedicated engineering team
Enrique and his support team built their own integrations using Plain's API. One example: a Deep Search bot that queries Sourcegraph's codebase to help troubleshoot customer issues.
The bot takes the full Plain thread as input and searches Sourcegraph's codebase, log activity, and database for what might be causing the problem.
This pattern is something many Plain customers build toward, and Plain's Bring Your Own Agent feature is what makes it possible:
One integration point. Because Plain holds the entire conversation in a single thread, you only need to connect your tooling once.
Technical depth without engineering overhead. Agents can surface context from your own systems without pulling in an engineer for every question.
Support as a product feedback channel. Enrique has used bot results to bring threads to engineering leads and make the case for product changes.
Plain's API gives teams room to keep building without waiting on a vendor to ship what they need.
"I don't need to hook up Deep Search to Plain or to Slack. I can do it to Plain and it grabs everything I need."
Enrique Gonzalez
Head of Support Engineering at Sourcegraph
Results: Four Months In
67% improvement in Median First Response Time
Response time dropped from roughly 2 hours to 40 minutes driven by Plain's AI tooling, the switch to automatic thread creation, and not having to jump between three systems to handle a single conversation.
3 tools replaced by Plain
Zendesk, Foqal, and [NAME] were consolidated and replaced by Plain, saving thousands of dollars a year in licensing alone.
AI participation rate: 50% to 62% (targeting 75%)
Enrique reports this number to leadership weekly. AI is now involved in a growing share of threads through labeling, suggested responses, and the Deep Search bot. He's pushing toward 75% by end of quarter.
Less engineering time spent on support context
Engineers read the full thread in Plain instead of piecing together Slack and Zendesk histories. Even a 50% reduction in time spent per escalation adds up fast at an average engineering salary.
Where Sourcegraph goes from here
For Enrique, the next phase is solidifying Plain as the system of record for customer context. If the right data lives in the thread before a customer writes in, support stops being reactive by default.
He's building thread events into Plain so agents can see that a customer hit an error or requested elevated access minutes before they wrote in. Combined with the account data already surfaced in customer cards, agents arrive at a conversation knowing what happened, who the customer is, and what they're likely writing about. The conversation starts informed rather than cold.
What Enrique is building toward is where support is heading: a platform that works as infrastructure, not just a place to close tickets. Four months in, it's keeping up.
