Vertical B2B · USA

Decagon

A customer-support agent that grew through a chain of references rather than virality; the logo and funding evidence is strong, the revenue and resolution-quality evidence is only secondary.

Verdict: Gaps remain

Updated: 2026-10-06

Profile

What it does: AI agent for customer support; resolves issues across voice, chat and email, with workflows defined as "Agent Operating Procedures". In 2026 it expanded towards sales, onboarding and collections and added a "Personal Agent Gateway" that recognises a customer's own agent [3]
Founded: August 2023, San Francisco (secondary data) [8]
Founders: Jesse Zhang (CEO) and Ashwin Sreenivas (President) [8]
Pricing model: Not public; per-conversation and per-resolution models are debated in the sector. Decagon has not disclosed its own.
Funding: 27 January 2026: Series D $250M at a $4.5B valuation (Coatue, Index); earlier investors a16z, Accel, Bain Capital Ventures, Elad Gil, Forerunner, Ribbit [1]
June 2025: Series C (secondary data; the round release could not be opened). The Series D tripled the valuation to $4.5B [8]; the prior valuation is derived from that, roughly $1.5B (derivation) [1]
4 March 2026: employee tender at a $4.5B valuation [2]

Growth timeline

2025-02-26
Founder interview: logic flows and gradual rollout
"we can roll a change out to a small percentage of our users, it's not an all or nothing thing"
[6]
2025-06
Series C (secondary data; the round release could not be opened)
[8]
2026-01-27
Series D $250M at $4.5B; more than 100 new enterprise customers in the fiscal year
"more than 100 new global enterprise customers"
[1]
2026-08-12
ARR $100M (citing Newcomer, secondary data)
[5]
2026-09
Brazil and Benelux launches; Databricks and Plaid partnerships
[4]
2026-10-01
Dialogues: "AI concierge for every customer [and their agent]"; OpenAI B2B marketplace
[3]

Growth loop

1Enterprise customer defines AOPs
2Agent resolves conversations
3Resolution-rate and cost evidence
4Expansion and reference logo
5Conversation data, self-improving agent
Leak: resolution rate not public; the evidence stays between customer and vendor

The loop runs on enterprise references, not viral output; the red arrow marks the quality evidence nobody outside can see.

The drawing shows the shape of the loop; each node is explained in the list below. The red arrow marks value leaking out of the loop.

Mechanism

Acquisition channels: Direct enterprise sales and a chain of logos: Avis Budget Group, Block, Deutsche Telekom, Oura Health, Affirm, Chime [1]; each big brand is a reference for the next (opinion). In 2026, AWS Marketplace, the OpenAI B2B marketplace and data partnerships with Databricks and Plaid form the distribution layer [4].
Loop: The enterprise customer defines its workflow, the agent resolves conversations, resolution-rate and cost-reduction evidence brings channel and usage expansion in the same account plus new logos, more conversation data feeds the self-improving agent [4]. The loop runs on references, not virality (opinion).
Activation: Gradual rollout: "we can roll a change out to a small percentage of our users, it's not an all or nothing thing" [6]. This is how enterprise trust is built (opinion).
Retention evidence: No figure. Expansion signal: the product's move into sales, onboarding and collections means more work from the same customer [3] (opinion). No retention evidence.
Role of pricing: Undisclosed. If outcome-based pricing is used, revenue is tied directly to a measurable result; who defines a "resolution" is an open question (opinion).
Distribution moves: Geographic launches (Brazil, Benelux), marketplaces [4] and the "Dialogues" customer event [3]. Zhang's positioning: reduce dependence on forward-deployed engineers, differentiate on speed and product design [5].
Team: Headcount does not appear in the sources that could be opened.

Real reading

The five questions of the Growth Claim Check, applied to this company's public claims.

01
What was measured, and by whom?
unclear
ARR figures come from investor and press reporting; the company's own blog gives no revenue figure, only the count of new customers.
02
Perception or outcome?
unclear
Customer count is an outcome; quality metrics such as resolution rate are not public.
03
Is there a stability or quality metric?
no
Resolution rate, conversation volume and NRR are undisclosed.
04
Is there a counterfactual or cohort?
no
No control group or cohort.
05
Nominal or real?
unclear
Whether enterprise pilots count towards ARR is unknown; reading more than 100 new customers against $100M ARR gives roughly $1M per customer, which fits a large-contract profile.

