How does the AI probing work?
The AI listens to each participant response and generates a follow-up probe based on your probing goal (defined in the Probing Guide field), the participant's actual answer, the overall research objective, the interview audience, and the configured tone of voice. The AI tracks all previous answers in the conversation to avoid redundancy. Probes stay strictly on-topic and dig deeper rather than changing the subject.
Can I switch off AI probing?
Yes. Set the number of probes to 0 for any question or all questions. The AI will then operate in scripted-only mode with no follow-up probes.
Will the AI ask off-topic questions?
No. The AI is constrained by strict system prompts developed over years of refinement. It stays within the scope of your discussion guide and redirects participants who go off-topic back to the relevant questions.
Why did the AI probe more times than I configured?
The system has quality guardrails that may trigger additional probing when a participant gives a very short or unclear answer. This typically resolves when participants give more substantive responses. If it is a persistent issue, contact [email protected].
Does the AI learn from my projects?
No. The AI does not learn from individual projects. Your interview data is never used to train Tellet's AI or any third-party models.
Will my data be used to train AI models?
Tellet is built on Microsoft Azure OpenAI, which guarantees that none of the data processed will be used for training purposes by Tellet or by Microsoft.
Can I see all the probes the AI asked?
Yes. All AI probes appear in the full transcripts (Transcripts tab) and in the Questions section under the Answers tab alongside the participant responses.
Can the AI misinterpret what someone says?
Like any moderator, the AI can occasionally mishear something. When unsure about a response it will typically ask for clarification rather than make assumptions. You can review all transcripts to verify the accuracy of transcriptions.
What is the difference between the AI moderator and a human moderator?
The AI moderator is consistent across every participant, available 24/7, processes interviews in parallel at any scale, and delivers analysis automatically. A human moderator brings intuition, cultural nuance, and deep topic expertise. Tellet is most powerful when the AI handles the scale and consistency layer while researchers focus on strategy, discussion guide design, and insight interpretation.
Where do I add topic restrictions or legal guardrails that don't fit in the Research Objective field?
Add them as a context snippet in Context & Training, not in Research Objective. Research Objective is capped at 700 characters, which is too short for detailed compliance rules or a list of prohibited topics. A context snippet holds up to 1,400 characters, and you can add more than one, so put your guardrail instructions there, for example: "Do not discuss competitor pricing or pending litigation. Decline requests for medical, legal, or financial advice and redirect to the next question." The AI follows snippet content as instructions for the whole interview, the same way it follows your terminology and product background. See Context & Training: Complete Guide.