AI at Usertick

1. Features that use AI

• Reading of responses. On every Plan, Usertick’s AI reads the responses that have enough to read, such as a written answer. It gives answers and responses readings such as category, theme, sentiment and urgency, which appear in the inbox and on the Results page and are available to workflows, the REST API and webhooks.
• Written readings of results. When a period ends with enough responses, AI writes a short reading of what changed and what to do, in English.
• Help with building. When a person asks for it, AI drafts surveys from a chat, imports pasted questions, suggests and rewrites questions, answer options and texts in the editor, translates surveys, drafts dashboards, workflows, folders and categories, and answers questions in the assistant. Each Plan includes monthly allowances, listed on the pricing page.
• Checks of drafted surveys. Before Usertick shows a survey or questions that AI drafted, AI checks them for requests for passwords or other secret values, and for hateful or deceptive content.

The people who answer surveys never talk to an AI. Surveys show the questions and texts that the customer published and Usertick’s fixed built-in texts, also in the Chat layout.

2. Where AI runs

• Generative features, such as drafting, translation, written readings and the assistant, run on Microsoft Azure OpenAI Service in the United Kingdom. If a model is not offered in the United Kingdom, Usertick may run it in a Microsoft Azure region in the European Union.
• The reading of responses runs at TypeSafe AI, Inc. in the United States, under a data processing addendum that includes the EU standard contractual clauses and the UK Addendum to them.

Both providers are listed on the sub-processors page, and section 13 of the DPA describes the transfer to the United States.

3. What data the AI receives

• For the reading of a response, Usertick sends its answers with the texts of its questions, the survey’s name and the organization’s categories and themes, without the respondent’s identifiers, the identity a website sends or the contact details a contact question collects. Written answers are sent as the respondent wrote them.
• For the generative features, Usertick sends what a person types, the content of the survey, dashboard or workflow concerned, and figures such as response counts. Written readings and the assistant also receive short excerpts of written answers.
• The outputs, such as readings, drafts and translations, are stored in Usertick as the customer’s data, under the same protection and deletion rules as the rest of it.

4. No training on customer data

Usertick does not train AI models on its customers’ data, and its contracts with its AI providers don’t allow them to. Usertick may create anonymized, aggregate statistics that identify no person and no customer, as the Terms of Service describe.

5. Review what AI produces

AI outputs are generated automatically and can be wrong or incomplete. Customers should review an output before relying on it, and must not use one as the sole basis for a decision that has legal or similarly significant effects on a person. The readings make no decisions about the people who answered.

Usertick shows in the product which readings, drafts, suggestions and translations come from AI. The REST API and webhooks return the readings in their own field, which shows when a person corrected them.

6. Human oversight and control

• Drafts wait for a person. A survey that AI drafts opens in the editor and is created as a draft, which a person publishes. A drafted workflow is created switched off, and a drafted category is added only when a person adds it.
• People can correct the readings. Members whose role can manage responses can correct most readings of a response. A correction replaces the AI’s value everywhere and can be undone. Readings that describe how the respondent felt, such as satisfaction and effort, can’t be changed.
• Roles decide who uses the building features. The survey, dashboard and workflow helpers need the permission to manage surveys, dashboards or workflows. The assistant is available to every member and answers with the data that member’s role can see.
• The reading of responses is always on. It runs on every survey and every Plan, and no setting turns it off for a survey or an organization. Customers decide what it reads by choosing the questions they ask, and they tell their respondents about it in their own privacy notice.

7. Questions

Questions about how Usertick uses AI can be sent to privacy@usertick.com.