--- slug: decagon name: "Decagon" vendor: "Decagon" page_url: https://ai-agent.review/agents/decagon record_url: https://ai-agent.review/agents/decagon.md vendor_website: https://decagon.ai/ tested_performance: not-yet-tested facts_checked_at: 2026-09-20 page_updated_at: 2026-10-03 --- # Decagon > An AI agent platform for customer-service workflows across chat, email and voice. Vendor: Decagon. Website: https://decagon.ai/ Decagon connects customer conversations to a company’s support tools and data. Teams define workflows through Agent Operating Procedures and use testing and monitoring tools to review behavior. Its Curology story gives concrete examples such as shipment-status checks and order replacements; these vendor-published results have not been reproduced by this portal. ## Tested performance Source-based review. This portal has not run a hands-on benchmark, so nothing here reports observed performance. No record contains a score, a rating or a benchmark result, and none will until the agents have actually been tested. ## Fit verdicts Each state describes what the reviewed sources establish about a workflow, not whether the product is good at it. Vendor-documented means the vendor documents the workflow. External report means a source outside the vendor discusses it. Caveat documented means a cited source describes a limitation or failure in a specific context; read it alongside any capability evidence. Not established means the reviewed sources do not settle it, which is not the same as the product being unable to do it. States are derived from the cited claims below, never stored. ### Billing — Vendor-documented - **Vendor capability:** Decagon’s technology page lists billing management, plan upgrades and account updates. It does not specify which billing operations are prebuilt for a particular payment or subscription system. _(Vendor-reported, checked 20 Sept 2026; source: [Technology offering (Decagon)](https://decagon.ai/industry/technology); claim id: `decagon-billing-tech-page`)_ > Illustrative evaluation: ask about a duplicate subscription charge. Verify access to the billing record, policy for credits and approvals, and whether the final action is recorded in the billing system. ### Orders — Vendor-documented - **Vendor capability:** Decagon’s Curology case study lists shipment-status checks, order replacements and shipment cancellation or postponement. These are specific examples from Curology’s configured service workflows. _(Vendor-reported, checked 20 Sept 2026; source: [Curology customer story (Decagon)](https://decagon.ai/case-studies/curology); claim id: `decagon-orders-curology`)_ > Illustrative evaluation: a customer requests an address change after dispatch. Check the carrier and order-system state, then distinguish an allowed update from a request that needs human help. ### Returns — Vendor-documented - **Vendor capability:** Decagon’s retail article describes answering return-eligibility questions and taking actions through systems such as Shopify while following company policies. It does not define a universal policy for refunds or exchanges. _(Vendor-reported, published 5 Aug 2025, checked 20 Sept 2026; source: [From WISMO to upsells: How retail brands are using AI agents (Decagon)](https://decagon.ai/blog/retail-cx-use-cases); claim id: `decagon-fit-returns-vendor`)_ > Illustrative evaluation: return one item from a discounted bundle. Check eligibility, refund calculation, approval rules and the recorded return state. A refund request alone is not proof that a physical return was completed. ### Product questions — Vendor-documented - **Vendor capability:** Decagon’s technology offering describes surfacing relevant documentation during onboarding and troubleshooting. This supports a documented product-information workflow, not a measured accuracy result. _(Vendor-reported, checked 20 Sept 2026; source: [Technology offering (Decagon)](https://decagon.ai/industry/technology); claim id: `decagon-fit-pq-vendor`)_ > Illustrative evaluation: ask whether a plan includes a feature and whether that feature works in a particular integration. Check that the answer uses current product documentation and distinguishes missing information from a confirmed limitation. ### Account issues — Vendor-documented - **Vendor capability:** Decagon’s financial-services page lists password resets, balance inquiries and dispute workflows. Those examples require a configured identity and account-access flow; the page does not prove that a given deployment implements it correctly. _(Vendor-reported, checked 20 Sept 2026; source: [Financial-services offering (Decagon)](https://decagon.ai/industry/financial-services); claim id: `decagon-account-issues-finserv`)_ > Illustrative evaluation: a user who cannot sign in requests an email-address change. Verify identity checks, permission boundaries and escalation when verification fails. ### Complaints — Vendor-documented - **Vendor capability:** Watchtower documents identifying frustration and other conversation flags for team review. This is monitoring capability; the source does not demonstrate an agent completing a complaint remedy. _(Vendor-reported, checked 20 Sept 2026; source: [Watchtower (Decagon)](https://decagon.ai/product/watchtower); claim id: `decagon-complaints-watchtower`)_ > Illustrative evaluation: a customer repeats an unresolved refund complaint. Check recognition of prior attempts, escalation rules and the promised follow-up. A sentiment flag alone does not resolve the issue. ### Human handoff — External report - **Vendor capability:** Decagon documents live-chat escalation, email handoff and voice transfer to human agents. