Decagon Teardown — AI Customer Support at Scale ($30M+ ARR, Klarna + Bilt Customers, $500M Val)
Copyable to YOU
Sign in with Google to see your personal Copyable Score - a 5-dimension breakdown of how likely you (with your budget, tech stack, channels, network, and timing) can replicate this product.
Decagon Teardown — AI Customer Support at Scale
TL;DR — Decagon is an AI customer support agent platform built for B2C scale. Founded August 2023 by Jesse Zhang (ex-Citadel quant, ex-Lowkey founder acquired by Niantic) and Ashwin Sreenivas (ex-Helia, acquired by Scale AI). Raised $65M Series B May 2024 led by Bain Capital Ventures and Accel, with a16z, A* Capital, and notable angels participating. Reported ARR near $30M with >$500M valuation. Product replaces tier-1 ticket handling for companies like Klarna, Bilt Rewards, Eventbrite, Rippling, and Notion, where 30-70% of inbound chat or email can resolve without a human.
1. The Numbers That Matter
| Metric | Value | Source / Signal |
|---|---|---|
| Founded | August 2023 | YC alum profile, founder LinkedIn |
| ARR (reported) | ~$30M | Press leak, May 2024 Series B coverage |
| Series A | $35M (Jan 2024) | Accel led, six months after founding |
| Series B | $65M (May 2024) | Bain Capital Ventures + Accel, six months after A |
| Valuation | $500M+ post-money on B | The Information + Forbes coverage |
| Total raised | ~$100M | A + B + seed |
| Headcount | ~40-60 (estimated late 2024) | LinkedIn employee count band |
| Customers (named) | Klarna, Bilt, Eventbrite, Notion, Rippling, Substack, Webflow | Customer page + case studies |
| Pricing | Not public; enterprise contracts ~$50K-$500K ACV | Inferred from ARR / customer count |
$30M ARR with ~50 customers implies blended ACV near $600K. Enterprise sales motion, not PLG. Needs real AE team, real SE team, real SOC2 audit before customer 10. That is the gating cost.
For indie operator, $30M ARR is wrong target. Right read: customer support AI for single vertical (returns for D2C ecommerce, fraud disputes for fintech, scheduling for healthcare) where buyer is head of CX with $30K-$80K budget and workflow is narrow. There blended ACV drops to $5K-$15K, sales cycle drops from 6 months to 6 weeks, moat shifts from "we serve Klarna" to "we own returns-handoff workflow for Shopify Plus brands."
2. Product Surface
Decagon sells "AI Agent" — not chatbot, not copilot, but autonomous resolver. Distinction is technical and commercial.
Chatbot uses decision tree. Copilot drafts replies for human agent. AI agent reads ticket, queries internal systems (order DB, refund engine, knowledge base, CRM), takes action (issue refund, update shipping address, escalate to fraud team), writes back to customer in brand voice. Decagon claims 30-70% full resolution without human review across customer cohorts.
Product layers observed from public docs, customer case studies, careers page job descriptions:
- Ingestion — Knowledge base scraping (Zendesk Help Center, Intercom Articles, Notion docs, Confluence), past ticket history, macros and saved replies, brand voice tone documents.
- Routing brain — Hybrid LLM call (GPT-5 and Claude based on task, inferred from job postings). Custom intent classifier on top.
- Action layer — API integrations with Shopify, Stripe, internal customer DBs, fraud tools (Sift, Forter), shipping carriers. This is the moat. Each integration takes 2-6 weeks of engineering to harden.
- Guardrails — Confidence scoring, automatic human handoff below threshold, escalation rules per customer, audit log for every action taken.
- Reporting — Resolution rate, CSAT post-resolution, cost-per-ticket vs human baseline. Dashboards are what buyer signs renewal on.
IP that compounds is not the LLM call. It is action layer and guardrails. Any new entrant who skips guardrails will lose first time AI refunds $10,000 order to wrong customer.
Sign in to read this report
You have read your 1 free report. Sign in with Google to unlock 2 more.
Sign in with Google