Static survey tools render one question at a time, which already feels lighter than a long form, but they cannot react to an answer. SleekAI can ask a follow-up on a low NPS score, skip a question when the previous answer already covers it, and acknowledge the reply before moving on. The data shape is similar at the end, but the conversation in the middle is what lifts completion rates and the depth of free-text answers.

Yes. The bot can ask any 0-10, 1-5, or 1-7 scale and store the result as a numeric field. NPS, CSAT, and CES are just different scale plus prompt combinations. The free-text reasoning the bot captures alongside the score is usually the part that delivers the actual product insight: a 7 with 'wizard kept resetting' is a roadmap item, while a 7 with 'no reason really' is noise.

Yes. Display conditions scope a chatbot by post type, taxonomy term, user role, logged-in state, or URL pattern. Common patterns: trigger on the dashboard after 14 days of usage, on the cancel-page before a churn event, or on a docs page after a known support workflow. Each scoped survey runs as its own chatbot via multibot, with its own questions and tone.

Conversations are stored in your WordPress database with structured fields for each numeric or multi-choice answer, plus the full transcript. From there, common paths are CSV export for ad-hoc analysis, a webhook to Segment or RudderStack, or a direct push to your data warehouse. The structured fields mean the warehouse table can stay clean even though the underlying conversation is free-form.

Yes. The bot only stores user identity if you ask it to. For anonymous research surveys you can configure the prompt to never ask for name or email and to attach a synthetic id to the record. For logged-in customer surveys, the WordPress user id is associated with the record automatically, which is usually what NPS reporting needs.

The initial message and the system prompt should be transparent: 'this is a SleekAI-powered survey'. That clarity actually tends to lift response quality, because respondents understand they can be specific and the response will be processed structurally. The conversational pace then does the rest. Pretending to be a human is both unnecessary and against the guideline filter defaults.

Yes. Multibot scopes a different chatbot per cohort, page, or campaign, each with its own discussion guide, tone, and exit criteria. A win-back survey for cancelled users, an onboarding survey for new signups, and an annual NPS for all customers can run as three separate bots from one WordPress install, all logging into the same database table for cross-cohort comparison.

Partial conversations are still logged with whatever answers the visitor gave before leaving. For NPS that often means the numeric score arrived but the reasoning did not, which is still useful data. The conversation can be resumed within the same browser session via the widget's persistence. After the session ends, an incomplete record is just an incomplete record, the same as a partial Typeform submission.