Governance should empower, not restrict: Prasanna Krishnan, Snowflake
When enterprises embrace AI, they inevitably face a paradox of how to innovate at speed while staying compliant and trustworthy. For Snowflake, the answer somewhere lies in making governance not an afterthought but an intrinsic part of the data journey.
“We want to ensure that everything you do in Snowflake, whether it’s AI or analytics, is trusted and governed,” says Prasanna Krishnan, Head of Apps, Collaboration, and Horizon at Snowflake. “Our value proposition is easy, connected, and trusted, and the ‘and’ matters as much as each word.”
Krishnan leads the product teams driving Snowflake’s applications, collaboration ecosystem, and Horizon, the governance and discovery layer that underpins the AI Data Cloud. Together, they reflect the company’s core belief: that the future of enterprise AI depends on seamless collaboration and embedded governance.
Embedding governance into the platform, without lock-in
While many technology vendors advocate for platform-agnostic governance frameworks, Snowflake takes a more integrated stance. Horizon is built directly into the Snowflake platform, ensuring every action, from AI queries to data sharing, adheres to role-based access controls and policies.
Yet, Krishnan is quick to clarify that integration doesn’t mean restriction. “Horizon’s governance isn’t limited to Snowflake data. We also support open table formats like Apache Iceberg,” she explains. “You can use Snowflake’s governance capabilities with open data, or even run your own Apache Polaris catalogue if you want to go fully vendor-agnostic.”
This dual approach, deep integration with openness, helps customers strike the delicate balance between control and flexibility.
Balancing speed and trust
In today’s data-driven businesses, speed is everything. But speed without governance can lead to chaos. Snowflake’s approach, Krishnan emphasises, is to make governance simple enough to accelerate innovation, not hinder it.
“Our role-based access control has always been powerful but also simple, it’s just a few lines of SQL,” she says. “Now, we’re taking it further with Horizon Copilot, which lets data stewards ask questions in natural language, like ‘Who has access to this table?’ or ‘Which tables contain sensitive data?’”
By combining automation with ease of use, Horizon turns governance into what Krishnan calls “a tailwind,” a driver for democratising access to data and AI, rather than a bureaucratic bottleneck.
The context behind AI accuracy
AI outputs are only as reliable as the context surrounding the data. Horizon’s semantic views address this by layering business context over physical data schemas. “A table might store customer data, but how a business defines a customer can vary,” says Krishnan. “Semantic views help AI models understand that context, making text-to-SQL conversions and insights more accurate.”
Horizon also integrates retrieval-augmented generation (RAG) techniques, grounding AI responses in actual, governed data while ensuring role-based access controls remain intact. In Krishnan’s words: “If I’m a marketing analyst, I’ll only see the data my role allows, nothing more.”
Redefining collaboration: From data to AI products
Snowflake has long been known for transforming data sharing by eliminating duplication and delays. But the company’s ambitions have expanded far beyond that.
“What started as sharing structured data has evolved into sharing rich application logic, and now, AI products,” says Krishnan. “We’re seeing customers move from sharing raw data to sharing insights, even packaging semantic models and agentic AI applications as data products.”
This evolution, powered by Snowflake’s Native App Framework, allows enterprises to distribute secure, sandboxed applications through the Snowflake Marketplace. Each app specifies exactly what privileges it requires, similar to how mobile apps request permissions before installation.
“Customers love this transparency,” Krishnan notes. “They can see exactly what the app can do in their account before deploying it. That’s how we ensure trust and control coexist.”
Ensuring trust and auditability in AI search
With the integration of Neeva’s search technology, Snowflake’s AI-powered discovery capabilities have grown more sophisticated, but trust remains the cornerstone.
“When you search in Snowflake, you only discover objects you’re authorised to see,” Krishnan explains. “Our ranking takes into account multiple trust signals, like how often a table is queried, to ensure results are both relevant and reliable.”
For regulated industries like banking and healthcare, where transparency and auditability are paramount, this blend of precision and governance is non-negotiable.
Adapting to India’s DPDP Act and a localised future
As India enforces the Digital Personal Data Protection (DPDP) Act, Snowflake is ensuring that its architecture aligns with the country’s privacy and localisation mandates. “Customers can choose to create an account in our India region, on AWS or Azure, so data never leaves the country,” Krishnan says.
Features like data clean rooms further support compliant collaboration, allowing organisations to share aggregate insights without exposing underlying data.
Beyond compliance, Krishnan sees vast potential in India’s data ecosystem. “India is already a data-strong nation, from Aadhaar to UPI,” he says. “The future of collaboration here will be about business-to-business ecosystems grounded in governance and powered by AI-ready data.”
In a world where enterprises are racing to integrate AI, Snowflake’s vision kind of offers a timely reminder that governance is not a barrier, it’s the enabler of responsible innovation.
As Prasanna Krishnan puts it, “Our goal is to make governance so powerful yet easy that it becomes the foundation for democratising access to data and AI.”
