Font Vella

A listening system assisted by AI to turn questions, habits and consumer contexts into more relevant editorial decisions

Font Vella coexists with an everyday category, transversal and full of contexts: well-being, sport, family, sustainability, heat, restoration and active life. The challenge was to go from a mention-centric monitoring to a listening capable of ordering signals and explaining why some deserved a response, a piece or an investigation.

Karmina AI Studio implemented a Social Listener connected to approved sources, an editorial framework and a human review circuit. The AI grouped conversations and detected patterns; the team interpreted the context, checked the evidence, and decided what to enter the strategy.

Main CTA: Let’s talk about the listening system. CTA Secondary: Discover the Social Listener

Summary Fact Sheet

• Brand: Font Vella
• Sector: Food and Beverages
• Challenge: convert scattered conversations into verifiable editorial opportunities
• Principal Agent: Social Listener
• Complementary agents: Market Scout · Social Publisher
• Service: Karmina Always-On Social AI
• AI layer: Operational agent for listening, classification and detection of signals
• Human role: Interpret, prioritize, validate and decide on any public response

El challenge

Be present in the conversation without chasing every trend

A mass consumer brand can accumulate many mentions and, at the same time, continue to have little useful information to decide. The volume mixes real doubts, weather content, consumption habits, comparisons, incidents, jokes, modes and conversations that only contain a word related to the category.

The challenge was not to listen anymore. It was to separate the noise from the signals that could help to better understand the moments of consumption, the recurring questions and the territories where the brand had authority to participate.

To listen well is not to react before anyone else. It's understanding what's going on before you decide if the mark needs to come in.

The starting point

Information could come from social platforms, search queries, community messages, campaign reports and team knowledge. Without a shared criterion, each source generated a different list and decisions depended too much on who reviewed the data that day.

A common taxonomy was defined: theme, origin, time of consumption, public, territory, risk, degree of verification and possible use. This allowed the same conversation not to be interpreted as an editorial opportunity, an incident or a trend according to the document where it appeared.

The Karmina AI Studio approach

The service combined three layers:

1. Listen to the government: approved fonts and keywords, with exclusions to reduce false positives.
2. Assisted Classification: grouping of conversations, detection of repetitions and summary with original examples.
3. Human Editorial Committee: context review, data contrast and decision on response, content or discard.

The agent did not calculate an emotional truth from a sentiment indicator. It showed evidence, uncertainty and changes so the team could interpret them.

How the Social Listener worked

1. Collect content from authorized sources.
2. It removed duplicates and classified the signals according to taxonomy.
3. It detected volume changes or recurring questions.
4. He prepared a chart with examples, links, date and degree of confidence.
5. He proposed three possible exits: observe, investigate or activate.
6. It transferred only the validated signals to the Social Publisher and the calendar.