Zurvey.io has long empowered users to gain a deeper understanding of their textual data through advanced aggregations and chart visualizations. Now, with AI-powered insight generation, this process is even quicker.
Please note: AI Insight features are premium functionality and may not be available in every Zurvey.io group. If you would like to try these features, please contact your account manager
Large Language Model - Tech & Security
Feature accessibility & AI Insight panel
Prompting tips for better results
Large Language Model - Tech & Security
The AI-powered features in Zurvey.io are driven by open-source models developed by leading technology providers such as Microsoft and Meta. These models are optimized for generating actionable insights from textual data.
Key Technical Highlights:
- Exclusive Use: The model is hosted on our dedicated virtual servers and is exclusively used by Neticle Plc., ensuring complete control over its operation.
- Hosted Locally: All data processing occurs within the EU, guaranteeing compliance with data protection regulations.
- Data Privacy: Your data is never used to train the model or shared with external systems, safeguarding your information.
- Secure Communication: Interactions with the AI engine are routed through a secure gateway, maintaining the integrity and confidentiality of your data.
Neticle's AI Engine may produce occasional inaccuracies. Please always verify insights with your actual data.
Feature accessibility & AI Insight panel
The AI Insight feature is accessible through the AI Insight card, located at the top of every page within the text analysis dashboard.
Limitation: If the average length of verbatims exceeds 2,000 characters, the AI summary function will not be available.
Clicking on this card opens a panel where you can choose from several AI-powered functionalities:
- Get a summary of verbatims.
- Reveal pain points in verbatims.
- Reveal love points in verbatims.
- Define a custom prompt to ask specific questions about your textual feedback.
Each AI function uses your dataset (verbatims) as context and generates insights accordingly.
Using filters for AI Insights
To achieve higher precision and efficiency in AI insight generation, we recommend applying filters to narrow down the data provided to our Large Language Model (LLM). Filters can be accessed at the top of the dashboard and will remain active within the AI Insight panel.
When you open the AI Insight panel, all active filters will automatically apply to the dataset used for insight generation. The selected filters are always visible in the filter bar, even while the AI Insight panel is open.
To make filtering easier, the system provides recommended filter options based on your dataset’s metadata:
- If your dataset includes Customer Experience (CX) metrics such as NPS, CSAT, or CES, the system will suggest segmentation filters for these metrics.
- To help you focus on topics rather than metadata, the system offers filters for the top five most recognized topics in your dataset.
- For continuous datastreams (e.g., API integrations, media intelligence, or email traffic), predefined interval filters are available to simplify segmentation.
Once you've selected an AI function and applied filters (optional), click the Get AI Insight button at the bottom of the panel. This button remains disabled until an AI function is selected.
AI insights are typically generated from a sample of your dataset, representing a focused yet representative subset. For more details on our sampling methods, visit the Sampling Methodology entry in the knowledge hub.
While your AI insight is being processed, indicated by a spinning icon on the button, you can close the panel and navigate elsewhere within the dashboard. A notification will alert you when the insight is ready. Please note that only one insight generation can be active per user per dataset at any time.
Custom prompt creation
In addition to predefined AI functions like summary generation, pain point reveal, and love point reveal, you can create custom prompts to ask specific questions tailored to your needs. Similar to other AI functionalities, your verbatims are provided as context alongside your custom prompt to the AI engine.
Please note your custom prompt can be up to 1,000 characters.
After saving your custom prompt, you can edit it anytime by clicking on the custom prompt element. Once modified, you can regenerate insights with the updated prompt.
Prompting tips for better results
To help you craft effective prompts and maximize the potential of the AI engine, we recommend the following tips:
Be specific and clear
- Use straightforward and concise language. Avoid vague terms or overly complex phrasing.
- Example: "Summarize the top three positive comments from the customer feedback provided as context."
Set expectations for the format and length
- Clearly specify both the structure (e.g., bullet points, numbered lists, or paragraphs) and the length of the response you want. This helps ensure the output is aligned with your needs.
