A leading global manufacturing enterprise operating across more than 100 sites in over 25 countries faced a growing challenge - systematically gathering and consolidating stakeholder inputs for an ambitious multi-industry Experience Center. With a pressing need for a scalable process to capture and consolidate feedback from diverse virtual consultations, they turned to Coforge.
The Challenge
Building an Experience Center spanning healthcare, packaging, technology, automotive, and appliances is no small feat. Each sector brought unique requirements and distinct stakeholder expectations, making coordination both critical and complex.
Scheduling and conducting virtual consultations with clients, internal teams, and industry experts across time zones was highly resource-intensive. Nuanced feedback and ideas shared during video-based discussions were often lost, unstructured, or scattered across multiple platforms. Manually reviewing hours of video recordings to extract suggestions, ideas, and feedback was tedious and error-prone.
Without a systematic process to consolidate inputs, the risk of misalignment between stakeholder expectations and loss of valuable insights increased significantly. Delays in processing feedback threatened to slow the overall timeline for the Experience Center launch. They needed a scalable, intelligent solution that could automate the extraction of insights from video consultations, freeing teams to focus on analysis and action rather than manual transcription and collation.
Our Approach
Coforge designed and implemented an end-to-end AI-powered Video Intelligence Platform that combines the strengths of AWS AI services with the contextual understanding of Anthropic's Claude large language model. The solution transforms raw stakeholder video recordings into structured, searchable, and actionable intelligence automatically.
Automated Speech-to-Text Transcription
Amazon Transcribe converts speech from stakeholder video discussions into text. The service supports multiple languages and provides automatic punctuation, speaker identification, and custom vocabulary to enhance accuracy.
Contextual Speech and Sentiment Analysis
Amazon Comprehend analyzes transcripts to identify key phrases, sentiment, entities, and language, providing deeper insights into video content.
AI-Driven Feedback Extraction and Top Modeling
Anthropic's Claude LLM processes transcripts to extract stakeholder suggestions and feedback related to the Experience Center, and further identifies and extrapolates the main topics of conversation at different stages of each video, helping categorize and summarize discussion content effectively.
Semantic Vector Embedding and Intelligent Search
The Amazon Titan Embeddings Model converts processed text into high-dimensional vectors that capture semantic meaning. These embeddings are stored in Amazon OpenSearch, which acts as a knowledge base for efficient retrieval and search of discussion content.
.jpeg?width=1600&height=897&name=bottom-image%20(3).jpeg)
Impact to Date
50%
300+
Business Impact
Since deploying the platform, the organization has eliminated the need to manually traverse large volumes of recordings to extract discussion points. Key outcomes include:
-
All suggestions, feedback, and actions across every aspect of the Experience Center are now automatically extracted and presented through a custom Angular-based interface, improving alignment between stakeholder inputs and project outcomes.
-
The solution's ability to handle any type of video content makes it operationally scalable as consultation volumes grow.
- Configurable extraction ensures outputs remain tailored to each industry context. Teams can now focus on analysis and informed decision-making rather than manual transcription and collation.
By combining deep domain knowledge with AI-led engineering, Coforge delivered a production-grade platform that turned a fragmented, manual process into a scalable intelligence system, purpose-built for the complexity of a global manufacturing enterprise.
.jpeg?width=1920&height=1080&name=hero-image%20(5).jpeg)