Researched
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
What is Decagon?
Decagon builds autonomous AI support agents for large enterprises — chat, email, voice, and SMS, handling customer issues end to end rather than deflecting to a help article. Founded 2023, around $481 million raised, $4.5 billion valuation as of March 2026, with customers including Duolingo, Chime, Hertz, Affirm, Dropbox, and Notion.
Its pricing philosophy is deliberate and worth stating clearly: Decagon prices AI agents as workers, not software seats. Traditional SaaS charges per user because humans operate the tool. Decagon charges for work the AI performs — usage-based, no seat fees.
The practical version of that: there's no pricing page. It returns a 404. Every deal is a sales conversation, an annual contract, and — by third-party contract data — a six-figure commitment.
A note on sources before we go further
Something you should know while reading anything about Decagon's cost, including this page.
Nearly every "Decagon pricing" article ranking on Google is published by a direct competitor — Fin, My AskAI, eesel, Quiq, Featurebase. All of them sell cheaper alternatives. All of them have an obvious interest in making Decagon's contracts look enormous.
That doesn't make their numbers wrong. Contract figures from Vendr and similar sources hold up across otherwise-competing publishers, which is decent evidence they're roughly accurate. But the framing around those numbers — the comparison tables, the "here's a better option" conclusions — is marketing, and you should read it that way.
What follows uses the figures that corroborate across sources and flags where they diverge.
What Decagon actually costs
No published rates. Third-party contract data gives a range rather than a number:
| Source basis | Reported figure |
|---|
| Median annual contract | ~$386,000-433,000 |
| Reported range | ~$95,000 to $923,000+ |
| Entry point (frequently cited) | ~$95,000/year |
| Per-conversation estimate | ~$0.99 (third-party, unconfirmed) |
| Per-resolution (negotiated) | ~$0.50 reported in at least one case |
Sources disagree on the median — one puts it near $386,000, another near $433,000. Both cite Vendr. Treat the spread as the honest answer: low-to-mid six figures annually, varying substantially by volume, channels, and integrations.
Billing is usage-based with no per-seat component, and you choose between two models.
Per-conversation vs per-resolution: the math that decides it
This choice matters more than the negotiated rate, and it hinges on one number.
Per-conversation charges a fixed rate for every interaction the AI touches — resolved or not. Predictable, forecastable against ticket volume, no ambiguity about what counts. Decagon says most customers pick it for exactly that reason.
The catch: if the AI fails and escalates to a human, you still pay the AI fee. You've paid twice for one ticket.
Per-resolution charges more per unit but only when the AI fully handles an issue without escalation. Better aligned with outcomes, worse for forecasting.
The crossover is roughly a 70% resolution rate. Below that, per-conversation means paying for a large share of interactions that produced no automated outcome. Above it, per-conversation's predictability starts winning because you're rarely paying for failures anyway.
So the question to answer before negotiating isn't "which model is better" — it's "what resolution rate will we realistically hit on our ticket mix?" And remember that month-one rates are always below steady state.
The word to negotiate hardest
If you go per-resolution, get "resolution" defined in writing.
Decagon determines it algorithmically. Multiple reviewers flag this as a source of billing disputes. The genuinely ambiguous case: a customer gets a partial answer, doesn't escalate, and simply gives up. Resolved?
That question is cheap to argue about at 500 tickets and expensive at 50,000 — which is exactly when seasonal spikes hit. Decagon has written thoughtfully about resolution-based pricing as a concept; the practical accounting is what belongs in the contract, in plain language, before you sign.
Who is it for?
Decagon suits large enterprises with genuinely high support volume, complex queries that go beyond FAQ deflection, and budget for a six-figure annual commitment plus the existing helpdesk it layers on top of. The customer list — Duolingo, Hertz, Affirm, Dropbox — reflects that: consumer-scale support operations where automation percentage points translate to real headcount.
Agent Operating Procedures make it a fit for teams whose support logic is complicated enough that decision trees break down, and who want operations people rather than engineers defining agent behavior.
It's not sold to: mid-market companies (the entry point starts around $95,000), anyone who wants published pricing without a sales cycle, teams needing something live this week rather than in six weeks, or organizations whose support is mostly simple FAQ deflection — you'd be paying enterprise rates for a job cheaper tools do adequately.
