Skip to main content
AboutProblemsGuidelinesTerms & conditions

KCC Agent Interface – AI Assisted Expert Support System

Problem Statement​

Kisan Call Center (KCC) experts handle thousands of farmer calls every day and must provide accurate agricultural advice in real time. During live conversations, experts largely depend on memory and personal experience, which can vary between agents. Even trained experts may find it difficult to recall the most accurate or up-to-date recommendations instantly. Currently, there is no real-time support system that assists experts during calls using verified agricultural knowledge. Incorrect or incomplete advice can directly affect farmer yield and income, highlighting the need for a real-time decision support mechanism that enhances expert responses without disrupting call flow.

Objective​

To provide real-time AI assistance to KCC experts during live farmer calls by converting speech to text, understanding agricultural intent, and suggesting expert-verified answers from a trusted knowledge base.

Work Done So Far​

  • Live audio capture from expert–farmer calls.

  • Real-time speech-to-text using Sarvam API (supports Malayalam, Hindi, and regional languages).

  • Continuous transcription with 15-second rolling analysis.

  • Semantic understanding of farmer queries (crop, symptom, issue, stage).

  • Similarity search over a Golden Knowledge Database of expert-verified Q&A.

  • Display of relevant, pre-approved answers on the expert dashboard.

  • Human-in-the-loop system where the expert decides the final response.

Key Features​

  • Real-time transcription during live calls.

  • Semantic similarity search (not keyword-based).

  • Only expert-verified answers shown.

  • Low-latency decision support for agents.

  • Improves accuracy, consistency, and confidence of expert responses.

Technologies Used​

  • Speech Recognition: Sarvam Speech-to-Text API

  • NLP: Embedding models for semantic understanding

  • Vector Database: Similarity search on Golden Q&A

  • Backend: AI processing engine (Python)

  • Frontend: Web-based expert dashboard

Future Scopes​

  • Regional language translation.

  • Crop-specific dashboards.

  • Auto-highlighting of diseases and nutrients.

  • Call quality scoring and expert performance analytics.

  • AI-flagging of risky or incorrect advice.

  • Predictive alerts for outbreaks and weather-linked risks.

  • Voice-based AI co-pilot for experts.

Impact​

This system transforms every KCC agent into a knowledge-augmented expert, reduces errors in agricultural advice, standardizes service quality, and improves outcomes for farmers at scale.