How Docusign connects AI to trusted data with Fivetran + dbt

"Fivetran gives us the velocity we need to ingest data from disparate sources, while dbt helps us transform and govern that data with consistent metrics and KPIs. Together, they bring pieces of the puzzle together, connecting the entire journey from data ingestion to consumption, so our teams can get more value from our data."
— Bala Sahejpal, Chief Data Officer at Docusign

Docusign is transforming how the world does business. More than 1.9 million customers and 1 billion users across 180+ countries rely on its platform to create, manage, and act on agreements, from everyday transactions to complex enterprise contracts. Now, the company is taking agreement management further with AI, helping organizations unlock insights and automate workflows across the agreement lifecycle.

Behind that global scale is a complex data ecosystem. Docusign needed to connect information across its applications, databases, and business functions to understand performance, make informed decisions, and continuously improve its products and operations. As AI becomes increasingly central to its business, delivering trusted data and context efficiently is more important than ever.

The challenge: Scaling data access and analytics without adding engineering overhead

Docusign needed to solve 2 challenges without increasing engineering overhead or compromising governance: make enterprise data reliably available and give AI the trusted context to work with it.

As Docusign grew, its legacy data infrastructure struggled to keep pace. The company analyzed data from just 6 sources in SQL Server, limiting visibility into business performance and making it difficult for teams to connect information across functions. Adding new sources required custom integrations, with individual pipelines taking 3-6 months to develop and up to 20 hours per week to maintain.

These bottlenecks slowed access to business-critical data, limited cross-functional insights, and consumed engineering resources that could otherwise support innovation. Docusign needed a more scalable way to bring enterprise data together and make it readily available for analytics and emerging AI initiatives.

As its analytics environment matured, another challenge emerged: making AI useful for developing trusted data products. While AI could generate SQL, it lacked the context to understand existing models, dependencies, naming conventions, and testing requirements. Engineers still spent valuable time validating outputs, correcting inconsistencies, and ensuring AI-generated assets met production standards.

“The risk of doing nothing is that you won't help the company meet its growth, efficiency, and productivity goals,” says Bala Sahejpal, Chief Data Officer at Docusign. “If you don't change your data layer, you're not doing justice to all the other initiatives where generative AI is making a difference.”

The solution: Connecting AI to enterprise data through context engineering

Docusign built its data foundation on Fivetran, dbt, and Snowflake. Fivetran automates ingestion from disparate business systems into Snowflake, while dbt transforms that information into governed models with consistent definitions, metadata, and lineage. Together, they make enterprise data accessible and establish the context AI needs to use it reliably.

"Fivetran gives us the velocity we need to ingest data from disparate sources, while dbt helps us transform and govern that data with consistent metrics and KPIs. Together, they bring pieces of the puzzle together, connecting the entire journey from data ingestion to consumption, so our teams can get more value from our data."
— Bala Sahejpal, Chief Data Officer at Docusign

Docusign initially connected Fivetran to Snowflake in just 20 minutes, replacing custom integration work that previously took months. Today, the company is building on that foundation through context engineering: connecting AI agents to the metadata, business definitions, and governance rules needed to understand enterprise data and produce trustworthy results.

Docusign is harnessing metadata from Fivetran, dbt, and Snowflake, then enriching it with input from business users and product managers to clarify definitions and resolve ambiguities. This helps conversational AI agents understand not just where data comes from, but what it means, how it relates to other information, and which insights users are authorized to access.

The same emphasis on trusted context extends to analytics engineering. Using dbt Agent Skills, dbt MCP, dbt Fusion, and Claude Code, Docusign has developed an AI-assisted workflow that connects agents to existing models, dependencies, and engineering conventions. Instead of generating SQL in isolation, agents can follow established standards and help engineers deliver validated data products faster.

The results: Faster analytics, greater efficiency, and trusted AI

Expanding data access without doubling the BI team

With Fivetran, Docusign expanded its analytics environment from 6 data sources to more than 3x that number, bringing previously disconnected business information together in Snowflake. The broader data foundation gives teams more comprehensive visibility into operations and supports the company's growing analytics and AI ambitions.

Automated ingestion also eliminated significant engineering overhead. Docusign estimated that its BI team would have needed to double in size to support the same integration requirements without Fivetran. Instead of building and maintaining pipelines, engineers could focus on governance, data modeling, and higher-value initiatives, including developing the trusted data products that AI depends on.

Improving analytics engineering productivity by 50% with AI and dbt

In an AI-assisted engineering pilot, Docusign achieved a 50% productivity improvement from requirements to validated dbt assets. By connecting AI to existing models, metadata, lineage, and engineering standards, the company reduced manual handoffs, repetitive validation, and context switching while improving the consistency of its data products.

The workflow combines dbt Agent Skills to guide engineering best practices, dbt MCP to provide project metadata and lineage, and dbt Fusion to compile and validate outputs. Claude Code coordinates the process, carrying context from initial requirements through development and validation.

For example, when an engineer requests a new data model, the workflow examines existing sources, models, and dependencies before generating SQL and YAML files. dbt Fusion then validates the output earlier in development, helping catch errors before they cause failed builds, rework, or unnecessary compute costs.

Docusign also maintains automated testing, continuous integration checks, and human approval before production deployment. Rather than bypassing engineering controls, AI works within them, helping developers deliver reliable data products faster while spending more time on complex business challenges.

Bringing trusted business insights to conversational AI agents

Docusign is also applying context engineering to conversational analytics, moving beyond static dashboards toward AI agents that can help business users explore data through natural language.

To make those interactions reliable, Docusign is combining metadata from Fivetran, dbt, and Snowflake with business definitions, governance policies, and access permissions. This helps agents interpret questions accurately, understand what information users are authorized to see, and deliver relevant insights without requiring users to navigate multiple reports.

For Docusign, the opportunity goes beyond faster answers. By making enterprise data easier to explore conversationally, the company aims to help more employees make informed decisions and uncover insights that might otherwise remain buried in dashboards.

What's next: Scaling context-aware AI across the business

Docusign is continuing to expand its AI initiatives, from more sophisticated conversational data agents to AI-assisted engineering workflows and semantic modeling. The focus is on making trusted data and business context accessible wherever AI can help teams work faster and make better decisions.

“I'm very excited about how you can start with a conversation and get to an insight so quickly,” Sahejpal says. “Think about doing that at scale. You're really unleashing the data that's been sitting in your data stores, previously visible only through static dashboards.”

For Docusign, AI readiness is about more than connecting models to enterprise data. It's about giving agents the context, definitions, and governance needed to use that information responsibly. By combining automated ingestion with governed transformation and context engineering, Fivetran and dbt help Docusign build toward AI that delivers reliable insights, accelerates development, and supports business growth.

[CTA_MODULE]