Docusign faced a common challenge in adopting AI technologies; after launching coding assistance tools across its engineering teams, productivity rates stagnated by mid-2025. Although 49% of engineers were using these tools, initial gains did not translate into sustainable improvements in delivery speed or software quality.
The "Multipliers" Strategy
Instead of mandating tool usage, the company adopted an organizational approach centered on a core team of six experts (Eng AI Champions) to set standards, supported by 75 "Multipliers" from within the engineering teams. These Multipliers were responsible for deploying technical solutions and gathering direct feedback, which helped raise active usage rates to 93% in just two months.
Building Agents That "Speak Docusign"
The company moved from simply using coding assistants to developing custom AI agents (Krewmates) that understand the company's internal context, including infrastructure, coding standards, and error logs. Notable tools include:
- Kody: For Docusign-specific coding tasks.
- Chronicle: For monitoring and investigating technical incidents.
- DocuWiki: For internal code-aware documentation.
The Krew platform allows engineers to orchestrate these tools into automated workflows; for example, an AI agent can scan error logs overnight and suggest Pull Requests ready for human review in the morning, reducing time spent on routine tasks from days to minutes.
Why This News Matters
This shift demonstrates that the true value of AI in software engineering lies not in merely providing tools, but in fully integrating them into the Software Development Life Cycle (SDLC). Docusign's experience confirms that success requires:
- Continuity: Having a team responsible for maintaining and evolving workflows.
- Transparency: Making agents and solutions discoverable and reusable by other teams.
- Human Oversight: Ensuring that AI does not make final decisions (such as merging code) without human review, while maintaining quality and security standards.
Sagnik Nandy, Chief Technology Officer at Docusign, emphasizes that the focus must shift from measuring the "number of AI-generated lines of code" to measuring actual outcomes such as feature delivery speed, system reliability, and customer satisfaction.