Qualtrics

Once approved, Experience Agents perform ticket actions, including status updates, escalation and reassignment, and they create tickets of their own. These are writes into the XM Platform's system of record, not lookups. They read across survey, call, chat and digital experience analytics data, so a single agent draws on several data types in one decision.

Outside the agent layer, the platform has a public API with SDKs, Postman collections and workflow tooling for managing surveys, users and results, which gives a customer's own systems a route in and out. The agent pages show no connectors into third party systems of record such as a CRM or commerce platform, so the agents act in Qualtrics' own platform, not the customer's wider stack.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

In Agent Studio a specific agent is created for each workflow and trained on that company's policies and runbooks. The Ticketing Agent runs a real sequence with conditions. It summarizes the ticket, recommends an action, drafts the response, waits for approval, then executes the approved ticket action, escalating or reassigning where the case requires it.

Agents coordinate handoffs to people when oversight is needed. Survey agents resolve an issue inside the response flow before it becomes a ticket at all, so orchestration spans two surfaces. Experience Agents are built on LangChain's LangGraph platform, a framework under the hood that the customer does not work with directly.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

Qualtrics grounds Experience Agents in its proprietary human experience data, roughly twenty years of customer and employee feedback captured across hundreds of thousands of touchpoints. This maintained corpus is the product's main differentiator.

Qualtrics sets the agents against transactional agents built on systems of record, claiming the agent understands why an issue occurs, not only what to execute. Each agent is also trained on the individual company's policies and runbooks, so answers reflect that business's rules, not generic guidance. Agents draw on survey, call, chat and digital experience analytics across the platform. The corpus and the runbooks persist and are maintained.

Sourcequaltrics.com/platform/experience-agentsread 2026-09-05

A person approves before the agent sends or acts. The Ticketing Agent recaps the situation, recommends an action and drafts the email response, and a person approves before anything is sent or executed. Only once approved does the agent carry out ticket actions such as status updates, escalation or reassignment. Compensation has limits.

Agents apply make goods within guardrails the brand sets, so a service-recovery gesture cannot exceed what the business authorized. Handoffs are coordinated when human oversight is required, and Qualtrics cites Stanford Health Care testing patient care coordination explicitly under human oversight.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

Attestations include SOC 2 Type II covering security, availability and confidentiality, ISO/IEC 27001, 27017, 27018 and 27701, and ISO/IEC 42001:2023, the first international standard for AI management systems.

Qualtrics also holds HITRUST certification for healthcare, TISAX for the German automotive sector (retrievable through the ENX portal), and PCI DSS certification on the XM Discover voice-of-customer integration for call recordings containing card data. Customers can get the reports through a Trust Center and a self-service repository for the SOC 2 report, the ISO certificates, the HITRUST validated report and the PCI attestation.

Access controls include single sign-on authentication, enforced multi-factor authentication, user log and account activity monitoring through the Control Panel and APIs, password complexity controls, PII redaction and collection restriction, and customer-controlled retention with one-click deletion of personal data across responses, tickets and contacts for GDPR erasure requests. API-side controls add IP allowlisting and mutual TLS. FedRAMP High and IL4 authorizations add sovereign delivery environments.

Sourcequaltrics.com/security-statementread 2026-09-05

Every action an Experience Agent performs on a ticket is visible in the AI activity log within Tickets, and the agent's recommendations and actions appear directly inside the tickets it creates, so a supervisor can reconstruct what the agent did and why without leaving the workflow. Underneath, the platform offers platform-level user log and account activity monitoring through the Control Panel and APIs. That log does not capture all read access to brand directories, and Qualtrics routes such requests to its Security Operations Center.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

Context persists in the platform, but the agent has no memory layer of its own. Experience Agents read an individual's prior interactions from the XM Platform and predict trends from that experience context, so an agent meeting a customer for the second time knows what happened the first, and the record survives across sessions because the platform holds it.

