Siena AI

Breadth across the commerce stack with authenticated write action.

Named integrations span helpdesks including Zendesk, commerce with Shopify, subscriptions with Recharge, reviews and loyalty with Yotpo, post-purchase with AfterShip and Loop, and fulfillment with ShipBob, delivered through an App Store reachable from the Siena dashboard, with a Custom Apps category alongside.

The agent writes rather than reads: it issues refunds, updates shipping addresses, generates return labels, sends replacements, and skips, pauses or swaps subscriptions, all of which require authenticated action in the brand's own systems.

Sourcesiena.cx/integrationsread 2026-09-05

Multi-step resolution inside the customer's commerce stack, published as complete flows rather than capabilities: from a where-is-my-order question through to starting a return, Siena updates addresses, generates labels, issues refunds and sends replacements in a single flow, and subscription management runs skip, pause and swap while capturing the reason and offering alternatives. Order tracking is native across more than two thousand carriers. Agents also compose commercially, making in-the-moment product recommendations against a customer's order history.

Sourcesiena.cxread 2026-09-05

Grounding runs on the brand's own material through a named knowledge product: Docs is published as the operating system for a brand's knowledge, and agents answer from the brand's knowledge, standard operating procedures and automation library, with Memory supplying customer-level context to the same answer. No index, embeddings layer, retrieval architecture, refresh behavior or source-level citation is documented anywhere, so a maintained retrieval structure is asserted rather than shown, and the Docs product has a launch post rather than documentation.

Sourcesiena.cxread 2026-09-05

Oversight is documented across the handoff and the configuration surface. Siena publishes enterprise-grade safety and brand control as guardrails plus granular controls that keep every interaction on brand and on policy, escalates to human agents, and routes handoffs through Collaborative AI Routing which attaches an explanatory note so the person picking up knows why.

Copilot drafts on-brand replies and translations for human agents rather than sending autonomously, and AI Assistant lets a team build or adjust workflows in minutes. Guardrails and granular controls are named rather than described, and no pre-send approval step, sign-off surface or policy threshold holding a sensitive action for a person is documented.

Sourcesiena.cxread 2026-09-05

Compliance is documented thoroughly; customer-facing controls are not. A live trust center publishes SOC 2, GDPR, CPRA and HIPAA as compliant, with SOC 2 Type II reports for 2023, 2024, 2025 and 2026 available under access request alongside a Data Processing Agreement and NDA, an AICPA SOC badge carried in the site footer, and 62 named controls published in full across product, data, network, application, endpoint and corporate security.

Governance roles are named: an Information Security Officer, a Data Protection Officer and an EU Representative, with annual risk assessments, vendor risk assessments and a Data Protection Impact Assessment, and the terms of service describe personnel training on AI governance including EU AI Act requirements.

All 62 published controls are internal governance over Siena's own personnel and infrastructure, covering staff access provisioning, offboarding, endpoint encryption and production database restriction. Not one is a control the customer configures or operates, and no customer-facing single sign-on, SAML, customer-administered role-based access, customer-visible audit log or configurable retention appears anywhere.

Sourcesienaai.trust.siteread 2026-09-05

Reporting and handoff transparency are documented, reconstruction of a run is not.

Topics Explorer is published as a voice-of-customer tool built on real conversations, surfacing emerging trends and issues; Ask Siena answers questions about the conversation corpus and produces shareable reports; Memory Insights gives cross-functional teams visibility into what customers said; and Collaborative AI Routing explains every handoff to a human with a note, which is a legible reason at the one moment a person takes over.

No step-level decision trace, tool-call log or retrieval record for an individual agent action is documented, so a buyer can see what the agent concluded and reported but not reconstruct how a given interaction was resolved. The QA Agent that would score every conversation is marked coming soon.

