Zapier is a no-code automation platform that enables millions of users and businesses to connect apps, streamline workflows and operate more efficiently. As AI raised expectations for speed and personalization, Zapier’s legacy data stack created bottlenecks. Siloed systems and heavy reliance on one team limited what could be explored or built. After adopting the Databricks Platform, Zapier unified its data and unlocked self-serve, AI-powered analytics across the business. What was once “not worth it” due to cost and time is now routine, from dynamic product personalization to conversational data access.
What data access challenges did Zapier face with its legacy data stack?
Zapier democratizes automation so that anyone, not just developers, can streamline work and save time. The company is focused on enabling teams across the business to work more independently with data, while also making the product itself more intelligent, responsive and personalized.
Internally, Zapier uses AI agents within Zapier Central to remove long-standing workflow bottlenecks and expand access to insights across every function. Sales, marketing, support and product teams are analyzing customer feedback, campaign performance and product usage patterns directly, often through knowledge retrieval copilots that reduce the need for deep technical expertise. Externally, Zapier is improving the customer experience, turning product data into a personalization layer that helps customers build automated workflows (“Zaps”) faster and with less friction.
Before adopting the Databricks Platform, these use cases were either difficult to achieve or not worth pursuing.
High latency
Zapier’s legacy data warehouse, AWS Redshift, limited its ability to handle low-latency data, which meant real-time personalization and responsive product experiences required entirely separate systems. That introduced complexities like duplicated pipelines and fragmented sources of truth.
Innovation constraints
Even when alternative technologies could address specific gaps, the engineering effort required to integrate and maintain them made experimentation costly and slow. Many potential use cases were never attempted because the return did not justify the investment.
Data silos
Business teams relied on the insights team to model data, design experiments and generate reports. In many cases, teams either waited for support or moved forward without data. As Lukas Toma, Director of Data and Machine Learning at Zapier, explained, “There were so many questions teams either didn’t have the skills to answer or knew wouldn’t get prioritized by the data team, so they just never got explored.” This slowed feedback loops across product, marketing and sales, making it harder to act on customer signals in real time and support more sophisticated go-to-market strategies.
Zapier needed a data foundation that could support both internal self-service and external product innovation, without the tradeoffs in speed, cost and complexity that defined its previous stack.
Why did Zapier choose the Databricks Platform?
Zapier adopted the Databricks Platform to unify its data and make information more accessible across the organization. Databricks allows teams to experiment more freely and deliver new capabilities faster. Here’s how.
Data storage, processing and access are conducted in one place, lowering latency and eliminating the need for multiple pipelines. According to Lukas, data processing is flexible enough to support batch and real time as needed. “We process data once and can decide how quickly it needs to be available,” he said. “If a use case requires data within seconds, we can make that happen without building separate systems.”
How did Databricks enable real-time analytics and AI-driven workflows?
Through the MCP connector, AI agents receive governed, real-time access to Zapier’s own data. Built on top of Unity Catalog, this ensures every query an agent makes follows the same fine-grained permissions and access controls as human users. This enables employees to query, analyze and act on information without relying on prebuilt dashboards or data engineering support. These agents allow customer-facing teams to explore customer trends, quickly optimize campaign performance and surface deal blockers and expansion signals in real time.
Vector Search allows Zapier to index and retrieve both structured and unstructured data at scale, handling over 600k requests per day. “With Vector Search, we now have a single pipeline to ingest and index the data, making semantic search easier and more accessible across the company,” added Lukas. This powers internal RAG-based knowledge retrieval systems, storing embeddings (e.g., help documentation, customer data from third-party applications like Zendesk) to provide richer context.
MLflow and Model Serving operationalize their machine learning workflows. MLflow provides a centralized system for tracking experiments, managing model versions and monitoring performance. Model Serving enables teams to deploy models into production with consistent, reliable access. This includes models that support personalization and analytics use cases.
By keeping model development, deployment and data processing within the Databricks Platform, Zapier reduces complexity and ensures that models are tightly integrated with the data they depend on.
What measurable results did Zapier achieve with Databricks?
At a glance, Zapier successfully:
Reduced dashboard creation time from 2–3 days to a few hours
Achieved up to 10x faster turnaround for analytics workflows
Enabled self-service analytics across product, marketing, sales, and support teams
Shifted from 0% to ~75% of queries executed via AI-assisted workflows within 6–9 months
Rather than measuring success purely in efficiency gains, the bigger shift is that teams can now pursue use cases that were previously too complex, too costly or simply out of reach. Databricks supports the real-time decisions that accelerate revenue growth and improve customer experiences.
One of the clearest changes is the move from bottlenecked insights to self-service analytics at scale. Previously, there was a limit to how many questions could be explored and how quickly teams could act. Today, employees across product, marketing, sales and support can independently query data, generate analyses and build dashboards using AI-powered workflows. This has not only removed delays but also expanded the scope of what gets analyzed.
Over the past six to nine months, the volume of queries run against Zapier’s data platform has grown significantly, with approximately three-quarters now originating from their MCP. “The number was zero last summer,” said Lukas. Business users interact with data conversationally and generate insights in real time.
This transformation has also led to substantial gains in speed and productivity. Tasks that once required dedicated engineering or insights support can now be completed in a fraction of the time. For example, building operational dashboards has been reduced from 2–3 days to just a few hours, representing up to ten times improvement in turnaround time. Many requests that would have previously been deprioritized are now handled directly by business users, accelerating decision making across go-to-market and product teams.
Due to this, Zapier has been able to refocus its technical talent on higher-value work. Data experts can focus on building resilient systems, improving data quality and enabling the next generation of AI-driven workflows.
How is Zapier expanding its use of Databricks?
Looking ahead, Zapier is continuing to expand what it can do with Databricks. The team is exploring new ways to bring data even closer to end-user experiences, including experimenting with transactional capabilities that could support more responsive, real-time product features. As Lukas explained, “We are going to be leaning more and more into Databricks MCP and other Databricks APIs. You can accomplish so much.”
Databricks is the foundation for how Zapier will continue to scale intelligence, automation and decision-making across the business.