Decagon
B2B enterprise SaaS; usage-based pricing on a per-conversation and per-resolution basis, land-and-expand into large enterprise support orgs
Series A (~$35M, Jul 2024) valuation not disclosed, so the trail begins at Series B. Valuation roughly tripled from $1.5B to $4.5B in under six months (Jun 2025 to Jan 2026); the Mar 2026 secondary tender re-confirmed the $4.5B primary mark.
Earnings, margins, COGS & capex
Private, venture-funded, pre-profitability. ARR estimated ~$35M annualized as of Oct 2025 (Sacra), up from ~$10M exiting 2024, with Sacra reporting Q3 2025 revenue and ARR each grew more than 3x YoY, on the back of 100+ new enterprise logos added in the past fiscal year. Valuation ($4.5B) implies a very high ARR multiple (~130x on the $35M estimate), typical of frontier agentic-AI names but leaving little margin for growth disappointment. No audited financials, margins, or retention figures are public; Decagon has not officially disclosed revenue since late 2024.
Revenue trend
Margins
pressured by per-conversation LLM inference cost; improves as model costs fall and per-resolution pricing captures value
assumed deeply negative (growth-stage burn)
COGS structure
Primary COGS is third-party LLM inference (OpenAI, Anthropic, Google) plus cloud compute and voice/telephony infrastructure; a variable cost that scales with conversation volume, unlike zero-marginal-cost classic SaaS. Not separately disclosed.
Capex
Minimal — asset-light; no owned data centers disclosed. Compute is rented, expensed as COGS/opex.
Latest earnings
not applicable
not disclosed
- ARR (annualized, Oct 2025 est.)
- ~$35M (Sacra)
- YoY growth (Q3 2025 est.)
- more than 3x (Sacra)
- Enterprise customers
- 100+ large accounts added in past fiscal year
- Latest valuation
- $4.5B (Jan 2026, Series D)
- Total raised
- ~$481M (Series A-D)
- Employees
- 300+ (Mar 2026 tender)
- Best-case resolution rate (customer-cited)
- ~90% without human
Growth drivers
- Enterprise land-and-expand — 100+ new global enterprise customers (Avis Budget Group, Deutsche Telekom, Block, Mercado Libre, Notion, Duolingo, Rippling, Bilt) added in the past fiscal year
- Channel expansion from chat into email and voice (Decagon Voice 2.0 cited a 65% latency reduction, per Decagon), enlarging the addressable support surface per customer
- Per-resolution pricing that ties revenue to demonstrated deflection (customers cite up to ~90% resolution without human intervention, one reducing support headcount ~80%)
- Agent Operating Procedures (AOPs) lowering deployment/iteration cost, shortening time-to-value and enabling faster expansion
- Secular shift of enterprise support budgets from human BPO/headcount toward AI agents
Bull & bear
Decagon is an early leader in the largest, most obviously automatable enterprise-AI category — customer support — with real enterprise ARR growing more than 3x, marquee logos, and a multi-channel product; if agentic CX becomes the default, it compounds into a category-defining platform.
- ~$35M ARR growing >3x with 100+ blue-chip enterprise customers proves the product works at scale, not just in demos
- Customer support is a $100B+ global labor spend pool (Sierra's Taylor cites ~$400B); even modest share capture supports a large outcome
- Per-resolution pricing means revenue scales with proven ROI — customers citing ~90% deflection and ~80% headcount reduction create a durable expansion motion
- Voice + email expansion multiplies revenue per account beyond the initial chat wedge
- Top-tier syndicate and ~$481M raised give it capital to outspend smaller startups and out-service incumbents on white-glove enterprise deployment
A $4.5B valuation on ~$35M ARR (~130x) prices in near-flawless execution in a category being invaded from both sides — CRM incumbents bundling from above and foundation-model vendors from below — with an orchestration moat that may not hold and margins structurally capped by inference COGS.
