⇒ Overview
A new Conversational AI question type is being introduced to the Pollfish dashboard. This feature allows you to include dynamic, interactive AI-driven questions that engage respondents in deeper, contextual conversations.
Unlike traditional survey formats, Conversational AI can adjust based on respondents' answers, making the experience feel more natural and less scripted. These questions are useful for gathering nuanced feedback or exploring more complex topics in a way that feels intuitive to the participant.
As with all question types on Pollfish, Conversational AI can be used alone or alongside other formats within the same survey.
⇒ Conversational AI and Qual Research
Conversational AI helps bridge the gap between qualitative and quantitative research—offering the ability to gather deeper, more nuanced feedback while still reaching respondents at scale.
Instead of relying solely on structured question types, Conversational AI allows for dynamic interactions that surface richer context and more authentic responses. It brings the flexibility of interviews into the structured world of survey research, without compromising efficiency.
⇒ How It Works
You can insert Conversational AI questions into your survey the same way you’d add any other question type. For example, you can create a 10-question survey that includes 1 Conversational AI question, or even a 1-question survey consisting solely of a Conversational AI question.
You can add 1 Conversational AI question per survey, and there is a limit of 6 goals per Convo AI question.
⇒ Setting Up a Conversational AI Question
Researchers start by setting a primary goal (objective) for the AI-driven interaction — this is the main insight they want to gather.
They’ll also need to add a set of sub-goals (between 1 and 6), either manually or with help from our AI question builder. The previous will guide the AI in shaping the conversation and extracting relevant responses.
Conversational AI will respect the survey language selected in the Audience tab and conduct the conversation accordingly.
Media files are not currently supported within Convo AI questions.
Example Q1 (Conversational AI):
Objective: Understand how the survey participantst are reallocating their budgets in response to inflation.
Goal 1: Gauge the impact of inflation on brand loyalty and product preferences.
Goal 2: Identify changes in shopping habits (e.g., bulk buying, purchase frequency, switching to cheaper alternatives).
Goal 3: Explore how rising gas prices are influencing travel and transportation decisions.
Goal 4: Assess growing interest in private label or store-brand products.
⇒ Respondent Experience
Respondents will answer a series of open-ended questions in a conversational format. Initially, their responses will be provided in text, with a future version supporting voice responses. The Conversational AI technology will capture their responses in real-time.
At the end of each question, the system will evaluate the response to determine the outcome of the participant’s engagement. If the response is deemed successful, the system will proceed to the next question. If the response is considered of insufficient quality, nonsensical, or unrelated, the participant will be disqualified.
If a participant responds with a one-word answer that doesn’t address what the AI is trying to learn, Conversational AI will follow up and prompt them to elaborate, ensuring it gathers meaningful information.
If participants provide information relevant to multiple goals while answering a single question, the Conversational AI will recognize this and avoid asking repetitive questions—creating a more seamless and efficient experience from the participant’s perspective.
⇒ Results & Exports
The raw conversation between the convo AI system and respondents will not be accessible to you. For each respondent, the Conversational AI technology will analyze their answers and provide a summary of each goal, as well as an overall summary of the main objective.
Additionally, on the results page, summaries of the main objective and its associated goals will be available for all selected respondents.
On the results page, you'll find:
- Objective summary per respondent
- Summary per goal per respondent
- Summaries for selected respondents (Objective and Goals)
In your exports, you will find both a summary per goal/respondent and an objective summary per respondent, along with a summary of goals and the overall objective for selected respondents.
For our Elite clients using BigQuery and our Insights Builder, a summary per goal/respondent and an objective summary per respondent will be available.
Crosstabs are not available, as open-ended questions cannot be used for crosstab analysis.
⇒ Skip Logic
Logic cannot be applied during the AI conversation to direct participants based on their open-ended responses. However, once participants have completed all AI questions and the interaction is finished, logic can be applied to guide them all to a next question if needed—just as Logic can be used with question groups or A/B test blocks.