Skip to content

Navigating the Challenges of Scaling Generative AI: Insights from Vijoy Pandey, SVP of Outshift by Cisco

In an insightful discussion about generative AI (gen AI), Vijoy Pandey, SVP of Outshift by Cisco, highlights the complexity, data-related issues, and the need for a software-centric approach in successfully deploying and managing gen AI at scale.

Enterprises are rapidly embracing generative AI (gen AI) for diverse applications, from ideation to decision optimization. However, as companies venture into scaling gen AI initiatives, they face numerous challenges, according to Vijoy Pandey, SVP of Outshift by Cisco. Pandey highlights the complexity, data-related issues, and the need for a software-centric approach in successfully deploying and managing gen AI at scale.

Key Insights:

  1. Rapid Evolution and Managing Complexity:
    • Gen AI is considered one of the most significant technological transitions in recent years, impacting creativity, productivity, innovation, and revenue streams.
    • The rapid evolution of gen AI introduces challenges in managing its complexity, making it difficult for organizations to keep up with the diverse and multi-pronged developments in the space.
  2. Skills Gap and Transformative Models:
    • The lack of skill sets becomes a barrier to the successful adoption of transformative large language models.
    • Transformative models, such as those used in gen AI, require specific expertise for deployment, updates, and management.
  3. Challenges in Data Management:
    • Data presents a major challenge, with organizations often having disparate data puddles rather than a centralized data lake.
    • Quality issues arise in cleaning and utilizing data for gen AI, with security, responsible AI, and usability concerns complicating the process.
  4. Privacy Considerations:
    • Privacy becomes a critical issue as gen AI models learn from user interactions, raising concerns about sensitive data and responsible AI.
    • A responsible AI approach is essential in addressing privacy, eliminating bias, and ensuring explainability for customers.
  5. Model Selection and Complexity:
    • The proliferation of model options complicates decision-making, requiring organizations to choose, customize, and fine-tune models for optimal performance.
    • The complexity of the gen AI space makes it challenging to keep up with evolving models and boundary conditions.
  6. Cost and Compute Power:
    • Gen AI is associated with high costs, both in terms of initial setup and ongoing expenses related to user requests and compute power.
    • The significant compute power required for generative AI solutions contributes to the overall cost of implementation.
  7. Software-Centric Approach:
    • Successful deployment and management of gen AI projects require a software-centric approach, focusing on building or utilizing software frameworks.
    • Abstraction layers and software frameworks help standardize prompts, manage bias, improve fairness, and simplify interactions with multiple providers.
  8. Narrowing Focus for Success:
    • Organizations must narrow their focus by selecting specific use cases that align with their strengths and value propositions.
    • Success in gen AI involves excelling in chosen use cases, building abstraction layers, and collaborating with experts to simplify the complexity of the space.

As gen AI technology continues to evolve, Pandey emphasizes that it will become simpler, cost-effective, accurate, trustworthy, and inclusive. He encourages companies to embrace AI-focused digital transformation, emphasizing the importance of crystal-clear KPIs to ensure a significant return on investment. Pandey concludes by highlighting the transformative potential of gen AI in reshaping productivity, creativity, and innovation.