Q&A with Sanjay Deshmukh, Senior Regional Vice President, ASEAN and India, Snowflake.
What trends do you see driving data and cloud in ASEAN and Singapore?
Enterprise Artificial intelligence (AI) is a key trend driving the need for AI data cloud platforms. Generative AI is revolutionising business operations by enabling enterprises to create and deploy innovative solutions rapidly. However, the effectiveness of these AI models is fundamentally dependent on the quality and management of the data they use.
A strong data foundation is crucial for ensuring that data is clean, reliable, accessible, and governable. This foundation enables organisations to break down data silos and ensure seamless data flow across departments and applications. In ASEAN and Singapore, companies are increasingly recognising the importance of this approach as they strive to make accurate predictions and informed decisions. With up to 90% of data – such as videos, images, and documents – being unstructured, there is a significant opportunity to unlock value from these underutilised resources.
Data monetisation is another key trend in this evolving landscape. By leveraging platforms like Snowflake, organisations can transform their data into strategic assets. This involves not just internal optimisation but also creating new revenue streams through data products and services. Data collaboration across enterprises allows for the development of richer datasets, providing deeper insights and fostering innovation. Monetising data effectively requires a robust infrastructure that ensures data security, compliance, and governance, all while enabling seamless sharing and utilisation.
At the same time, ensuring data is protected for secure use is increasingly a priority for decision-makers. A recent Snowflake survey found that more organisations are using new features to tag and classify data so that appropriate access and use policies can be applied.
What are the challenges faced by organisations when it comes to leveraging these technologies?
There are four key challenges: rising complexity, increasing costs, how to build trust and a lack of skills.
The first challenge is the complexity of building, managing, owning, and supporting the infrastructure needed to adopt these technologies. For example, an organisation may not have the capabilities to support and host the models, the GPU’s, the vector databases, and all the other components required to build an AI application. Snowflake’s core design principle has been to simplify which enables wider adoption. Snowflake offers a fully managed AI platform with all the required components which helps the customers to eliminate complexity and they don’t have to invest time in stitching together various tools or managing the underlying infrastructure.
This also allows developers to leverage near-unlimited scalability and concurrency and streamlined data pipelines without the burden of site reliability engineering (SRE)/DevOps to launch new features faster with improved engineering efficiency.
This platform is offered as pay-per-use where you only pay for the actual consumption and it also comes with built-in cost management and performance optimisation. This helps shield organisations from unpredictable demand fluctuations associated with many AI platforms that result in cost spikes – meeting the second challenge of increasing costs.
The third challenge hinges on building trust – specifically by ensuring that customer data is not compromised, and the organisation data is not going to go to training an external model. Snowflake solves this problem by bringing the models and the AI applications to the data and not the other way around. This ensures data stays within the confines of the organisation’s security perimeter and that all the security policies and controls applied to the data are protected.
Finally, there is the challenge of lack of relevant skills. A typical business may have business analysts and data scientists, but they are not likely to have the skills to build and manage AI platforms. The SnowflakeAI Data Cloud enables data analysts, data scientists, and data practitioners to build AI applications and help organisations improve productivity and efficiency without having to learn and build complex AI platforms or stitch together multiple complex Data and AI tools.
How can organisations effectively implement Gen AI and what needs to be in place before organisations leverage on Gen AI?
The first step is to create a data foundation that offers a single-pane view of structured, semi-structured, and unstructured data. A lot of the value provided by large language models (LLMs) comes from unstructured data, be it text or images. The Snowflake platform enables this by anchoring the data foundation on unified data and universal governance. This eliminates silos to enable any architectural pattern, while supporting all data types and workloads.
Before deploying any model, it is crucial to identify what information needs protecting and how to protect it – whether through masking, tagging, anonymisation, or other processes. The next step is to implement a governance layer around features such as low-level security and access control.
Organisations must be clear about what business outcomes they are expecting from their generative AI applications. This provides the foundation to determine the data required: structured/unstructured and if this data is available internally or needs to be sourced externally. If the data is available internally they may need to build additional pipelines to bring the data into the enterprise data platform or get it from an external source or from the Snowflake marketplace from one of the data providers.
External data inputs can help organisations get a 360 view of the business and help build actionable insights. For example: if a retailer wants to expand its presence and open new outlets, what more can it do besides turning to the internal data sources and performance of its current stores? The business could look at footfall traffic from telecommunications operators to parse data that covers age groups too and map this data using geospatial coordinates to determine the location of their new store. Training AI/ML models with these comprehensive datasets helps provide better business context, improves effectiveness and delivers differentiated business outcomes.
How has Snowflake been helping local organisations to unlock the value of data?
Snowflake’s unified platform streamlines complexity and reduces costs. Canva, the popular graphic design platform with over 150 million monthly active users worldwide, was able to build better AI products faster, helping more users design with ease. Canva also leveraged Snowflake to bolster organisation-wide decision-making – including improving feature launches, optimising marketing spending, and helping executives build more informed strategies.
Meanwhile, global communications powerhouse Zoom uses Snowflake’s AI Data Cloud to support internal business functions and develop customer insights. By combining Snowflake Cortex AI and Streamlit, Zoom can utilise pre-trained LLMs to speed up app building. Zoom’s teams were then empowered with easy and swift access to helpful answers, facilitating AI democratisation without compromising on data security and governance.
Another example is Spark New Zealand. Using Snowflake to integrate AI and ML, one of New Zealand’s largest telecommunications and digital services providers aligned key processes across its organisation, from workforce management to supply chain. Their proprietary platform, BRAIN, utilises ML models to deliver precise messaging to customers and has improved the performance of marketing campaigns by 20 times. This allows Spark to predict customer needs with higher accuracy, driving greater ROI and operational efficiency.
Apart from that, solutions like Snowflake Horizon make discovery and access to critical data assets more seamless, while maintaining privacy compliance. Innovations like the Cortex suite, meanwhile, ensure access to data is paired with no-code interfaces that simplify AI development while upholding responsible AI.
