Decagon — Defaults to per-conversation billing, so you pay for interactions the AI fails to resolve — on top of a reported $50,000 annual platform fee.
Short answer: Decagon builds AI customer support agents with two pricing models: per-conversation, the default for most customers, and per-resolution. Per-conversation is reported at roughly $0.99 per interaction including ones the AI fails to resolve. A platform fee of $50,000 a year is reported regardless of model. Marketplace data reports median annual contracts between $386,000 and $400,000, with most deals between $95,000 and $590,000. There is no free trial, no self-serve signup and no pricing page.
On this page
Verdict at a glance
Overall rating: 3.0/5 ★★★☆☆
Standout: Genuine choice of pricing model. Being able to select per-conversation or per-resolution lets you match the contract to your expected resolution rate, which most vendors do not offer.
Watch out: Per-conversation is the default and charges for interactions the AI fails to resolve — so a low resolution rate costs you twice, once in fees and once in escalations.
Decagon builds AI customer support agents and offers something its closest competitor does not: a genuine choice of pricing model. You can be billed per conversation, which is the default for most customers, or per resolution. That choice matters because the right answer depends on your resolution rate — per-resolution only pays off if the AI resolves a high proportion of interactions, and below roughly 50% resolution, per-conversation is likely cheaper. Being able to align the contract to your actual performance is a real advantage over a single imposed model. The default, however, cuts the wrong way for most buyers. Per-conversation at a reported $0.99 charges for every interaction including ones the AI fails to resolve — so a poor resolution rate costs you twice, once in fees and again in the human escalations you still handle. On top of either model sits a reported platform fee of $50,000 a year regardless of which you choose. Marketplace data from Vendr and reported buyer figures put median annual contracts between $386,000 and $400,000, with most deals landing between $95,000 and $590,000 and a hard minimum around $50,000. There is no free trial, no self-serve signup, no public documentation and no pricing page — a number requires a discovery call and a custom proposal.
Quick facts
| Attribute | Detail |
|---|---|
| Ideal team | Enterprises with very high support volume and strong resolution rates |
| Learning curve | Moderate — agent configuration against your support content |
| Setup time | Weeks, following a discovery call and custom proposal |
| Free trial | ⚠️ None — no trial, no self-serve signup, no pricing page |
| Free tier | None |
| Default model | ⚠️ Per conversation, reported ~$0.99, including failures |
| Platform fee | ⚠️ Reported $50,000 a year regardless of model |
| Reported median | $386,000-$400,000 annual contracts |
Overview
Decagon builds AI agents for customer support, competing directly with Sierra for the same enterprise buyers. Its distinguishing commercial feature is offering two pricing models rather than imposing one: per-conversation, which is the default for most customers, and per-resolution. That choice has real value, because the correct model depends entirely on how well the agent performs against your specific support content — per-resolution only pays off if the resolution rate stays high, and below roughly 50% resolution, per-conversation would likely be cheaper.
Per-conversation pricing is reported at roughly $0.99 per interaction, with large enterprise customers at higher volumes able to negotiate lower rates. The important characteristic is that it charges for every interaction, including ones the AI fails to resolve — so a low resolution rate is doubly expensive, costing you both the per-conversation fee and the human handling of the escalation. On top of either model, a platform fee of $50,000 a year is reported regardless of which pricing model is chosen.
Total contract values are substantial. Marketplace data from Vendr and reported buyer numbers put median annual contracts between roughly $386,000 and $400,000, with most deals landing somewhere in the $95,000 to $590,000 per year range and a hard minimum of around $50,000. Nothing about this is available without a sales process: there is no free trial, no self-serve signup, no public documentation and no pricing page on the website. Obtaining a number requires scheduling a discovery call, sharing your ticket volume and waiting for a custom proposal. That opacity is standard at this end of the market but it means, as with Sierra, that no buyer can form an independent view of the product before entering procurement.
