Delivered governed, self-service analytics to over 5,000 users across the organization
Achieved near 100% accuracy in AI-generated insights by grounding models in Looker's semantic layer
Automated secure credential management to meet strict InfoSec and DLP compliance standards
Accelerated dashboard creation for developers, reducing build time from hours to minutes
PayPal adopted Looker's Model Context Protocol (MCP) Server to provide secure, conversational analytics to its workforce, overcoming complex security challenges by automating credential encryption. By grounding AI interactions in Looker's governed data models, PayPal ensures accuracy and compliance while enabling 5,000 employees to query petabytes of data using natural language.
PayPal adopted Looker's Model Context Protocol (MCP) Server to provide secure, conversational analytics to its workforce, overcoming complex security challenges by automating credential encryption. By grounding AI interactions in Looker's governed data models, PayPal ensures accuracy and compliance while enabling 5,000 employees to query petabytes of data using natural language.
Democratizing data at scale with
uncompromising security
The real win here was delivering conversational analytics to 5,000 employees, fully secured and governed, by extending what we already had. No migrations, no rewrites, just smart architecture on top of Looker's semantic layer. That's how PayPal innovates at scale.
Vaishali Walia
Senior Director of Engineering, PayPal
As a global leader in commerce, PayPal manages a data estate of immense magnitude and complexity. Their operations involve querying billions to trillions of records across petabytes of data stored in BigQuery.
As PayPal scaled, it encountered a classic business intelligence challenge which required analysts to pull full table extracts before any analysis could begin—not because live querying was impossible, but because those tools were architecturally optimized for extract-based Looker, with its query engine generating optimized SQL on the fly, pushing computation directly to the database and delivering faster, more efficient results on live data. Apart from performance, there was also a governance challenge.
Without a shared semantic layer, foundational metrics like "revenue" or "active users" risk being calculated differently across teams, creating fragmented logic and eroding trust in the numbers. PayPal adopted Looker to address both—bringing live query performance and a single, governed source of truth under one platform.
However, as the company moved to integrate generative AI deeper into their business, a new, critical challenge emerged. PayPal's leadership envisioned a future of Conversational Analytics, where employees could simply "chat" with their data using natural language to uncover insights instantly. The goal was to deploy a "Cloud Desktop" experience, powered by Large Language Models (LLMs), that would allow five thousand users, from data scientists to Customer Success Managers, to query data without writing a single line of SQL.
One remaining hurdle, a foundational requirement for the world of finance, was ensuring security. Standard implementations of conversational AI often require storing API keys, client IDs, and secrets in plain text on user machines, a practice strictly prohibited by PayPal's rigorous security standards. Additionally, opening a direct line between an LLM and the company's data warehouse raised significant data loss prevention concerns. PayPal needed a way to expose the intelligence of the Looker semantic layer to an AI interface without compromising on security or accuracy. They needed a solution that could translate natural language into precise, governed metrics while keeping authentication credentials invisible and encrypted - a capability that did not exist out of the box.
The solution: A secure, automated path to conversational intelligence
To realize their vision of secure conversational analytics, PayPal's engineering team turned to Looker's Model Context Protocol (MCP) Server. MCP acts as a standardized bridge enabling Large Language Models to read and understand the context of Looker's semantic models. To meet their strict enterprise security requirements, PayPal built on top of this open-source foundation.
The team developed an internal architecture that completely automated the security handshake, abstracting it away from the end-user, and deployed the Looker MCP Server directly into PayPal's MCP Hub—sitting behind the AI proxy layer for full enterprise security, end-to-end auditability, and real-time observability.
By connecting PayPal's enterprise Looker BI MCP with Claude, we turned complex data analysis into effortless conversation—accelerating insights by more than 50% and enabling teams to make smarter, faster business decisions with confidence and security.
Vinod Ganesan
Director, Analytics Platform, PayPal
This "wrapper" performs a sophisticated sequence of operations behind the scenes. When a user initializes the tool, the script executes a secure API call to Looker to fetch the necessary credentials. Crucially, instead of storing these credentials in a config file or plain text, the system pushes them directly into a secured vault/database with encryption, while access tokens are encrypted and stored in Redis cache—keeping sensitive data protected at every layer of the stack. This ensures that sensitive Client IDs and secrets remain fully encrypted at rest and inaccessible to the user, satisfying InfoSec's most stringent requirements.
For the Engineering organization at PayPal, this complex backend engineering results in a seamless "one-click" experience. A user simply installs the Looker MCP Server from the internal software center, and the environment is instantly configured. They can then open their AI-powered Claude Desktop and begin asking questions like, "What was the Total Payment Volume (TPV) for the UK market last quarter?"
The true power of this solution lies in Looker's semantic layer. Unlike generic AI to SQL tools that often hallucinate or generate incorrect queries by guessing at column names, PayPal's MCP implementation grounds every AI response in governed data definitions. When the LLM receives a prompt, it doesn't query the raw database; it queries the Looker model. This ensures that the AI respects the exact business logic defined by data owners, resulting in near 100% accuracy for generated insights.
PayPal Senior Software Engineer Venkata Sangaraju said, "In the world of financial services, we can't afford hallucinations. By anchoring conversational analytics to Looker's governed LookML, we transformed AI from a novelty into a trusted tool that delivers almost 100% accuracy against our defined metrics."
Beyond ad hoc queries, PayPal is using this infrastructure to revolutionize developer workflows. In early pilots, the team has demonstrated that the time required to build complex dashboards can be compressed from hours to minutes. A developer can describe the visualization they need in natural language, and the MCP-enabled AI generates the corresponding LookML and visualization code instantly. By securing the "front door" to these capabilities with their custom MCP implementation, PayPal has successfully democratized data access, proving that with the right architecture, the agility of AI and the rigor of enterprise governance can coexist.
For over 25 years, PayPal has revolutionized global commerce with innovative experiences that make shopping, selling, and sending money simple, personalized, and secure. PayPal empowers consumers and businesses in nearly 200 markets to thrive in the global economy.
Industry: Financial Services
Location: United States
Products: Google Cloud, Looker, BigQuery
