DocuSign Reports 50% Productivity Gain by Combining Claude Code with dbt Tools

DocuSign has cut the time its analytics engineers spend from requirement to a validated, review-ready dbt asset by half, according to Bishal Gupta, who worked on trusted data products and analytics engineering modernization at the company.

Gupta said SQL generation was never the real bottleneck - context switching was. Engineers bounced between Jira requirements, dbt conventions that lived mostly in people's heads, and repeated validation, with no clean handoff between stages.

His approach puts Claude Code in the orchestrator role. dbt Agent Skills supply project conventions and naming patterns, the dbt MCP server provides structured metadata and lineage, and dbt v2 validates compiled SQL before a pull request goes out.

Instead of one large agent doing everything, the workflow is split into discovery, dbt development, and orchestration phases, so a failed compile or skipped unit test is retried on its own rather than restarting the whole pipeline. In a live demo, the workflow built a new quarterly aggregation model, generated the model and YAML, validated it against dbt v2, and produced a pull request and matching orchestration job - with branch isolation and human review still required before anything ships.

Gupta reported the piloted 50 percent productivity gain came alongside more consistent naming, testing, and semantic patterns, letting reviewers focus on engineering judgment instead of style fixes. His framing: AI proposes, and engineers stay accountable for what actually merges.

On guardrails, Gupta said test runs are capped at 1,000 rows so an agent can't accidentally trigger a billion-row build, and described an early, still-evolving approach to governing which teams' custom skills get layered onto the shared workflow via a master skills file.

Source: dbt Labs Blog