Key features
- Two pricing models — ⭐ per conversation or per resolution
- AI customer support agents — conversational resolution at scale
- Per-conversation default — ⚠️ reported ~$0.99, including failures
- Per-resolution option — pays off above roughly 50% resolution
- Volume negotiation — lower per-conversation rates at higher volumes
- Platform fee — ⚠️ reported $50,000 a year on either model
- Enterprise scale — reported contracts to $590,000 a year
- Custom proposals — priced from your ticket volume
- Integration with support platforms — including Intercom
- No public documentation — ⚠️ evaluation is sales-led throughout
Pricing (as of 2026-09-29)
Model: per-conversation or per-resolution, plus a reported annual platform fee · Free tier: ⚠️ none — no trial, no self-serve signup, no pricing page · Entry paid: ⚠️ Quote only — reported $50,000 annual platform fee minimum
| Plan | Price | Includes |
|---|---|---|
| Per conversation (default) | ⚠️ Reported ~$0.99 per interaction | Charges for every interaction including ones the AI fails to resolve · lower rates negotiable at high volume |
| Per resolution | Reported as an alternative model | Only pays off if resolution rate stays high — below roughly 50% resolution, per-conversation is likely cheaper |
| Platform fee | ⚠️ Reported $50,000 a year | Charged regardless of which pricing model you choose |
| Reported contract range | $95,000-$590,000 a year | Median between $386,000 and $400,000 per marketplace data, with a hard minimum around $50,000 |
⚠️ Per-conversation is the default and charges for every interaction including ones the AI fails to resolve. A low resolution rate therefore costs twice — once in the per-conversation fee and again in the human handling of the escalation you still have to do.
⭐ Being offered a genuine choice between per-conversation and per-resolution is a real advantage over a single imposed model. The correct answer depends on your resolution rate: per-resolution pays off when it stays high, while below roughly 50% per-conversation is likely cheaper.
⚠️ A platform fee of $50,000 a year is reported regardless of pricing model, and marketplace data puts median annual contracts between $386,000 and $400,000. There is no free trial, no self-serve signup, no public documentation and no pricing page — a number requires a discovery call and a custom proposal.
Ratings breakdown
| Criterion | Score | Why |
|---|---|---|
| Ease Of Use | 4.0/5 | Configuration against support content is straightforward once scoped |
| Features | 4.0/5 | Strong conversational support agents with model choice |
| Value | 2.0/5 | ⚠️ Default model bills failures, plus a $50,000 platform fee |
| Support | 3.5/5 | Enterprise-grade, though everything is sales-led |
Pros
- Genuine choice between per-conversation and per-resolution billing
- Contract model can be matched to your actual resolution rate
- Lower per-conversation rates negotiable at higher volumes
- Strong conversational resolution for high-volume support
- Integrates with existing support platforms including Intercom
- Custom proposals priced from your real ticket volume
- Enterprise scale, with reported contracts to $590,000 a year
- Per-resolution option available if your resolution rate is high
Cons
- ⚠️ Per-conversation default bills interactions the AI fails to resolve
- ⚠️ Reported $50,000 annual platform fee regardless of model
- ⚠️ No free trial, self-serve signup, public documentation or pricing page
- Median reported contracts of $386,000-$400,000 a year
- Low resolution rates cost twice — in fees and in escalations
- Evaluation requires a discovery call before any number is available
- Hard minimum around $50,000 excludes all but enterprise buyers
Who it’s for
- Enterprises with very high customer support volume
- Organisations confident their resolution rate will exceed 50%
- Teams who want to choose their pricing model rather than accept one
- Companies with procurement capacity for a six-figure contract
Who it’s not for
- Any organisation below enterprise support volume
- Teams whose resolution rates are unproven, given per-conversation defaults
- Buyers who need documentation or a trial before a sales call
- Businesses unable to absorb a $50,000 platform fee before usage
Best use cases
- Resolving very high volumes of customer support conversations
- Choosing per-resolution billing when resolution rates exceed 50%
- Negotiating lower per-conversation rates at high volume
- Integrating AI agents with an existing support platform
- Modelling both pricing models against your actual resolution rate
Integrations
Decagon integrates with existing customer support platforms including Intercom, so agents work inside the systems a support team already operates rather than replacing them. Because there is no public documentation, verifying support for your specific stack requires asking during the discovery call — which is one more reason to scope integration requirements precisely before the first conversation rather than during it.
How it compares
| Decagon | Sierra | Beam AI | CrewAI | |
|---|---|---|---|---|
| Default model | ⚠️ Per conversation, ~$0.99 | ⭐ Per resolution, ~$1.50 | Per agent tier | Per execution |
| Bills failures | ⚠️ Yes on default model | No — resolution only | n/a | Yes — per kickoff |
| Model choice | ⭐ Conversation or resolution | Resolution plus routing rate | No | No |
| Platform fee | ⚠️ $50,000/yr reported | ⚠️ Yes, additional | Included in tier | Included in tier |
| Reported annual | $95,000-$590,000 | $200,000-$350,000+ | $11,880-$47,880 at listed tiers | $60,000-$120,000 enterprise |
| Free access | ⚠️ None | ⚠️ None | ⚠️ None | ⭐ 50 executions monthly |
| Best fit | Per-conversation support | Outcome-priced service agents | Agentic process automation | Multi-agent orchestration |
LangSmith figures are from published third-party reporting verified 2026-09-29. Sierra, Decagon and Beam AI publish no public pricing, and CrewAI publishes only its free Basic tier.
