Introduction
Cekura - Key Information Extraction
Cekura is a tool designed to automate quality assurance, testing, and observability for Webex AI Agents and, more broadly, voice AI agents. It focuses on ensuring the reliability and performance of conversational AI deployments.
1. Problem Solved:
Cekura addresses the need for reliable and consistent performance of voice AI agents. Specifically, it helps mitigate issues such as:
- An impatient, interruptive user causing disruptions.
- An old conversation continuously causing issues.
- Skipping compliance checks.
2. Capabilities and Features:
- Automated Testing: Cekura allows for automated testing against AI-generated and custom datasets, utilizing workflows, personas, and real audio data to create datasets for evaluation.
- Simulate Scenarios: The tool allows for simulating various user interactions, including scenarios with diverse personalities and interruptions.
- Start Call: Greet Customer: The system provides a function to initiate interactions with simulated customers.
- Replay Real Conversations: Users can replay existing conversations for testing and analysis.
- Evaluations: Cekura offers a mechanism for assessing agent performance based on defined criteria.
- Real-time Monitoring: Provides real-time insights, detailed logs, and trend analysis for optimal performance.
- Alerting: Instant notifications for errors, failures, and performance drops to ensure swift action.
- Intuitive Dashboard: Offers a performance visualization and data-driven decision-making for continuous improvement.
3. Key Aspects & Partnerships:
- Trusted by 50+ Conversational AI companies: Cekura is utilized across various industries.
- Cisco Partnership: Cekura collaborates with Cisco to provide end-to-end testing and observability solutions, particularly for Webex AI Agents, emphasizing reliable AI agent deployment at scale.
- Recent Funding: Cekura recently raised $2.4M in seed funding to further develop its reliability layer for conversational AI.
4. How it Works:
Cekura's functionality centers around creating datasets using workflows, personas, and real audio. These datasets are then used to evaluate agent performance, tracking metrics such as empathy, responsiveness, and the absence of hallucinations (incorrect information). The system’s dashboards provide real-time insights, enabling proactive monitoring and immediate alerts for any performance deviations.
