AI reads, transcribes and translates. Your team decides.

AI in EnaFeedback does three jobs: understand the tone of text, turn audio and video into text, and translate. These are managed with six feature switches: sentiment analysis, voice transcription, video transcription, dynamic content translation for visitors, comment and message translation for admins, and transcript translation. The first three are on by default; the translation features are off.

AI does not route, close or prioritize tickets. Priority follows a fixed rule based on the category the visitor chose; routing and closing are done by admins and resolution teams. AI helps the team read and understand faster. There is no automatic summarization feature.

Sentiment analysis reads the tone of the text a visitor writes with a hygiene rating or feedback. Admins see next to the record whether a comment expresses satisfaction or frustration. Whatever the category or score, the words themselves are often the clearest signal.

The same analysis drives the "Positive outcome" block, an optional thank-you message you set to Always, Conditional or Never for each rating or category. With Conditional, the thank-you content is shown based on the tone of the text, and a "tolerance" setting decides what happens when the sentiment is neutral or unavailable.

A voice recording or video attached to feedback is transcribed after it is processed on the server. Admins and resolution teams can read the content without playing the file, which saves real time in a noisy ward or between meetings. Voice notes that staff leave in the field app are transcribed the same way.

Transcription accepts audio up to 20 MB and video up to 50 MB. Video transcription also serves content pages: when you do not upload WebVTT captions for a video, captions are generated from the AI transcript where possible. With transcript translation on, you can read the transcript in your own language too.

With dynamic content translation on, content pages, surveys and module text are translated into the language the visitor selects. The visitor interface supports Turkish, English, German and Arabic, and translation targets those four languages. You write one text and reach international patients, students or passengers.

In the other direction, comment and message translation lets admins read feedback written in another language in the language of their panel. Translations and transcripts are cached for 14 days, so the same content is not sent twice. The "Translations" view under Records lists the translations made.

The default provider is the ENA AI Gateway, which needs no setup. You can also bring your own key for OpenAI (text and audio), Google Gemini (text, audio and video) or Anthropic Claude (text only). One provider is active at a time. Your keys are stored encrypted with AES-256-GCM.

Usage has built-in daily limits: 20,000 calls a day for sentiment, 5,000 for translation and 5,000 for transcription. Every call is written to the "AI invocations" log with provider, model, latency, characters, tokens and status. Prompts and content are not written to that log.