Alternatives
Sierra is the direct competitor and charges only on resolutions at roughly $1.50, so it does not bill for interactions its agent fails to resolve — a meaningful structural difference. Beam AI handles process automation rather than customer conversations. CrewAI and LangChain are building blocks with free tiers rather than managed enterprise services.
Recent changes
Decagon continues to offer two pricing models: per-conversation, which is the default for most customers, and per-resolution. Per-conversation pricing means paying for every interaction, including ones the AI fails to resolve, and public information suggests a cost of roughly $0.99 per conversation, with large enterprise customers at higher volumes able to negotiate lower rates. The per-resolution model only pays off if the AI resolution rate stays high, with per-conversation likely cheaper below 50% resolution. Decagon is reported to charge an annual platform fee of $50,000 per year regardless of the pricing model chosen. Based on marketplace data from Vendr and reported buyer numbers, median annual contracts run between roughly $386,000 and $400,000, with most deals landing in the $95,000 to $590,000 per year range and a hard minimum of around $50,000. There is no free trial, no self-serve signup, no public documentation and no pricing page on the website.
Frequently asked questions
How much does Decagon cost?
It publishes nothing. Per-conversation pricing is reported at roughly $0.99 per interaction, with a $50,000 annual platform fee reported regardless of model. Marketplace data puts median annual contracts between $386,000 and $400,000, with most deals between $95,000 and $590,000 and a hard minimum around $50,000.
Does Decagon charge for conversations it fails to resolve?
On the default per-conversation model, yes — you pay for every interaction including unresolved ones. That means a low resolution rate costs twice: once in the per-conversation fee and again in the human handling of the escalation. Sierra charges only on resolutions, which is a structural difference worth weighing.
Should I choose per-conversation or per-resolution?
It depends on your resolution rate. Per-resolution only pays off if the AI resolves a high proportion of interactions; below roughly 50% resolution, per-conversation is likely cheaper. Model both against a realistic resolution estimate for your specific support content rather than the vendor’s benchmark, since the answer flips around that threshold.
What is Decagon’s platform fee?
A reported $50,000 a year, charged regardless of which pricing model you choose. That sits on top of per-conversation or per-resolution usage charges, so $50,000 is effectively the floor before a single interaction is handled — which is why the reported hard minimum for contracts sits at around that figure.
Decagon or Sierra?
Both are enterprise customer support agent platforms with no published pricing. Sierra charges on resolutions only at roughly $1.50; Decagon defaults to per-conversation at roughly $0.99 including failures, with per-resolution available. Sierra’s model is better aligned; Decagon’s offers more choice. Quote both.
Can I try Decagon before buying?
No — there is no free trial, no self-serve signup, no public documentation and no pricing page. Obtaining a number requires scheduling a discovery call, sharing your ticket volume and waiting for a custom proposal. That means no independent evaluation is possible before entering procurement.
When does Decagon make commercial sense?
When support volume is very high, your resolution rate is demonstrably strong, and a six-figure annual contract is proportionate to what you currently spend on human handling. Below that, the $50,000 platform floor alone rules it out — and if resolution rates are unproven, the per-conversation default makes the downside worse rather than better.
Sources
- eesel AI — Decagon AI pricing in 2026: what it actually costs
- CloudTalk — Decagon AI pricing and plans: complete cost guide for 2026
- Coworker AI — Sierra AI pricing 2026: what it really costs
- AIChief — Beam AI review: cost, use cases and alternatives 2026
- Lindy — CrewAI pricing, features and alternatives explained 2026
Compiled from public sources; last verified 2026-09-29. Pricing is subject to change — verify with the vendor. Ratings reflect editorial opinion based on documented capabilities, published pricing, and aggregated user reviews. DigiJaws does not accept payment for review placement or rating.
Decagon Review (3/5): Pricing & Verdict (2026). DigiJaws, 2026. https://digijaws.com/reviews/decagon-review/