AWS re:Invent 2025 - From Vision to Value: Scaling Gen AI with Speed to Reduce TCO with AWS (COP215)

Scaling Gen AI with Speed to Reduce TCO: TechMahindra and AWS Partnership

Challenges in Generative AI Adoption

• Only 5% of generative AI proofs-of-concept (POCs) get converted to production
• Companies spend 25% of their time finding the right data and people to support daily work
• 7% of company budgets go towards improving employee productivity with co-pilots and other tools
• 1/3 of technology investments (e.g., GPUs, software) are not being utilized

TechMahindra and AWS Approach to Productionize Generative AI

1. Treat POCs as Full-Fledged Projects: Define clear deliverables, data requirements, and performance metrics.
2. Embed Security and Governance: Incorporate security and governance considerations from the POC stage.
3. Plan for Iterative Improvements: Adopt a staged approach, incorporating learnings from each phase.
4. Balance Human and AI Roles: Leverage AI to automate tasks, while maintaining human oversight and involvement.
5. Measure and Evaluate Continuously: Benchmark POCs and projects to track progress and outcomes.

TechMahindra's Orion Platform

• Orion is a platform developed by TechMahindra, in partnership with NVIDIA, to deliver AI solutions.
• Key capabilities:
  • Access to pre-trained models for fine-tuning
  • Drag-and-drop repositories for rapid deployment
  • Automated agent creation to streamline task automation
• Benefits:
  • Speed: Faster development and deployment of AI solutions
  • Interoperability: Seamless integration with existing systems
  • Security and Governance: Built-in security and governance controls

Use Cases: Enhancing Customer Experience and Fraud Detection

1. Engineering Company Case Study:

   • Challenge: High call volumes and long wait times for customer support
   • Solution: Deployed a generative AI-powered chatbot using Amazon Lex, AWS Lambda, and Anthropic's ClauseSet model
   • Outcome: Automated 50% of customer calls, enabling support agents to focus on sales and generate $50,000 in additional revenue

2. BFSI Company Case Study:

   • Challenge: Slow resolution of account takeover (ATO) fraud cases, leading to credibility loss
   • Solution: Automated the fraud detection process using AWS Step Functions and created a generative AI-powered chatbot for fraud analysts
   • Outcome: Reduced ATO case resolution time from 7 days to less than 24 hours, improving customer satisfaction by 10%

Key Considerations for Successful Generative AI Adoption

1. Focus on Business Outcomes: Align AI solutions with the customer's specific business needs and challenges.
2. Optimize for Cost: Consider the cost implications of token-based models and other AI components.
3. Adopt a Production-Ready Mindset: Treat POCs as full-fledged projects with clear deliverables and performance metrics.
4. Measure and Evaluate Continuously: Continuously track the performance and impact of AI solutions.

Beyond Generative AI: Exploring Metaverse and Quantum Computing

• TechMahindra is also exploring the integration of AI with immersive experiences, such as digital twins, and the application of quantum computing for use cases like fraud detection and route optimization.
• These efforts are being developed in collaboration with AWS, leveraging products like AWS Bracket for quantum computing.