Key Features
- Agent Operating Procedures (AOPs) — natural-language workflow definitions, configurable without code
- Multi-channel agents — chat, email, voice, and SMS from one platform
- Autonomous resolution — handles issues end to end rather than deflecting
- Usage-based billing — per-conversation or per-resolution, no seat fees
- AI-driven analytics — resolution reporting and performance measurement
- White-glove implementation — included in contract, roughly six weeks to deploy
- Helpdesk integration — layers on Zendesk, Salesforce, and similar rather than replacing them
- Enterprise scale — 100-plus enterprise customers on consumer-scale volume
Decagon vs Competitors 2026
| Tool | Pricing published | Model | Typical annual | Setup |
|---|
| Decagon | ❌ 404 page | Per-conversation or per-resolution | ~$95K-923K | ~6 weeks |
| Intercom Fin | ✅ Yes | $0.99 per resolution | Volume-dependent | Fast |
| Sierra AI | ❌ Custom | Custom | $200K+ reported | Weeks |
| Salesforce Agentforce | ⚠️ Partial | ~$2/conversation or credits | Six figures with Data Cloud | 5-11 months |
| Botpress | ✅ Yes | Usage tiers | From ~$89/mo | Self-serve |
Pricing checked August 2026. Decagon and Sierra figures come from third-party contract data, much of it published by competitors — treat conclusions accordingly. Competitor prices are approximate.
Decagon vs Intercom Fin: Intercom publishes $0.99 per resolution with no platform fee, integration fee, or setup charge — genuinely transparent, and outcome-aligned by default. Decagon is more configurable for complex enterprise support but opaque and far larger as a commitment. Note that Salesforce agreed to acquire Fin for roughly $3.6 billion in June 2026, folding it into Agentforce, which is worth weighing on a multi-year decision.
Decagon vs Sierra AI: Sierra operates the same way — enterprise-only, custom contracts, no public pricing, reported year-one costs above $200,000. They compete for the same buyer. If you're evaluating one, quote the other; opacity on both sides means competitive tension is your main leverage.
Decagon vs Salesforce Agentforce: Agentforce has the CRM data advantage if Salesforce is your system of record, but requires Data Cloud and runs 5-11 month implementations. Decagon deploys in around six weeks and is CRM-agnostic. For Salesforce-native orgs, Agentforce; for faster deployment across an existing helpdesk, Decagon.
Decagon vs Botpress: Botpress is self-serve from around $89/month with published pricing and developer depth. It's a different product for a different buyer — you build the agent; Decagon delivers one. For teams with engineering capacity and mid-market budgets, Botpress; for enterprise volume with white-glove delivery, Decagon.
Pricing 2026
Decagon publishes nothing. Here's the practical framework instead:
Budget expectation: low-to-mid six figures annually. Reported entry around $95,000, median somewhere in the $386,000-433,000 range depending on source, top end past $900,000.
Model choice: per-conversation (predictable, pays for failures) or per-resolution (outcome-aligned, definition matters). Decide based on your realistic resolution rate against the ~70% crossover.
What's included: white-glove implementation, roughly six weeks to deploy, no per-seat fees.
What's not: your existing helpdesk platform, which you keep paying for. Custom integrations and workflow development. Premium support tiers and dedicated CSMs, where offered.
Negotiation notes: contracts are annual with minimums. Get resolution defined in writing. Model a ramp period where month-one rates sit below target. And quote Sierra — competitive tension is the main lever available when neither vendor publishes a price.
Figures checked August 2026 from third-party contract data (Vendr and similar). No official pricing exists to verify against.
Use Cases
Consumer-scale support automation: A company handling tens of thousands of monthly tickets automates a large share end to end across chat, email, and voice.
Complex query handling: Support logic too intricate for decision trees gets defined as Agent Operating Procedures in natural language by ops staff rather than engineers.
Multi-channel consistency: One agent definition serves chat, email, SMS, and voice rather than maintaining separate bots per channel.
Layering onto an existing stack: An enterprise keeps Zendesk or Salesforce for human agents and ticketing while Decagon handles autonomous resolution on top.
Seasonal volume absorption: A business with sharp seasonal spikes uses AI capacity that scales without hiring, though this is exactly when per-resolution billing disputes surface.
Our Verdict
Decagon is a serious enterprise product with serious backing — $481 million raised, a $4.5 billion valuation, and a customer list of consumer brands running support at real scale. Agent Operating Procedures are a genuinely good idea, making agent behavior configurable in natural language by the people who understand the support process rather than only by engineers. Six-week deployment with white-glove implementation is fast for software at this price point.
The reservations are structural rather than about quality. Pricing is entirely opaque — a 404 where the pricing page should be — with contracts in the low-to-mid six figures, which puts it out of reach for anyone below genuine enterprise scale. The per-conversation model means paying for interactions the AI fails to resolve. The per-resolution model rests on an algorithmically-determined definition of "resolved" that reviewers repeatedly flag as a source of billing disputes. And it layers on top of your existing helpdesk rather than replacing it, so the six figures is additive.