Sourcesiena.cx/topics-explorerread 2026-09-05

Siena Memory ships as a named product described as the agentic brand intelligence layer, and it works as a memory layer rather than a store of connected records. Scope is not merely stated, it is configured by the customer: brands define Memory categories that set what is captured, with published live examples including preferences, feedback, milestones, birthdays, shade and color choices, subscription opinions, referral advocacy and inventory complaints, each written as a capture instruction.

Captured detail builds a 360-degree customer profile that the agent references in future interactions, and the vendor publishes a worked example spanning five weeks between two conversations in which the agent recalls a sizing remark the customer made in the first. Customers describe using Memory categories operationally, one routing tariff and price-increase mentions to a senior associate. No retention window, memory lifetime or per-customer deletion control is documented, so a buyer can name and scope the memory but not say how long it is held.

Sourcesiena.cx/memoryread 2026-09-05

Every entry on the trust center's published sub-processor list is located in the United States of America, with AWS named as the IT infrastructure provider, and no region menu, VPC, single-tenant, dedicated-instance or self-hosted option appears on the site or in the trust center's nine sections.

The one control touching location, Transfer of PII, commits to procedures for compliance with regulatory requirements when personal data moves outside the region it was collected from, which is transfer compliance rather than a location a buyer selects. The trust center itself is public, and the Data Processing Agreement sits behind an access request.

Sourcesienaai.trust.siteread 2026-09-05

A named set of agents the buyer adopts by role, published each with its own product page: a Customer Service Agent, a Social Agent and a Reviews Agent are live, with a Shopping Agent and a QA Agent marked coming soon. Alongside them sit AI Personas, which brands configure with their own tone and voice and deploy per channel, and Memory categories, which ship as named capture templates a brand selects and applies. The App Store supplies prebuilt connector apps from the dashboard.

Sourcesiena.cxread 2026-09-05

Breadth across five channel classes with the agent both receiving and initiating. Work reaches Siena over email, live chat, SMS, WhatsApp and social media direct messages and comments, in more than a hundred languages, and the agents initiate rather than only responding: the Social Agent engages shoppers on social channels and moderates comments, and the Reviews Agent puts review management on autopilot, both of which act on content the customer never sent to support. No voice or phone channel is offered.

Sourcesiena.cxread 2026-09-05

A single named model provider, and no customer choice. The trust center's sub-processor list names OpenAI under Artificial Intelligence as the only model provider serving the product, and no model selection, routing control or bring-your-own-model option is exposed to the customer on any surface. Anthropic also appears on the same list, categorized as Development software, which is Siena's own internal engineering tooling rather than a second provider serving customer conversations.

Sourcesienaai.trust.siteread 2026-09-05

No documented way into the platform from outside. An App Store exists and third parties are invited to build against it, with the announcement asking developers interested in building an integration or publishing their own app to write to a developer address, and a Custom Apps category appears on the integrations panel. But the route in is an email conversation: no public API reference, SDK, developer portal, documentation subdomain or Model Context Protocol server is documented. A partner program reached by contacting the vendor is not a documented interface.

Sourcesiena.cx/blog/siena-app-store-announcementread 2026-09-05

A test surface is documented and included in the commercial model: the pricing page states that the platform fee buys access to the core AI engine and an unlimited sandbox, so a brand can exercise agent behavior without limit before it reaches customers, and workflows can be built or adjusted in the admin interface with AI Assistant and refined against live performance.

No scored evaluation, scenario set, regression suite or readable verdict on a proposed change is documented anywhere, so the sandbox is a place to try things rather than a gate that produces a comparable result. The QA Agent that would score every conversation automatically is marked coming soon.

Sourcesiena.cx/pricingread 2026-09-05

No browser control, hosted session, virtual desktop or remote computer control appears anywhere on the site. Agents act programmatically instead, through App Store connectors into Shopify, the brand's helpdesk, subscription, returns and fulfillment systems, and reach customers over email, chat, SMS, WhatsApp and social channels.

Sourcesiena.cx/integrationsread 2026-09-05