Verdict

Gaps remain

Logo and funding evidence is strong; revenue and quality evidence is secondary. A company that gives no figure in its own sources should be cross-read through customer count and contract profile.

Test your own claim with the same five questions: Growth Claim Check →

Shadow side

Competitive density: Sierra in the same category, with Intercom Fin, Zendesk and Salesforce Agentforce creating price pressure (opinion).
The personal-agent channel: when the customer's agent talks to the company's agent, for whom is customer satisfaction being measured [3] (opinion).
The absence of a revenue figure on the official blog makes it harder to measure the gap between visible and claimed growth [1].

What it teaches a founder in Türkiye

This section is opinion; a reading, not a source.

01
(opinion) In an enterprise vertical the growth loop is a chain of references. The first three logos bring the next thirty; choosing the first customer is a growth decision.
02
(opinion) "Open to a small percentage, measure, expand" is the support-agent equivalent of an A/B test. It works even at low support volume.
03
(opinion) Outcome-based pricing depends on the vendor's definition of "resolution". Let the customer define resolution in the contract.
04
(opinion) Marketplaces are a cheap channel that opens the enterprise door for small teams.
05
(opinion) Cross-read the claims of a company that gives no ARR in its own sources through customer count and contract profile.

A question for you: remove this company's logo and drop the loop onto your own product. Which node stays empty?

Sources

Numbers are the references in the text. An 'opened' source was read by the researcher; 'secondary' ones are verified against their primary source before publication.

[1]
Decagon's $250 million commitment to the AI concierge future
Decagon blog · 2026-01-27 · opened
“more than 100 new global enterprise customers”
[2]
Decagon closes employee tender at $4.5 billion valuation
Decagon blog · 2026-03-04 · opened
“Typically, startup employees can spend years, and sometimes even a decade or more, pouring their talent, energy, and creativity into something they believe in, without knowing when they'll be able to turn that effort into financial flexibility.”
[3]
The AI concierge for every customer [and their agent]
Decagon blog · 2026-10-01 · opened
“Customers are starting to send personal agents like Instinct, Muse, and dots to book, buy, cancel, and negotiate on their behalf”
[4]
Decagon blog index (OpenAI, Databricks, Plaid, Benelux, Brazil, AWS, Autopilot)
Decagon · 2026-06 – 2026-10 · opened
“Introducing Duet Autopilot: The self-improving agent for conversational AI (2026-06-09) · Decagon is now available on AWS Marketplace (2026-07-20) · Decagon is partnering with OpenAI on its new B2B marketplace (2026-10-02) · Decagon partners with Databricks (2026-09-30) · Plaid and Decagon partner (2026-09-28) · Decagon expands into Brazil (2026-09-15) · Growing in Benelux (2026-09-21)”
[5]
Decagon ARR hit $100M with focus on speed and product quality
ARR Club (citing Newcomer) · 2026-08-12 · opened
“ARR hit $100M”
[6]
How Decagon builds customer support AI agents that really work
Amplify Partners (Barrchives) · 2025-02-26 · opened
“we can roll a change out to a small percentage of our users, it's not an all or nothing thing”
[7]
Jesse Zhang — Building Decagon (podcast)
Invest Like the Best · 2025-10-06 · secondary
Note: Transcript could not be opened.
[8]
Decagon (company)
Wikipedia (secondary compilation) · accessed 2026-10-06 · opened
“founded in August 2023 by Jesse Zhang and Ashwin Sreenivas · San Francisco, California · tripled the company's valuation to $4.5 billion”

Open questions

Official ARR statement; Newcomer's source.
Pricing model (per conversation or per resolution) and the definition of resolution.
Resolution rate, total customer count, NRR.
Series C amount and valuation from a primary source; headcount.
Sales motion and first-customer story from the podcast transcript.