YC Root AccessRebuilding Customer Support for the AI Era
CHAPTERS
- 0:05 – 1:56
What Pylon is today: AI-native B2B customer support platform (scale, customers, funding)
Harj introduces the Pylon founders and frames their story as a multi-pivot journey. The team explains Pylon’s product positioning versus incumbents and shares current company scale, notable customers, and funding history.
- •Pylon replaces Zendesk/Intercom/Service Cloud with an AI-native B2B support platform
- •Company stats: ~79 people, 8-digit ARR, 5.35x growth last year, 1,000+ customers
- •Notable customers: Honeycomb, Linear, Incident.io, Applied Intuition, Deel
- •Funding: YC + Seed (General Catalyst) + Series A (a16z) + Series B (BCV), $51M total
- •Setup for the core theme: long path to the right idea via pivots
- 1:56 – 4:41
Advith & Robert’s origin story: Caltech, hackathon ops, and startup pull
Advith and Robert describe meeting at Caltech, bonding through building/operating their hackathon, and gaining early exposure to startups via internships and KP Fellows. They explain the personal “can’t not do it” motivation that pushed them toward founding.
- •Caltech connection: puzzle competition → strong working relationship
- •Hackathon operations as ‘running a company’ training (team, fundraising, logistics)
- •KP Fellows + internships (DoorDash/Slack) as founder exposure and learning
- •Decision driver: felt time was being wasted in big-company roles
- •Pandemic-era trigger: Advith’s existential moment prompts serious startup efforts
- 4:41 – 6:27
Marty’s path: from music dreams to founder mindset and leaving Airbnb
Marty shares how media and self-learning made startups feel accessible, then explains why he left Airbnb. He outlines his experimentation with many side projects and the “best case / worst case” framework that helped him take the leap.
- •Inspiration shift: movies like The Social Network made founding feel attainable
- •Early programming curiosity and self-teaching leading into startup ambition
- •At Airbnb: realized he didn’t want the standard career ladder outcomes
- •Ran many side projects across wildly different domains to explore ideas
- •Decision tool: explicitly writing best/worst outcomes to de-risk quitting
- 6:27 – 10:20
Starting while employed: COVID remote advantage, ‘Grind Time,’ and first edtech mistake
The founders describe how they explored ideas while holding full-time jobs, enabled by remote work. They recount their first serious attempt—an alumni portal for career advice—and the classic early mistake of building before validating demand.
- •Remote work enabled early-morning and late-night startup cadence
- •Calendar blocking (“Grind Time”) and stuffing days with discovery calls
- •Used The Mom Test approach and started with familiar user group: school admins
- •Built an alumni portal concept for a college career center
- •Key failure mode: built too early; learned about validation and edtech headwinds
- 10:20 – 13:25
Pivoting philosophy: refining validation loops and learning from dead ends
They compare approaches to idea testing and explain how each pivot improved their process. The key insight: many ideas can work with enough execution, but the team needed a method to converge faster on a truly large opportunity.
- •They lacked personal pain points, so sourced problems from networks and jobs
- •Each pivot taught a new lesson (especially: validate before building)
- •Improved idea-validation process iteratively rather than betting once
- •Observation: some explored ideas later became successful companies for others
- •Speed and diligence can make many ideas viable—selection still matters
- 13:25 – 18:55
Forming the founding trio: rejection, misalignment, and the ‘don’t pick cofounders for the idea’ lesson
Marty describes trying to join Advith and Robert, initially getting turned down, and the awkwardness of pre-idea teaming. They unpack what they learned about choosing cofounders, why domain-based cofounding can block pivoting, and why “passion” can mislead.
- •Marty sought cofounders based on who he wanted to work with long-term
- •Advith/Robert initially resisted expanding the group pre-idea (coordination cost)
- •Near-miss with logistics ‘sales-veteran’ cofounders: equity + trust issues
- •Core rule: optimize for relationships and adaptability because pivots are likely
- •Caution: domain-passion can blind you and reduce willingness to pivot
- 18:55 – 24:12
Selecting a big market: working backwards from $10B outcomes and comparing ideas on a spreadsheet
They explain a structured approach to market selection: define “big company” quantitatively and analyze public B2B SaaS winners. By independently researching different ideas and then debating them, they used bottoms-up analysis to see market-size ceilings clearly.
- •Defined ambition: target $10B+ company / $1B revenue as a planning anchor
- •Studied a small set of $10B+ public B2B SaaS companies to infer patterns
- •Each founder brought an idea; they debated and ‘battle-tested’ them
- •Built a Google Sheet: bottoms-up market sizing tabs per idea
- •Lesson: horizontal SaaS markets can be orders of magnitude larger than vertical niches
- 24:12 – 25:49
The ‘why now’ and the breakthrough insight: B2B support moving to shared Slack channels
The team lands on a compelling ‘why now’: businesses increasingly communicate with customers in Slack, breaking traditional email-based support workflows. They validate quickly through their founder network and recognize broader multi-channel implications across Teams, Discord, WhatsApp, etc.