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-fit-handoff-vendor`)_ - **Independent report:** Krishan Y.’s August 12, 2026 review says their team can step in when human attention is needed. This corroborates takeover in one self-reported deployment, without measuring routing accuracy or waiting time. _(Independent report, published 12 Aug 2026, checked 20 Sept 2026; source: [Krishan Y., G2 review syndicated by AWS Marketplace](https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy); claim id: `decagon-human-handoff-g2-krishan`)_ > Illustrative evaluation: request a human after an unsuccessful troubleshooting step. Check queue routing, transcript transfer, repeated questions and the offline fallback. ### Small businesses — Not established No public source was found that settles this. > The reviewed sources do not establish suitability for every small team. Illustrative buying scenario: a two-person support team needs one channel and basic FAQ answers. Ask what setup, integration access and ongoing maintenance that scope requires. ### Mid-market — Not established No public source was found that settles this. > Enterprise marketing does not exclude mid-market use. Illustrative buying scenario: a growing subscription business needs account updates and refunds. Confirm the specific connectors, implementation effort and responsibility for maintaining policies. ### Enterprise — Vendor-documented - **Vendor capability:** Decagon publishes customer stories for companies including Chime, Notion and Rippling. These are vendor-published examples of enterprise use, not independent confirmation of every security, scale or workflow requirement. _(Vendor-reported, checked 20 Sept 2026; source: [Customer stories (Decagon)](https://decagon.ai/case-studies); claim id: `decagon-fit-ent-vendor`)_ > Illustrative buying scenario: several brands share a support operation but use different policies. Evaluate permissions, policy separation, reporting and escalation under that actual structure. ## Industries An industry state says where the cited sources place the product, not how well it performed there. The same four states and the same rubric apply. Industry entries cite claims published in the agent's own record, by id, so a source is never stored twice and the two copies can never disagree. An industry with no entry is unknown, which is a gap in the public record rather than a finding. ### Retail and e-commerce — Vendor-documented - Decagon’s retail article describes answering return-eligibility questions and taking actions through systems such as Shopify while following company policies. It does not define a universal policy for refunds or exchanges. _(Vendor-reported, published 5 Aug 2025, checked 20 Sept 2026; source: [From WISMO to upsells: How retail brands are using AI agents (Decagon)](https://decagon.ai/blog/retail-cx-use-cases); claim id: `decagon-fit-returns-vendor`)_ > Decagon publishes a retail offering describing order and return actions through systems such as Shopify. The source is the vendor's own; it names no retailer and reports no outcome. ### Financial services — Not established No reviewed source places this product in this industry. ### Telecommunications — Not established No reviewed source places this product in this industry. ### Travel and hospitality — Not established No reviewed source places this product in this industry. ### Healthcare — Not established No reviewed source places this product in this industry. ### Public sector and education — Not established No reviewed source places this product in this industry. ### Logistics and delivery — Not established No reviewed source places this product in this industry. ## Strengths - Decagon’s voice page lists 70+ languages with detection and switching. A multilingual team can evaluate its required language combinations, but this list alone does not establish equal accuracy across languages. _(Vendor-reported, checked 20 Sept 2026; source: [Voice AI for customer service (Decagon)](https://decagon.ai/product/voice); claim id: `decagon-languages-strength`)_ - Decagon lists help-desk, knowledge and telephony connectors, plus APIs and MCP for custom actions. Example use: retrieve an order record before giving a status update. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-integration-surface`)_ - The retail offering describes workflow instructions written in natural language, simulations and experiments. For example, a team can evaluate a revised return policy before rolling it out to all conversations. _(Vendor-reported, checked 20 Sept 2026; source: [Retail offering (Decagon)](https://decagon.ai/industry/retail); claim id: `decagon-workflow-controls`)_ - Decagon’s security page documents role-based access, single sign-on and short-lived API tokens. These are vendor-described controls; buyers still need to check how permissions are applied to their connected systems. _(Vendor-reported, checked 20 Sept 2026; source: [Security (Decagon)](https://decagon.ai/security); claim id: `decagon-security-controls`)_ ## Weaknesses - In a review dated August 10, 2026, Subhashree S. reports needing human correction for some complex questions. The review does not identify the deployment or measure an