Rebuilding Customer Support for the AI Era
Harj Taggar (host), Advith (guest), Marty (guest), Robert (guest), Harj Taggar (host)
In this episode of YC Root Access, featuring Harj Taggar and Advith, Rebuilding Customer Support for the AI Era explores pylon’s pivots to AI-native B2B support platform from Slack wedge Pylon is an AI-native B2B customer support platform (replacing Zendesk/Intercom/Service Cloud) that has grown to ~79 employees, 1,000+ customers, and eight-digit ARR after rapid scaling post-YC W23.
Pylon’s pivots to AI-native B2B support platform from Slack wedge
Pylon is an AI-native B2B customer support platform (replacing Zendesk/Intercom/Service Cloud) that has grown to ~79 employees, 1,000+ customers, and eight-digit ARR after rapid scaling post-YC W23.
The founders describe “pivot hell” and how their idea-validation process matured from building too early to disciplined, high-volume user discovery guided by The Mom Test.
They intentionally optimized for a huge, horizontal B2B market by benchmarking $10B+ public SaaS companies, doing bottoms-up TAM models in spreadsheets, and hunting for a compelling “why now.”
The initial breakthrough came from the emerging trend of B2B companies supporting customers inside shared Slack channels, creating visceral workflow pain and fast early traction via founder networks and cold LinkedIn outreach.
As AI capabilities surged and incumbents stagnated (e.g., Zendesk post-PE), Pylon expanded from a Slack-to-ticketing integration into a full system with AI-assisted workflows, aiming long-term to become “more like an AI Salesforce than an AI Zendesk.”
Key Takeaways
Pick co-founders for long-term fit, not the initial idea.
They found “idea-tied” co-founders constrain necessary pivots and can block trust/equity alignment; durable founder chemistry matters more because the company will evolve.
Your idea search improves fastest when you shorten iteration cycles.
They saw a step-change after quitting jobs: questions and hypotheses evolved daily (not weekly), compounding learning speed through more frequent discovery-loop updates.
Use bottoms-up sizing to kill emotionally appealing but capped ideas.
Their Google Sheet comparisons showed some ideas topped out at “high tens of millions,” while horizontal B2B markets (e. ...
Look for a “why now” that reflects a workflow shift, not hype.
The Slack-shared-channel trend was a concrete behavioral change (B2B communication moving from email to chat), producing immediate, repeated pain (broadcasting outages, losing context, manual copy/paste).
Early traction can come from a narrow integration wedge that proves urgency.
They built the first Slack-to-ticketing integration in about a week for Hightouch, validating willingness to deploy/pay before attempting the broader platform rewrite.
Treat “this failed before” as a prompt for differentiated timing/execution—not a veto.
They were “the 13th” attempt, but believed post-pandemic communication patterns plus rapidly improving conversational AI changed feasibility and customer urgency.
In B2B AI, prioritize correctness and relationship-preserving workflows over full automation.
They argue B2B has low error tolerance (a million-dollar customer can’t be answered wrong), so AI should often assist humans (drafting, structuring, routing) rather than fully replacing them.
Notable Quotes
“We replace products like Zendesk, Intercom, Salesforce Service Cloud, but very AI-native and built for a B2B persona.”
— Marty
“We made all the mistakes, going through Pivot Hell.”
— Robert
“You shouldn’t base your co-founders based on the idea.”
— Advith
“It took 30 seconds to find prospect number one after that phone call.”
— Marty
“For B2B… if a million-dollar customer asks you a question, you can never answer them wrong.”
— Advith
Questions Answered in This Episode
When you realized Slack-based support was the wedge, what were the exact “must-have” MVP requirements you used to beat other tools for Hightouch?
Pylon is an AI-native B2B customer support platform (replacing Zendesk/Intercom/Service Cloud) that has grown to ~79 employees, 1,000+ customers, and eight-digit ARR after rapid scaling post-YC W23.
You mentioned you modeled 26 public $10B+ B2B SaaS companies—what patterns (pricing, buyer persona, distribution, retention) most strongly guided your eventual choice?
The founders describe “pivot hell” and how their idea-validation process matured from building too early to disciplined, high-volume user discovery guided by The Mom Test.
How did you decide the Slack integration alone wouldn’t become a big company—what specific ceiling signals did you see (ARR, attach rate, churn, expansion limits)?
They intentionally optimized for a huge, horizontal B2B market by benchmarking $10B+ public SaaS companies, doing bottoms-up TAM models in spreadsheets, and hunting for a compelling “why now.”
What were the biggest product risks in moving from “integration” to “system of record,” and what did you ship first to avoid rebuilding Zendesk before you had PMF?
The initial breakthrough came from the emerging trend of B2B companies supporting customers inside shared Slack channels, creating visceral workflow pain and fast early traction via founder networks and cold LinkedIn outreach.
In B2B support, where do you draw the line today between agentic AI and human-in-the-loop workflows, and what failure modes worry you most?
As AI capabilities surged and incumbents stagnated (e. ...
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome