AI
Docusign Built an AI Operating Model Beyond Tool Rollout for 1,100+ Engineers
Instead of stopping at adoption metrics, Docusign shifted focus to embedding AI-powered workflows directly into how engineering teams work, measure value, and reuse solutions.
1/3
The New Stack
Docusign faced a familiar challenge: rolling out AI coding tools across engineering generated early momentum, but gains stalled. By mid-2025, roughly half its engineers were using the tools, and a 18% productivity bump had plateaued. The company realized that tool deployment alone wasn't enough—teams needed time to experiment, training suited to their codebase, and help weaving AI into their actual workflows.
The response transformed how Docusign approached AI at scale. Rather than mandating adoption, the company moved toward building an AI operating model, shifting from isolated tool use to integrated, reusable systems. CTO Sagnik Nandy emphasized the distinction: "There's a big difference between trying to get a few wins versus doing it consistently and doing it right."
From Champions to Multipliers#
Docusign established a six-person Eng AI Champions team to create shared guardrails, document practices, and run workshops. Adoption climbed to 65% by October 2025. The breakthrough came through embedding "AI Multipliers"—about 75 engineers embedded across individual teams—tasked with deploying central solutions within their groups and funneling problems back to leadership.
This two-way feedback loop proved critical. Active usage jumped to 93% within two months, and lines of code updated per developer rose roughly 52% from baseline. Nandy credits the durable success to this combination of top-down coordination and ground-level insight: rather than treating AI adoption as a one-time rollout, the model continuously captures what works, what breaks, and what teams actually need.
Teaching Systems to Know Your Business#
Once engineers were comfortable with AI assistance, Docusign expanded the ambition. The company built tools that could access its repositories, architecture, telemetry, and internal practices—essentially teaching AI to "speak Docusign." A suite of agents emerged: Kody for proprietary coding tasks, Chronicle for incident investigation, and DocuWiki for code-aware documentation.
These specialized systems converge through Krew, an orchestration platform that lets teams compose agents called Krewmates for recurring workflows. Nandy described one example: an overnight agent that scans error logs, proposes fixes, and presents a pull request to engineers by morning. Another agent handled infrastructure-file updates in 15 minutes—a task that previously consumed a week of manual work—though human review remained mandatory before any merge.

