By selecting the best-fit platform for each use case, from migrating SAP workloads to AWS to utilizing AI-driven analytics on Azure, the strategy enhances flexibility, security, and cost-effectiveness. Sengupta emphasizes how this approach allows the organization to stay agile, innovate effectively, and ensure robust data protection across various environments.
Excerpts from an interview:
Multi-Cloud Mastery: Lessons from a CIO on Choosing the Right Cloud for the Job
What factors influenced your organization’s decision to adopt a hybrid or multi-cloud strategy over a single-cloud approach?
We're effectively leveraging all three major hyperscalers—Microsoft Azure, Amazon, and Google Cloud—for different use cases. Rather than applying a single cloud provider to a specific function, we adopt a multi-cloud strategy based on the unique strengths each platform brings to the table. For instance, when we transitioned one of our business units' SAP environment from on-premises to the cloud, we chose Amazon Web Services. Meanwhile, for other applications, such as AI-driven analytics for trade promotions and product marketing mix optimization, we have built a data lake on Microsoft Azure.
Additionally, we are working with Google Cloud Platform (GCP) for other specialized use cases. This multi-cloud approach allows us to leverage the unique capabilities of each hyperscaler. For example, we’re setting up a landing zone to run large language models (LLMs) for generative AI applications. Each hyperscaler offers distinct LLM capabilities—OpenAI’s models through Microsoft Azure, Amazon’s proprietary LLMs, and Google’s Gemini—each tailored to serve different generative AI applications.
By implementing landing zones across all three hyperscalers, we ensure we can flexibly choose the LLMs that best meet our specific requirements. This strategy avoids the limitations of a single cloud provider by focusing on the technical competencies and algorithmic capabilities each one offers, allowing us to select the best fit for our diverse internal use cases.
How do you determine which workloads are best suited for on-premises versus public or private cloud environments?
Typically, for any solution where transaction volumes are steady and performance requirements are consistent, it’s most efficient to keep it on-premises. Doing so, however, requires a company to invest in hardware, services, support, and robust security infrastructure. Fortunately, we already have a highly advanced data center with extensive technical capabilities, providing secure compute resources fully managed by our in-house team. This setup allows us to maximize our existing infrastructure for stable, predictable workloads.
For the cloud, we prioritize use cases where we need bursts of compute power—for instance, a particular query, time-sensitive analysis, or specific periods when elevated compute and analytical capabilities are required. This enables us to tap into high-performance resources when necessary, without maintaining them year-round. The cloud’s flexibility in scaling up during peak demand and scaling down afterward makes it ideal for such dynamic requirements.
What are the biggest challenges in managing data and applications across a multi-cloud environment, and how are you addressing them?
Managing costs is one of the biggest challenges, especially in the beginning. However, companies are becoming smarter in addressing this. The key is to continuously monitor cloud consumption and regularly assess the costs to ensure that only essential applications are migrated. A benefit analysis is crucial to determine whether it’s more cost-effective to maintain and service applications internally or move them to the cloud.
Additionally, understanding how to optimize cloud usage is vital. Since companies don't operate 24/7 but still pay for cloud services around the clock, it's important to optimize usage. This can involve reusing cloud capabilities during off-peak hours or reducing compute and storage needs when demand is low. Effective management of cloud costs requires thoughtful planning and control.
How does your cloud approach enhance the resilience and flexibility of your IT infrastructure?
The pandemic highlighted the importance of cloud adoption for business continuity. With cloud, risk is distributed across multiple locations, ensuring that servers, storage, and compute resources are not confined to a single area.
This distribution enhances resilience. Additionally, the cloud provides robust security frameworks, with global best practices followed by hyperscalers to protect transactional data, master data, and other sensitive information. Cloud also enables remote work, allowing employees to be flexible in where and how they work. It shifts business processes from being dependent on physical presence to being aligned with outcomes, fostering greater agility.
How do you ensure data security when utilizing various LLMs across multiple hyperscalers?
To protect our data, we set up private instances of each hyperscaler’s LLMs in our own data center. This way, we can leverage their AI capabilities without transferring sensitive company data to their platforms. By using a landing zone within our secure environment, we avoid direct data sharing. Although this setup may mean working with slightly older versions of certain LLMs, it ensures data integrity and security.
Managing high computational requirements for LLMs can be costly. How do you address this?
We manage compute costs by applying prompt engineering to streamline our inputs, reducing the complexity and size of queries before sending them to LLMs. This allows us to limit the data volume processed, minimizing compute costs. By sharing only selected data with LLMs, we optimize both cost and performance.
How do you approach the risk of vendor lock-in while balancing cost-effectiveness and agility?
To avoid vendor lock-in, we distribute workloads across different hyperscalers based on use cases, maintaining flexibility and negotiating favorable terms. This approach ensures we receive competitive pricing and the necessary technical support. Additionally, we assess long-term partnerships to foster collaboration with the best provider for each solution.
Gartner recently forecasted a 17-18% increase in IT spending by 2025. How does your organization’s budget align with this trend?
Our IT budget growth aligns more with transformative initiatives than with basic infrastructure. Investments focus on improving productivity, business innovation, and security. We prioritize transformative solutions that drive tangible business outcomes, particularly in new areas like AI and enhanced cybersecurity.
Many AI-driven projects stall at the POC stage. What are your thoughts on advancing these initiatives?
The key to transitioning from POC to production lies in demonstrating clear productivity gains and business benefits. While some generative AI solutions are readily deployable, others require refinement. We aim to adopt AI solutions that have proven value and impact rather than adopting technology for its own sake.
Looking ahead to 2025, what innovations in cloud technology would be most beneficial to your organization?
To fully harness generative AI’s potential, we need solutions that ensure intellectual property protection and compliance, especially as data privacy regulations evolve. Cloud providers play a crucial role in securing AI-generated content, protecting IP rights, and maintaining authenticity. Additionally, advancements in handling and processing large datasets efficiently will be crucial for our continued cloud adoption.
What do you like most about your job as a CIO?
What I enjoy most about being a CIO is staying close to emerging technologies. These innovations are driving revenue and shaping the future of the business. However, it's also challenging to find the right talent. While there are many people in the tech field, it’s rare to find those with the right experience—individuals who can think critically, consider all perspectives, and implement solutions effectively. It’s not just about technical skills; it's about the ability to approach problems strategically and holistically.