One more time, because it matters when you research this further: most of what ranks for Decagon pricing is written by cheaper competitors. The numbers corroborate; the conclusions are sales copy.
For large enterprises with the volume and budget, recommend evaluating — with Sierra quoted alongside and "resolution" defined in writing. For anyone smaller, Intercom Fin's published $0.99 per resolution or a self-serve platform will serve you better.
Note: Decagon does not currently have an active affiliate program with AIVario. We earn no commission, and this rating carries no commercial incentive.
Best for: Large enterprises with high support volume, complex support logic beyond FAQ deflection, multi-channel operations needing one agent definition, companies with six-figure automation budgets
Not ideal for: Mid-market companies, anyone wanting published pricing, teams needing deployment in days, simple FAQ deflection use cases
Bottom line: Capable enterprise support automation sold entirely through sales, at low-to-mid six figures annually. Worth evaluating at real scale — negotiate the definition of "resolved" before anything else.
- Intercom — published $0.99 per resolution; being acquired by Salesforce
- Sierra AI — the direct enterprise competitor; quote both
- Salesforce Einstein — Agentforce if Salesforce is your system of record
- Botpress — self-serve agent building with published pricing
- Tidio — small-business support bots at the opposite end of the market
Frequently Asked Questions about Decagon
How much does Decagon cost?
Decagon publishes nothing — its pricing page returns a 404 and the site routes you to a demo request. Third-party contract data from Vendr puts the median annual deal somewhere between roughly $386,000 and $433,000, with reported ranges spanning about $95,000 to over $900,000 depending on which source you read. What's consistent across every source: enterprise-only, annual contracts, usage-based billing, no per-seat fees, and a sales-led process. Plan for a six-figure commitment.
Per-conversation or per-resolution — which is better?
It depends entirely on your AI resolution rate, and the math flips at a predictable point. Per-conversation charges a fixed rate for every interaction the AI touches, resolved or not — predictable and easy to forecast, but you pay for failures too. Per-resolution charges more per unit but only when the AI fully handles an issue. If your resolution rate runs below roughly 70%, per-conversation means paying for a meaningful share of interactions that produced nothing. Decagon reports most customers choose per-conversation for the predictability.
What counts as a 'resolution'?
This is the single most important thing to pin down in negotiation, and it's genuinely ambiguous. If a customer gets a partial answer and gives up, does that count as resolved? Decagon determines resolution algorithmically, and multiple reviewers flag this as a source of billing disputes and forecasting difficulty — particularly during seasonal volume spikes when the numbers get large. Get the definition written into the contract in plain language before signing anything.
What are Agent Operating Procedures?
Decagon's approach to defining agent behavior — natural-language workflow definitions rather than decision trees or code. You describe how a type of query should be handled and the agent follows that procedure, which makes the system configurable by support operations people rather than only engineers. It's the main technical differentiator from platforms built on rigid flowcharts, and it's part of why Decagon targets complex support rather than FAQ deflection.
Do I still need Zendesk or another helpdesk?
Yes. Decagon layers autonomous AI agents on top of your support stack rather than replacing the underlying platform — you maintain your existing helpdesk for human agents, ticketing, and the cases the AI escalates. That's a real budget consideration: Decagon's six-figure contract sits alongside your Zendesk or Salesforce spend, not instead of it. Model total support tooling cost, not just the AI line.
How long does Decagon take to deploy?
Roughly six weeks by most accounts, with white-glove implementation included in the contract price. That's fast for enterprise software and slow if you're expecting to switch something on. Factor in a ramp period beyond deployment too — resolution rates are rarely at target in month one, so early invoices under a per-conversation model can look worse than steady state. Build that into any ROI projection you take to finance.
Who actually uses Decagon?
The customer list skews toward well-known consumer and tech brands: Duolingo, Chime, ClassPass, Hertz, Oura, Affirm, Dropbox, Notion, and Rippling among others, with the company claiming over 100 enterprises. Founded in 2023, Decagon has raised around $481 million and reached a $4.5 billion valuation in March 2026 after a $250 million round. That's a serious, well-capitalized vendor — the question is scale fit, not stability.
What are the cheaper alternatives?
Several, and this is where you should read carefully. Intercom's Fin charges around $0.99 per resolution with no platform fee, and Salesforce agreed to acquire it for roughly $3.6 billion in June 2026 to fold into Agentforce. Smaller vendors advertise per-ticket rates around $0.10-0.40 with published pricing. Be aware that most 'Decagon pricing' articles online are published by these competitors, so treat their framing accordingly — the underlying contract figures are corroborated, the conclusions are marketing.