- ~130x ARR multiple leaves no room for a growth slowdown or a valuation-multiple reset in AI names
- Salesforce Agentforce, Microsoft, Zendesk and Intercom Fin can bundle 'good-enough' agents into existing seats at near-zero marginal price, compressing standalone-vendor economics
- The core IP is orchestration on top of third-party LLMs — replicable, and foundation-model providers are moving up-stack into vertical agents
- No disclosed gross margin; if inference costs stay high while per-resolution prices fall, unit economics may never reach software-grade
- Rival Sierra has pulled ahead — $15.8B valuation and $150M+ ARR (May 2026) — turning the same enterprise deals into bake-offs against a better-capitalized incumbent-in-the-making
- Support automation risk: brand-damaging AI failures or a single large-logo churn event could stall the expansion narrative
What it is worth
Private-market last-round + ARR-multiple sanity check (no public price)
Overvalued if AI multiples reset, incumbents (Salesforce/Microsoft) bundle away pricing power, or a better-funded Sierra wins the category — a down round toward the $1.5B Series C level is the realistic downside.
$4.5B holds as a fair private mark
if growth stays strong but decelerates toward 2x and margins prove workable; multiple compresses toward the ARR as revenue scales into it.
Justified/underpriced if ARR keeps compounding ~3x+, voice/email expand ARR-per-account, and Decagon becomes the category platform — path to a $10B+ next round or a premium IPO/strategic outcome.
Last priced round: Series D, Jan 2026, $250M primary at $4.5B post-money (led by Coatue Management + Index Ventures; prior: Series C $131M at $1.5B Jun 2025 co-led by Accel + a16z; Series B ~$65M at ~$650M Oct 2024; Series A ~$35M Jul 2024). A March 2026 employee tender offer cleared at the same $4.5B mark, corroborating the private valuation. On the ~$35M ARR estimate (Sacra, Oct 2025), $4.5B implies ~130x ARR — an aggressive frontier-AI multiple. No public equity; not directly tradable.
SWOT
Strengths
- Marquee enterprise logo book (Avis Budget Group, Deutsche Telekom, Block, Mercado Libre, Notion, Duolingo, Rippling, Bilt) signaling real large-account traction, not just SMB
- Rapid >3x ARR growth and deep-pocketed investor syndicate (a16z, Accel, Coatue, Index, Bain Capital Ventures, Ribbit, Forerunner, BOND) — ample capital runway
- AOP framework and multi-channel (chat/email/voice) product give a differentiated, enterprise-configurable deployment model
- Usage/resolution-based pricing aligns Decagon revenue with demonstrated customer ROI
Weaknesses
- Small absolute revenue (~$35M ARR est.) against a $4.5B valuation — a ~130x multiple with heavy execution expectations priced in
- No disclosed gross margin; LLM inference COGS structurally caps software-like margins vs. classic SaaS
- Undifferentiated at the model layer — relies on third-party foundation models, so the moat is the orchestration/data/workflow layer, which is contestable
- Single-category concentration (customer support) exposes it to platform bundling by CRM incumbents
Opportunities
- Expand from support deflection into broader agentic CX / revenue-generating concierge and proactive outreach use cases
- Voice is early — telephony/IVR replacement is a very large adjacent market
- International + vertical expansion (telco, fintech, travel already anchored) with reusable AOP templates
- Attach to enterprise data systems as the CX 'system of action,' deepening switching costs
Threats
- Salesforce (Agentforce) and other CRM/helpdesk incumbents bundling agentic support at near-zero incremental price into existing seats
- Sierra (Bret Taylor / Clay Bavor) is now a substantially larger, better-capitalized direct rival — $15.8B valuation and $150M+ ARR (May 2026) vs. Decagon's $4.5B / ~$35M — competing for the same enterprise white-glove deals
- Foundation-model providers (OpenAI, Anthropic) moving up-stack into vertical agents
- Margin compression if inference costs don't fall as fast as price competition, plus AI-hype valuation reset risk
Moats, dependencies & bottlenecks
Moats
AOPs wired into a customer's support stack, data, and processes are painful to rip out once live — but integration is replicable by rivals.
Volume of real resolved conversations can improve routing/resolution quality, a compounding advantage if retained and leveraged.
Marquee logo book (Deutsche Telekom, Block, Mercado Libre) is a reference moat for winning risk-averse enterprise buyers.
Relies on third-party foundation models; little defensibility at the model layer itself.
