Zest AI — Automatic Sentiment and Theme Analysis

4 min read

Take action easily thanks to automatic theme and sentiment analysis on the text-based responses (comments and verbatims) from your Surveys and Engageomètres.

What is Zest AI?

Zest AI is the artificial intelligence built by Zest to simplify the analysis of Survey and Engageomètre results. Today, it applies mainly to text-based data (comments and verbatims), along 3 main areas:

  • Detection and alerting on comments identified as sensitive
  • Automatic analysis of the themes raised in comments and responses
  • Automatic analysis of the sentiments expressed in comments and responses

Zest AI is custom-built to simplify results analysis for managers and HR teams, drawing on the expertise of the Zest teams.

How does Zest AI's Engagement / QWL analysis work on text-based data?

Automatic analysis of text responses relies on a neural network capable of understanding the text responses written by employees within surveys. Trained on thousands of responses and supervised by the Zest Lab teams, this network — focused on the topic of engagement and quality of work life (QWL) in companies — detects the themes and sentiments expressed in the responses.

1. The theme(s): these are predefined by our teams and used to categorize as many responses as possible into one or more themes.

For example, the response "The atmosphere at work motivates me a lot because we have a good team spirit, as well as the remote work policy introduced 6 months ago" could be categorized under both the "Workplace atmosphere" AND "Remote / Hybrid work" themes. In the results view, we provide the recurrence rate of themes across responses (for example: "Workplace atmosphere: 25% of comments categorized by Zest AI mention this theme").

2. Tone (sentiment): reflects the sentiments expressed (positive, neutral, or negative) across a set of responses. It is a key indicator of opinion trends that helps guide analysis and action choices.

While the categorization of responses by theme and tone is performed by an AI trained and hosted by Zest, human analysis must remain critical when reviewing the results provided. Zest AI is a tool and a support — it should not replace the human, expert analysis carried out by roles such as HR teams, managers, etc.

How do I read Zest AI's results?

No more need to read every response one by one! Zest AI's results are displayed by default on your survey or Engageomètre results page, allowing you to quickly analyze a large volume of data so you can take action easily, without having to review each response individually.

First, quickly view the themes to prioritize and to highlight in order to strengthen employee engagement, thanks to this intuitive, straightforward overview. You also get access to theme-specific content to support your reading of the results.

Next, to go further, Zest provides a prioritization matrix divided into 4 main areas, covering all the themes detected within the responses:

  • Themes to prioritize: frequently discussed negatively
  • Themes to highlight: frequently discussed positively
  • Themes to watch: rarely discussed negatively
  • Themes to develop: rarely discussed negatively

These first levels of analysis let you prioritize your reading of responses related to a specific theme. You can then go deeper by accessing the response details and the overall tone per theme.

This analysis is also available in the exports offered by Zest:

  • PowerPoint export: as a visual chart
  • Excel export: as a consolidated table of themes discussed, as well as response by response

Full list of Engagement / QWL themes in Zest AI

Below is the complete list of themes used to categorize responses and comments in the Engagement / QWL analysis.

AutonomyWorkloadWorking hours / SchedulingCooperation within the teamWorkplace safety
Interest and meaning of daily workCross-team cooperationWork-life balanceGoal setting and evaluationCompany project, strategy and values
Clarity of processes and everyone's rolesDigital environment (tools, communication...)Career developmentFeedbackInternal communication / Company rituals
Everyday recognitionPhysical work environmentTrainingOrganization / Distribution of work within the teamDiversity
Relationships with stakeholdersOther resources: headcount and budgetsSkills developmentManagement role and postureCSR policy
Compensation / Employee benefitsRemote / Hybrid workInterest in the jobInitiative-taking and right to make mistakesOrganization / Transformation / Change
Team functioning (meetings, rituals...)Onboarding of new hiresInterpersonal relationships within the companyWorkplace atmosphereSense of belonging / Employer brand
Inappropriate behaviorMental and physical healthCurrent topicsNo opinion / N/AOther topics
RecommendationGeneral feeling

Free-theme analysis by Zest AI

With free analysis, Zest AI automatically identifies and generates the themes emerging from text responses without relying on a predefined list, and also detects the sentiments expressed.

The analysis happens in two steps:

  1. Zest AI detects the different themes as responses come in.
  1. It groups them into broader themes and presents them at the end of the survey session for optimal analysis.

The output of the free analysis follows the same format as the Engagement / QWL analysis.

This analysis is useful for exploring broader topics and complements the Engagement / QWL analysis, allowing you to adapt the level of analysis to the type of questions and surveys. To use it, simply activate it from the Back Office; survey creators can then configure it on the questions of their choice when designing the survey.

💡
Good to know: it is not possible to activate free semantic analysis retroactively or after your survey has launched.

🧐 Any questions? Feel free to contact your Zest representative for more information

Related articles

Was this page helpful?