Conversation AI is a question type powered by Skipper, aytm's AI assistant, that engages respondents in dynamic conversations. Capable of extending a short 1:1 interview and asking follow-up questions, Conversation AI is fully integrated into the aytm platform, enhancing research efficiency by eliminating the need for separate qualitative study setup, analysis, and additional recruitment costs.
To request early access to Conversation AI, please complete the form linked here. Our team will then enable Conversation AI in your account. Please note, if your account does not support AI, you will not be able to use Conversation AI.
Want to learn more about Conversation AI? Enroll in Best Practices for Conversation AI in the Lighthouse Academy!

1. Building a Conversation Question
- In the Survey Editor, click Add a question.
- Select Conversation question and type or paste a topic or prompt into the question field.
-
Click the talking point text field and type or paste up to 15 talking points. These will be the main topics you wish to discover during your research.
Note: If the total N is larger than the n of conversation, the predicted LOI will be calculated proportionally. However, keep in mind that while expanding the question with additional talking points can enhance the conversation, each additional talking point contributes to the LOI (Length of Interview), which could lead to respondent fatigue.
|
 |
2. Conversation Details
Beneath the Talking Points use the Collection, Depth, Input, and Style settings to further enhance your question.
-
In the Collect up to field, type the number of responses you would like to collect.
Note: The upper limit of responses you can collect is less than or equal to your total sample size. However, the number of responses you collect is truly up to your discretion. To obtain comparable insights, we recommend conducting three times the number of classic 1:1 interviews. Explore other important factors to consider when deciding on the number of responses in our guide, Best Practices for Conversational AI.
- Depth refers to the number of follow-up questions Skipper may ask for clarification per talking point. Click the Depth drop-down menu to select from the following options:
-
-
No follow-up questions: 0 follow-ups per talking point.
-
Get some clarifications: ~ 2 follow-ups per talking point.
-
Pursue more details: ~ 3 follow-ups per talking point.
-
Discuss each topic in depth: ~ 4 follow ups per talking point.
Note: Keep in mind the LOI. In practice, Skipper may decide to ask more or fewer follow-ups, depending on the quality of the response given.
3. Click the Input drop-down field to choose how data will be collected: in text only, text and audio, or audio only. Note: Learn more about the respondent experience using audio here.
4. Click the Style drop-down to select how Skipper conducts the conversation:
-
-
Select Formal if you prefer a straightforward and conventional approach.
-
Select Engaging if you enjoy using smileys and slang in conversations.
-
Select Adaptive if you want Skipper to adjust to the respondent's tone.
|
 |
3. Conversation Quick Preview
It is always best to preview your entire survey before launching, and particularly important to preview more complex survey elements like Conversation AI.
Previewing allows you to:
- Experience the survey firsthand.
- Evaluate the flow.
- Ensure all desired information is gathered.
- Determine or experiment with the appropriate number of follow-up questions.
When you’re ready for a trial run:
-
Click Preview above the question to see the Conversation AI in action.
Note: Each preview and survey session is unique, much like a real human conversation. Keep this in mind while previewing, as Skipper may respond slightly differently each time, even when the topics, prompts, and replies remain the same.
|

|
4. Respondent Experience Using Audio
When Respondents arrive at this question and you have an audio only option enabled, they will first see a prescreener to confirm whether their audio input is working. After the prescreener, Skipper will begin the conversation
-
Response Options: Respondents can type or voice record their responses, as allowed in the question settings. If they choose the audio option they can:
- Click the microphone button to start speaking. Skipper records their voice for up to 60 seconds and transcribes the response. If the recording exceeds 60 seconds, it stops automatically and sends the response.
- Click the Play icon, allowing respondents to listen to the audio they just sent.
Note: Respondents may still choose to skip the question once they’ve started it by clicking the Skip button at the beginning of the conversation with Skipper.
|
 |
5. Viewing and Analyzing Responses
Once the survey is live, responses to the Conversation question will be displayed on the results page as an open-ended question. You will see an overview of direct quotes, or view responses as a chart categorized by topic, or as a word cloud.
- Click the first drop-down to isolate responses to individual talking points.
- Click the middle drop-down to change the visualization:
- Select Quote tiles to view individual quotes from respondents.
- Select Responses per topic to view a graphical representation of the various talking points
- Select WordCloud to view a visual display of the most frequently used words and common themes in respondents' conversations.
- Use the search field to search for keywords within responses.
Note: If you’d like to enable Skipper Autocoding for your survey responses, keep the Skipper Autocode box checked at the bottom of your survey editor page.
|
 |
6. Advanced Features and Audio Options on the Results Page
- To view the full conversations in the Open-Ended questions interface, click the comment bubble relative to a question, or at the top of the page. Responses are sorted by Respondent ID, each representing a single conversation that took place between a respondent and Skipper.
- If you have Sentiment Analysis enabled for your survey, sentiments may be applied to each individual talking point within a single conversation. As in other open-ended response questions, use the menu along the top of the response pane to sort and filter responses, and utilize the verbatim coding menu to tag and organize the data.
- Click the gear icon to select an accent for Skipper's voice and a different accent for the respondent's voice, then press the headset icon to begin playing the text. With a conversation playing, click the Stop button to stop the audio.
- To download the audio, click Download audio, and a separate tab will open, allowing you to download the audio as an .flac file.
|
 |