Fin vs Decagon vs Sierra
AI customer-service agents: the outcome-pricing shootout. Scores from the RXed AI Periodic Table v1.0 — identical 20-element scale, deterministic overall. How we audit →
At a glance
Scope = elements of the AI stack in the tool’s stated scope (dashed = honestly out of scope, never a penalty). Quality = how well it scores on what it covers. No single number — both axes, always.
The verdict
Fin (Scope 18/20 · Quality 7/10) is the only one whose price you can model before a sales call — $0.99 per outcome — as long as you audit the assumed-vs-confirmed resolution mix and budget on the 38-72% resolution independent tests find, not the 76% headline. Decagon (Scope 18/20 · Quality 7/10) runs 80% of its inference on models fine-tuned in-house and case studies show real actions, not just deflection — but there has never been a public price, and reviewers keep hitting the same wall: nobody can tell you why the agent did what it did (In 2). Sierra (Scope 18/20 · Quality 7/10) doubled ARR to $200M on outcome pricing with the deepest agent engineering of the trio (Ag 9.5, Ev 9.5) — behind a ~€130K+ platform floor and an NDA. Under ~15K tickets/month with a help center → Fin. High-volume operations with a sales cycle to spare → Decagon or Sierra.
Where the quality sits
Family mean over covered elements only — a dash means the family is honestly outside that tool’s scope.
| Fin (Intercom) Scope 18/20 · Quality 7/10 audited 2026-07-31 | Decagon Scope 18/20 · Quality 7/10 audited 2026-07-30 | Sierra Scope 18/20 · Quality 7/10 audited 2026-07-29 | |
|---|---|---|---|
| Vendor | Fin (formerly Intercom; Salesforce acquisition pending) | Decagon AI, Inc. | Sierra |
| Origin | US | US | US |
| Pricing | ~€0.85 ($0.99) per outcome (resolution, procedure handoff, disqualification); ~€8.60 ($9.99) per sales qualification. Standalone on other helpdesks: $49/month base incl. 50 outcomes. Inside Intercom: seats from ~€25 ($29)/seat/month on top. Copilot $35/seat. Fin Voice custom. 14-day free trial, unlimited outcomes | Never publicly published — confirmed via a zero-result Wayback Machine check on decagon.ai/pricing. Third-party deal data (Vendr, via aggregator reporting) puts a fixed platform fee around €44K/yr ($50K), usage at roughly €0.87/conversation ($0.99) or €0.44/resolution ($0.50, contract-negotiated), and a median enterprise contract around €340K/yr ($386K), ranging €84K–€519K+ ($95K–$590K+). All figures are third-party estimates, not vendor-confirmed, with approximate EUR conversion at the 29/07/2026 rate (~$1 = €0.88). | Outcome-based, not published — third-party estimates: roughly €130,000–650,000+/year platform minimum plus €45,000–175,000 implementation, and about €1.30 per resolved interaction (≈$1.50); no self-serve tier, no free trial. |
| Pr Prompts | 8 | 9 | 8 |
| Em Embeddings | — | — | 6 |
| Cx Context | 8 | 8 | 9 |
| Tr Tracing | 8 | 7 | 9 |
| Lg LLM | 8 | 9 | 9 |
| Fc Function calling | 8 | 8 | 9 |
| Vx Vector store | — | — | 6 |
| Rg RAG | 9 | 8 | 9 |
| Gr Guardrails | 8 | 8 | 8 |
| Mm Multimodal | 7 | 6 | 6 |
| Ag Agents | 8 | 9 | 10 |
| Ft Fine-tuning | 3 | 7 | — |
| Fw Frameworks & harnesses | 8 | 9 | 9 |
| Ev Evaluations | 8 | 9 | 10 |
| Sm Small models | 6 | 9 | 6 |
| Ma Multi-agent | 6 | 5 | 7 |
| Sy Synthetic data | 5 | 4 | — |
| Pc Protocols | 7 | 6 | 7 |
| In Interpretability | 5 | 2 | 4 |
| Th Thinking models | 5 | 4 | 6 |
Hover any score for the written reason. Full detail: Fin (Intercom) passport → · Decagon passport → · Sierra passport →