Dependencies
Critical supplier / core input Product quality, latency, and COGS are all gated by third-party LLM APIs; pricing or access changes hit margins directly, and providers are potential up-stack competitors.
Infrastructure supplier Compute and hosting are rented; scales as variable cost.
Twilio, ElevenLabs-class TTS) Voice channel depends on third-party telephony + speech infra; latency/cost pass through to the product.
Integration surface & competitor Decagon must integrate with the systems whose owners are also building competing native agents — a channel and a threat simultaneously.
Pre-profitability; dependent on continued fundraising or a path to self-funding before capital markets tighten.
Advantages
- First-mover enterprise traction with blue-chip logos in a category with clear, budget-backed ROI
- Multi-channel (chat + email + voice) breadth vs. single-channel point tools
- Outcome-aligned per-resolution pricing that lowers buyer risk
- Well-capitalized (~$481M raised) relative to most direct startups, though now out-raised by Sierra
Weaknesses
- Tiny revenue base vs. valuation — extreme multiple compression risk
- No proprietary model layer; dependent on and potentially disintermediated by foundation-model vendors
- Undisclosed / structurally-capped margins
- Category exposed to incumbent bundling from Salesforce/Microsoft/Zendesk, and to a now-larger Sierra
Bottlenecks
- Gross margin ceiling set by LLM inference cost per conversation
- Enterprise sales cycles and white-glove deployment capacity limit how fast logos convert to revenue
- Reliability/accuracy bar — enterprise buyers won't tolerate brand-damaging hallucinations in customer-facing support
- Scarce, expensive AI/ML talent in a competitive hiring market (the tender offer was framed partly as retention)
Top signals & trends
Top signals
Strong investor conviction and momentum; also raises the bar and reset risk.
Signals investor demand and aids talent retention, but early secondary liquidity can also flag a long path to IPO.
Broad, real enterprise adoption across telco, fintech, travel.
~130x multiple prices in near-perfect execution.
The nearest direct competitor is now materially larger and better-funded, intensifying the enterprise bake-off.
Trends
Directly expands Decagon's TAM; the core secular tailwind.
Improves gross margin over time and widens the ROI gap vs. human support.
Compresses pricing power for standalone vendors.
Threatens the orchestration layer's defensibility.
Enabled the raise; also the main downside if sentiment turns.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Private; foundation-model API — a core inference supplier and potential up-stack competitor.
Private; foundation-model (Claude) supplier for agent reasoning.
Foundation-model and cloud-infrastructure supplier.
Cloud infrastructure / hosting.
Cloud + model hosting; also a competitor via Dynamics.
Telephony / voice infrastructure for the voice channel.
Telco enterprise customer (ADR).
Fintech customer (Cash App / Square parent; ticker changed to XYZ in 2025).
Latin American e-commerce/fintech enterprise customer (cited in Sacra 2026).
Travel/car-rental enterprise customer.
BNPL fintech customer (per Decagon's published customer materials).
Neobank customer (Chime IPO'd 2025; per Decagon's customer materials).
Consumer app customer.
Mostly private high-growth tech and consumer brands cited as customers; Notion, Rippling, and Bilt are named in Sacra's 2026 profile.
Private; AI agent platform led by Bret Taylor & Clay Bavor targeting the same enterprise white-glove CX buyer. Now materially larger than Decagon — $15.8B valuation and $150M+ ARR after a $950M round (May 2026, Tiger Global/GV).
Agentforce embeds AI service agents in Service Cloud with native CRM data; can bundle into existing enterprise seats at low incremental price.
Private; Fin offers usage-based (~$0.99/resolution) AI support, strong in SMB-to-mid-market and on the Intercom stack.
Private (taken private by Hellman & Friedman, 2022); incumbent helpdesk embedding AI agents into its large installed base.
Enterprise workflow platform pushing AI agents into customer/IT service management — competes for enterprise agentic budgets.
Bundles AI customer-service agents into Dynamics/Copilot across its enterprise footprint.
Public helpdesk/CX vendor adding AI agents, strongest in mid-market.
Private; established AI customer-service automation vendor competing on deflection.
Private; AI for contact centers (agent assist + autonomous), overlapping in voice/contact-center CX.
Private; generative-AI support automation competitor.