error rate, so it is a reported experience rather than a product-wide failure finding. _(Independent report, published 10 Aug 2026, checked 20 Sept 2026; source: [Subhashree S., G2 review syndicated by AWS Marketplace](https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy); claim id: `decagon-complex-query-limits`)_ - LOKESH G.’s August 20, 2026 review says configuring uncommon scenarios can take time. This suggests asking who will maintain workflow exceptions; it does not establish a standard setup duration or that other deployments have the same difficulty. _(Independent report, published 20 Aug 2026, checked 20 Sept 2026; source: [LOKESH G., G2 review syndicated by AWS Marketplace](https://aws.amazon.com/marketplace/reviews/reviews-list/prodview-pnbnafhjefhhy); claim id: `decagon-setup-burden`)_ - Watchtower documents flags and review of negative sentiment. That provides monitoring evidence, not proof that the agent resolves a complaint. Evaluate the escalation and remedy workflow separately. _(Editorial assessment, checked 20 Sept 2026; source: [Editorial assessment of Watchtower scope](https://decagon.ai/product/watchtower); claim id: `decagon-complaint-monitoring-limit`)_ ## Product information ### Deployment model - Decagon connects agents to existing support systems through prebuilt connectors and custom API or MCP integrations. The required setup depends on the systems and actions in scope. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-deployment`)_ ### Channels - Decagon’s homepage lists chat, email and voice. SMS appears in its Curology customer story; confirm whether that channel is available for your proposed deployment. _(Vendor-reported, checked 20 Sept 2026; source: [Decagon homepage and linked customer examples](https://decagon.ai/); claim id: `decagon-channels`)_ ### Languages - The voice page advertises 70+ languages, automatic detection and switching. Confirm language support for each required channel and evaluate real conversations in those languages. _(Vendor-reported, checked 20 Sept 2026; source: [Voice AI for customer service (Decagon)](https://decagon.ai/product/voice); claim id: `decagon-languages`)_ ### Help desk and CRM integrations - The integration page names Salesforce, Intercom and Zendesk for tickets and customer data, and Amazon Connect and RingCentral for telephony. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-helpdesks`)_ ### Knowledge sources - Decagon documents knowledge synchronization from Confluence, Contentful and Kustomer. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-knowledge`)_ ### Actions in backend systems - APIs, custom tools and MCP connections provide access to data and actions. The integration must expose the operation the workflow requires. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-actions`)_ ### Human handoff - Documented options include chat escalation, email routing and transfer of voice calls to a human agent. _(Vendor-reported, checked 20 Sept 2026; source: [Integrations (Decagon)](https://decagon.ai/product/integrations); claim id: `decagon-handoff`)_ ### Target customer size - Decagon presents an enterprise offering and publishes customer stories for Chime, Notion and Rippling. That focus does not establish a minimum team size or exclude smaller deployments. _(Vendor-reported, checked 20 Sept 2026; source: [Customer stories (Decagon)](https://decagon.ai/case-studies); claim id: `decagon-target-size`)_ ### Vendor-reported performance claims - Decagon reports a 65% reduction in support operating costs at Curology. This is a customer-specific vendor claim, not a resolution rate or a benchmark result; the page does not provide a common basis for cross-vendor comparison. _(Vendor-reported, checked 20 Sept 2026; source: [Curology customer story (Decagon)](https://decagon.ai/case-studies/curology); claim id: `decagon-vendor-metrics`)_ ## Product positioning Product focus: Service + sales. Product breadth: Connected suite. Editorial view of the documented offering, not a performance ranking or a revenue measurement. - Decagon’s retail offering combines billing and account support with upsell workflows. It also documents integrations, workflow configuration, simulations and analytics. We place that scope in service + sales and connected suite: for example, a retailer could evaluate subscription support and upgrade offers within one configured agent. This placement does not measure revenue or resolution quality. _(Editorial assessment, checked 20 Sept 2026; source: [Editorial assessment of Decagon retail offering (vendor website)](https://decagon.ai/industry/retail); claim id: `decagon-positioning`)_ Landscape: https://ai-agent.review/positioning.md ## Evidence base The independent evidence behind these states comes mainly from practitioner forums and buyer reviews. Independent sources behind these verdicts: aws.amazon.com. By kind: practitioner forums and buyer reviews (1). Verdicts across agents are not directly comparable while the evidence behind them comes from different kinds of source. See https://ai-agent.review/methodology#evidence-asymmetry. ## Change history Every change to this record is logged in the public changelog: https://ai-agent.review/changelog ## Corrections Factual corrections with a public source are accepted at corrections@ai-agent.review and answered within 14 days. Quote the claim id shown with each statement above. Method: https://ai-agent.review/methodology