Docusign now has all of its more than 1,100 engineers actively using AI coding tools, up from just 49% when the company hit a plateau in June 2025, according to an engineering account published by The New Stack. The document management firm turned the stalled rollout around by creating a team of AI Champions and embedding 75 "Multipliers" across development groups, rather than making the technology mandatory. CTO Sagnik Nandy says the shift moved the company beyond basic coding assistants toward reusable agents and automated workflows tailored to Docusign's own codebase.
By June 2025, Docusign reported that roughly half of its engineers were actively working with AI coding tools, while a reported 18% gain in developer productivity had stopped climbing. The company responded by establishing a six-person Eng AI Champions team to build guardrails, share practices, and run workshops, which brought adoption to 65% by October 2025. After extending that effort through AI Multipliers embedded in individual teams, active usage jumped to 93% within two months, while lines of code updated per developer rose by approximately 52% from the starting baseline. The report notes that the code-volume metric does not, by itself, prove a corresponding improvement in delivery speed or software quality.
According to Nandy, many AI programs "start well, and then fizzle out," but Docusign sustained momentum by combining central coordination with feedback from individual teams. The company developed internal tools it jokes "speak Docusign," including Kody for company-specific coding work, Chronicle for observability and incident investigation, and DocuWiki for code-aware internal documentation. The report describes an overnight agent that searches error logs for problems and prepares proposed fixes, so engineers arrive in the morning to find a pull request already drafted. In one example involving infrastructure configuration files, an agent correctly modified 10 of 12 files on its first pass in roughly 15 minutes, using a sample commit as guidance—a process the company says previously consumed about a week of engineer time.
Docusign's orchestration platform, Krew, lets engineers discover, copy, and extend these agents for recurring tasks, with frequently used tools becoming more visible to other teams. Nandy explains the company focused on making successful solutions reusable rather than teaching everyone individually: "If something is good, you can try to teach 1,000 other people to do it, or you can make it easy for 1,000 other people to inherit it." The company emphasizes guardrails before speed, requiring human review on every pull request and blocking agents from merging code or taking irreversible actions without approval. Nandy says Docusign now measures feature velocity, customer satisfaction, and reliability rather than relying solely on adoption percentages or lines of AI-generated code, and he cautions against deploying AI where conventional software would perform better. For engineering leaders, the report recommends choosing a well-defined workflow, agreeing on the outcome to measure, assigning an owner, and specifying where human approval is required before scaling. The broader lesson is that adoption alone won't produce lasting results—teams also need shared tools, clear ownership, engineer feedback, and metrics that show whether the work is actually improving. Organizations that treat AI rollout as a one-time deployment risk watching early wins evaporate, while those that invest in reusable infrastructure and continuous feedback loops stand a better chance of sustaining gains across hundreds of developers.