- Example: "Provide your answer in a bullet-point list. Each point explained in one to two sentences."
Specify the language
Clearly state the language you require in the results. It’s best to emphasize this at both the beginning and the end of your prompt.
- Example:"You are an English-speaking, helpful AI assistant. How can the company improve customer service based on the feedback provided as context. Your answer will be needed in English."
Define the role for the AI
Provide a clear role description at the beginning of your prompt to guide the AI in its task.
- Example:"You are a Customer Experience Analyst tasked with generating actionable insights based on the customer feedback provided as context."
Refer to the context
Always mention the context (verbatims) in your prompt to ensure the AI focuses on the provided data.
- Example:"What were the main issues mentioned in the customer feedback provided as context?"
Avoid overloading the prompt
- Focus on a single task per prompt. If multiple insights are needed, consider breaking them into separate prompts.
- "Identify the top issues in the feedback provided as context."
- "Provide recommendations to address the call center problems ."
Test and refine your prompt
- Start with a broad question, review the AI’s response, and refine your prompt to improve focus or clarify ambiguities.
- Example: If the first response feels generic, add more details: "Focus on feedback related to customer service experiences."
Results
Once you receive your AI insights, you can rename them to better reflect your topic by clicking on the pencil icon next to the title. This updated name will also appear in the Insight History, where you can easily search for it later.
You can instantly copy your results for use in a presentation or other documents. Additionally, you can translate the insights to match the language of your input dataset. If your input dataset is also in English or contains multiple languages, the translation function will remain disabled.
You can download your AI Insight results as an .xlsx export, which includes:
- All verbatims from the sample used as context for the AI engine (located in a separate sheet under "Data subset").
- The selected AI function (or the full custom prompt, if applicable).
- The selected verbatim column, if multiple columns are available in your dataset.
- A record of all applied filters.
- The AI Insight result itself.
Insight History
To ensure your valuable insights remain accessible, you can revisit or regenerate them using the Insight History, accessible from the top left corner of the AI panel.
Within the Insight History, you can search for any previously generated insight by its name or creator. Each entry in the history list provides the following details:
- The name of the generated insight.
- The creator of the insight.
- The applied filters at the time of generation.
- The time of generation.
- The verbatim column used.
- The selected AI function (or the full custom prompt, if applicable).
From the history list, you can:
- Download exports of the AI Insight directly.
- Delete insights you no longer need.
By opening a previously generated AI Insight, you can adjust its settings, such as selecting a different AI function or modifying the custom prompt, and regenerate the insight. However, please note that regenerating an AI Insight will automatically override the previous version.
Drill Down
To enhance data discovery, AI-generated summaries are available for drill-down user flows. If you'd like to gain a deeper understanding of a specific datapoint on your dashboard, simply click on any label or chart. This will allow you to quickly generate a summary of the verbatims behind the numbers with the help of AI.
In addition to summarization, all other AI functions are available in this context. If you want to ask a specific question related to the drill-down data, you can define a custom prompt as well.
Keep in mind that you can close the drill-down popup after initiating the AI insight generation. A toaster notification will appear in the top right corner once your insight is ready.
You can also find your previously generated result in the Insight History, regardless of whether it was generated from a drill-down.
In the history view and within the AI panel, the applied filters will reflect the data you defined during the drill-down, ensuring that the AI summary is generated from the same data when regenerating the insight.
AI Comparison
AI support extends to another critical user flow: comparing two subsets of your dataset.
When the Comparison mode is active and exactly two subsets are selected, you can access this functionality through the AI widget.
Clicking on the AI Insight button will display the current subsets along with the filters defining them. From there, you can generate an AI Insight.
The AI engine will analyze the two subsets and provide a summary of their main similarities and differences. For clarity, the subsets will be referred to as Subset 1 and Subset 2 in the results.
As with other AI-generated insights, results from dataset comparisons are stored in the Insight History, allowing you to revisit or regenerate them later.