That is retrieval from a store the customer already owns, not state the agent forms and keeps. There is no memory product, scope, lifetime or deletion control for state the agent holds. Qualtrics says each interaction adds to the intelligence, which describes learning absorbed into a model, not a memory layer.

Sourcequaltrics.com/platform/experience-agentsread 2026-09-05

Qualtrics has a dedicated Data Sovereignty page. It operates multiple data centers with load balancing between them and independent third-party audits of the data center providers, and an additional isolation tier keeps a customer's data fully walled off in a virtual environment. Placement is set per account and visible to the customer.

Each account sits in a specific named data center, and developers select the API server that matches it. Distinct government environments sit above that, FedRAMP High for federal civilian agencies and Impact Level 4 for the Department of War, so a public-sector customer is deployed into a separately authorized environment, not the commercial one.

Sourcequaltrics.com/platform/security/data-sovereigntyread 2026-09-05

Specialized agent types ship ready to configure, but they are types, not a catalog. Survey Experience Agents act inside a response the moment feedback is given, and a Ticketing Agent summarizes, recommends actions, drafts responses and performs approved ticket actions, with more specialized agents interacting across touchpoints.

A customer starts from a working agent shape, not an empty canvas. There is no template gallery, agent library or marketplace that a customer browses and puts into service, and the build route points the other way, since a specific agent is created for each workflow in Agent Studio.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

Experience Agents act inside a survey the moment feedback is given, so the trigger is the customer's own words arriving, not a ticket being raised. Digital agents detect friction across a journey and step in instantly, so the platform starts a conversation off a behavioral signal the customer never reported. Ticketing agents fire within agent-enabled ticketing workflows as cases arrive and change state. All three starts are automatic and event-driven. Channels are narrow, running through surveys, tickets and instrumented digital touchpoints, not voice, SMS or messaging apps such as WhatsApp.

Sourcequaltrics.com/platform/experience-agentsread 2026-09-05

There is no model selection, routing configuration or bring-your-own-key path, and no model provider is named. Experience Agents are built on LangChain's LangGraph platform, an orchestration framework Qualtrics builds on, not a model the customer can choose or point their own account at. The open source framework Qualtrics says it is developing concerns interoperability between agents, not choice of model, and it has not shipped.

Sourcequaltrics.com/platform/experience-agentsread 2026-09-05

Qualtrics runs a Developer Portal at developer.qualtrics.com with registration, tooling and documentation for building in its ecosystem. A public API reference at api.qualtrics.com offers development kits for building applications, public API collections in Postman, and advanced connection controls including IP allowlisting and mutual TLS.

The support documentation has an interactive API tester with per-request parameter tables, cURL examples and code snippets in Python and Java for managing surveys and users and retrieving results. All of it is reachable without a login. Qualtrics also says it is developing an open source framework for cross vendor agent interoperability, which has not shipped.

The developer surface belongs to the XM Platform, and there is no agent-specific API for authoring or invoking Experience Agents from outside, so an integrator reaches the platform, not the agents.

Sourceapi.qualtrics.comread 2026-09-05

There is no simulation, test environment, scenario suite, regression run or scored pre-release evaluation of an Experience Agent, and Agent Studio is an authoring environment with no published test surface. Synthetic Research, which tests CX scenarios artificially and gives feedback on them, is a preview, not a shipped capability. Qualtrics has deep testing tooling for surveys, but what it tests is a questionnaire, not the agent.

Sourcequaltrics.com/support/experience-agents/experience-agents-overviewread 2026-09-05

There is no browser control, hosted session or interface automation. Experience Agents act inside surveys, tickets and instrumented digital touchpoints, and their outward reach is programmatic, through approved ticket actions written into the XM Platform and the platform's public API for anything beyond it. Digital experience monitoring observes a customer's journey to detect friction, which instruments the customer's own session and does not mean the agent operates a browser.

Sourcequaltrics.com/platform/experience-agentsread 2026-09-05