- •Trigger insight: inter-company communication shifting into Slack post-pandemic
- •Visceral pain: broadcasting outages or updates requires manual copy/paste per channel
- •Validation via network: many startups were already in-market for a solution
- •‘Why now’ mindset: tie a big category to an emerging trend you can ride
- •Broader thesis: email-era workflow models don’t fit new conversational channels
- 25:49 – 30:27
Discovery engine: LinkedIn outreach at scale and how they ran calls to surface real problems
They describe their daily outbound routine—40 personalized LinkedIn messages per founder—and how low response rates still create a full pipeline. The chapter details their call structure (tools, metrics, hates, boss priorities) and how quitting jobs increased iteration speed from weekly to daily.
- •Systematic outreach: 40 personalized connection requests each per day
- •Acceptance rates were low (1–5%) but still yielded enough discovery calls
- •Call format: map responsibilities, tooling, incentives, pain points, and metrics
- •80/20 rule: probe your current hypothesis but leave room for unexpected insights
- •Quitting jobs increased cycle time dramatically: iterate questions daily vs weekly
- 30:27 – 40:46
Early validation and ‘don’t do this idea’ warnings: passing the Mom Test and ignoring naysayers (wisely)
They explain how the idea passed the Mom Test: prospects were actively searching, attempted in-house solutions, and found vendors lacking. Marty recounts immediate validation from Hightouch—even though they warned the idea had failed before—leading to the principle of evaluating ideas from first principles and understanding why it can work now.
- •Mom Test in practice: customers actively looking, tried in-house, evaluated alternatives
- •Hightouch as instant proof point—and a cautionary ‘we tried this, it failed’ story
- •Advice: many failures are execution/cofounder issues; don’t overweight prior attempts
- •They estimate being ~13th team on the idea; several prior teams later became customers
- •Their ‘it’ll work for us’ rationale: timing + strong execution + compounding trends
- 40:46 – 48:37
YC entry story: using a two-week deadline to close a first customer and sharpen the pitch
They share a non-traditional path into YC via pre-idea office hours, then negotiating time to close a customer before the interview. They describe the scramble to win Hightouch, the fast-paced 10-minute interview, and why they chose YC (community, learning velocity) despite equity concerns.
- •Pre-idea outreach led to a YC conversation and a fast-track interview offer
- •They asked for two weeks to close a customer—used as an execution deadline
- •Closed/beat competing tools for Hightouch with a scrappy Gmail-level setup
- •Interview prep: rapid-fire question simulator; emphasis on knowing your space deeply
- •YC value seen as community + speed-learning; alumni strongly recommended it
- 48:37 – 50:47
From Slack integration to full platform: the product expansion path and PMF signals
Pylon started as a narrow integration that piped Slack support into Zendesk/Intercom, reaching ~400K ARR in year one. Customer pull—requests for core ticketing features, startups wanting Pylon as the system of record, and dissatisfaction with incumbents—pushed them to build a full B2B-native support platform.
- •Year-one wedge: ‘Slack channel support’ → tickets into existing systems
- •Early traction: ~400K ARR from the integration-first product
- •PMF signals: customers demanded adjacent core-ticketing features
- •Market opening: Zendesk PE acquisition + slower innovation + modern expectations
- •Strategic shift: build a complete B2B support platform, not just an integration
- 50:47 – 54:37
Becoming AI-native: skepticism, sequencing, and what works (and doesn’t) for B2B support
They explain how they adopted AI pragmatically: not as a buzzword, but as the best tool for specific pain points. In B2B, stakes are high and personalization matters, so they emphasize human+AI workflows, careful sequencing, and leveraging their core conversational dataset where LLMs add structural value.
- •Started skeptical of AI; added it only when it solved concrete problems better
- •Separating hype from signal: customers ask for ‘magic AI’; you must ground in workflows
- •B2B constraints: low tolerance for wrong answers, high need for personal touch
- •Core advantage: support data is unstructured conversation—ideal for LLM structuring
- •Advice to incumbents: identify core data/workflows, pick low-risk AI wedges, and sequence rollout as tech improves
- 54:37 – 57:47
What’s next: expanding from ticketing to customer success/CRM, plus scaling the company intentionally
They outline a long-term product vision: move from ticketing to a broader customer-facing workflow system, trending toward an AI Salesforce-like CRM built on conversational data. On the company side, they discuss multi-region expansion, adding leadership layers, and adopting ‘normal’ management processes only when they understand the first-principles need.
- •Product direction: ticketing → customer success/account management → cross-functional workflows
- •Use cases: feature request intelligence, product feedback loops, targeted customer outreach
- •Long-term framing: less ‘AI Zendesk,’ more ‘AI Salesforce’ (CRM-like system)
- •Company scaling: New York office, potential Europe office, leadership layers
- •Operating philosophy: adopt processes intentionally, not by copying big-company habits
- 57:47 – 1:00:21
Founder growth reflections: people leverage, staying close to the work, and owning the ship’s direction
Each founder reflects on how their role changed as the company grew. The central theme is leverage through people—moving from doing to enabling—while still staying close to execution and recognizing that founders alone must steer the company’s long-term trajectory.
- •Biggest leverage is people: aligning the right work to the right owners
- •Leadership reality: more time talking/coordinating as distance from IC work increases
- •Still valuable to dive deep and edit/evaluate work directly (‘stay in the weeds’)
- •Key realization: no one else has full context—founders must steer continuously
- •Founder responsibility scales with headcount and becomes existentially real
