# Decagon (Decagon) Category: AI agents. Public reviews about the AI: 10. Profile updated 2026-10-08. An enterprise AI agent for customer service across chat, voice and email, configured with natural-language procedures. ## What it does Decagon sells an AI agent for customer service over chat, voice and email [1]. Teams define the agent's workflows in natural language as Agent Operating Procedures (AOPs), and Duet can draft AOPs from uploaded conversation transcripts [1][2]. Engineering teams can version workflows with Git-based tracking [2]. Channels: Chat, Voice, Email. Works with: Decagon names Salesforce, Zendesk and Intercom for tickets and routing, Kustomer for knowledge sync, and Amazon Connect and RingCentral for voice [3]. ## Handoff Decagon says email runs through Zendesk and Intercom "with smooth handoffs when needed" and voice calls transfer to human agents when escalation is required [3]. What the human agent receives is not published [3]. ## Controls - Guardrails for brand voice, escalations and hallucinations across chat, voice and email [2] - Simulations test agent behavior before production and with each update; Duet can generate tests, and checkpoints confirm the agent used the right actions, data and tools [4] - Users can see why a test passed or failed, schedule recurring runs, and send alerts to tools such as PagerDuty [4] - CI/CD integration to test every agent version, and Git-based version tracking for workflows [4][2] - Watchtower for always-on QA and Experiments for live A/B tests [1] - Tracing of the agent's reasoning at any point in a conversation [2] ## Setup Decagon says most enterprises go live "in weeks rather than months" and that a forward-deployed team is included with every engagement [5]. ## Billing By contract. Not published; sold by contract. Decagon offers per-conversation pricing (a fixed rate per incoming conversation) or per-resolution pricing (a higher fixed rate per fully resolved conversation, no charge for escalations), and says most customers choose per-conversation [6][5]. What counts: Decagon frames resolution as a conversation resolved by the AI agent without escalating to a human; it publishes no formal billing definition [6]. ## Vendor claims (not results) - Decagon cites customer results including 70% chat and voice resolution at Chime and an 80% deflection rate at Duolingo [1] - Decagon cites a 95% cost reduction at ClassPass and a 65% cost reduction at Curology [1] - Decagon says Duet can "compress work that used to take weeks into minutes" [2] ## Not found in public sources - Public prices - What a human agent receives at handoff - An approval workflow for changes to AOPs - A formal definition of a billed resolution or conversation ## What reviewers say - Does the job: 100% positive (10 of 10; 95% interval 72%–100%) - Accuracy: 0 positive, 5 negative (too few to rate) - Handoff to a person: 1 positive, 2 negative (too few to rate) - Cost: 0 positive, 0 negative (too few to rate) - Setup and upkeep: 1 positive, 4 negative (too few to rate) > "Sometimes the AI needs additional context or a human agent to step in, so better handling of edge cases and smoother handoffs would make the support workflow more efficient." (Akash R., G2 (via AWS Marketplace), 2026-09-03: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "One thing I think Decagon could improve is the accuracy and consistency of responses in some situations." (Akash K., G2 (via AWS Marketplace), 2026-08-26: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "The initial setup was fairly straightforward, and the platform was intuitive enough to get started without a lot of friction." (Sumedha R., G2 (via AWS Marketplace), 2026-08-21: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "If I notice any problems or mistakes in company policies, I find that the resolutions provided by Decagon might not be proper and correct." (CA Rahul B., G2 (via AWS Marketplace), 2026-08-20: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "It may make mistakes or misunderstand unusual customer problems , so a human support agents may be still be needed" (Jatin K., G2 (via AWS Marketplace), 2026-08-13: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "It handles a large volume of routine requests quickly while giving our team the flexibility to step in when human attention is needed." (Krishan Y., G2 (via AWS Marketplace), 2026-08-12: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "Sometimes the responses need to be reviewed or corrected by a human, especially when the conversation requires deeper context." (Subhashree S., G2 (via AWS Marketplace), 2026-08-10: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "Sometimes the client’s issues aren’t understood, and the response comes back about something else." (abhay c., G2 (via AWS Marketplace), 2026-07-31: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "At the beginning, you need to review the responses quite closely to make sure they match your tone and business rules." (Logistics and Supply Chain, G2 (via AWS Marketplace), 2026-07-28: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) > "The UI during monitoring and conversations can feel a bit overwhelming, and navigation sometimes forces me to drill down into specific interactions because I have to do so many clicks." (Adam R., G2 (via AWS Marketplace), 2026-07-21: https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy) ## Sources 1. [Decagon home page](https://decagon.ai/), Decagon, accessed 2026-10-08 2. [Agent Operating Procedures](https://decagon.ai/product/aop), Decagon, accessed 2026-10-08 3. [Integrations](https://decagon.ai/product/integrations), Decagon, accessed 2026-10-08 4. [Testing & QA](https://decagon.ai/product/testing-qa), Decagon, accessed 2026-10-08 5. [Decagon vs. Sierra](https://decagon.ai/vs/sierra), Decagon, accessed 2026-10-08 6. [Pricing the AI Agent Economy](https://decagon.ai/blog/pricing-ai-agents), Decagon, accessed 2026-10-08 Canonical page: https://get-support-ai.com/products/decagon