
Unprompted: How DocuSign Is Putting AI to Work
In this episode of Unprompted, Glean COO Amar Maletira sits down with DocuSign CIO Shanthi Iyer, to hear how DocuSign has moved from broad experimentation to focused, governed AI adoption, ensuring trust as a condition for scale.
AI transformation rarely starts with a single breakthrough. At DocuSign, the journey began with years of work applying AI to agreements and business processes, then accelerated as employees explored new tools and the company evolved toward a broader agreement platform. That combination of business change, employee curiosity, and pressure to do more with existing resources created an opportunity to rethink how work gets done with Glean - not simply add another technology initiative.
Start with a business priority, not an AI showcase
DocuSign initially provided employees with widely available AI tools. But broad access alone made it difficult to understand where productivity was changing. The team’s approach became more focused: connect AI work to company priorities, identify processes where it could make a meaningful difference, and begin with a well-defined use case.
One example was sales development. The company started with older, untouched leads—a contained place to learn—and expanded the effort over time to inbound and outbound activity. The lesson is practical: choose a business problem with clear boundaries, learn from real use, and scale what proves useful rather than treating every experiment as a company-wide transformation.
Make room for learning and course corrections
Iyer was candid that the journey included missteps. Some initiatives did not work and were stopped. That is part of the work of moving beyond pilots: teams need room to test ideas, understand what is and is not helping, and redirect effort without treating every experiment as a success.
Iyer also described a gradual expansion of agent use in sales workflows. For example, AI could help surface signals about an opportunity, while the seller remains involved in deciding whether to advance it. Keeping people in the loop can make automation useful while preserving human judgment where it matters.
Treat adoption as a capability to build
Access to tools is not the same as knowing how to apply them. DocuSign found that employees needed help translating new capabilities into everyday work. Leaders began sharing practical examples in company meetings, showing concrete ways colleagues were using AI. The company also introduced role-based learning pathways through an AI Academy initiative.
This approach pairs bottom-up experimentation with structured enablement. Employees can see relevant examples, build skills for their roles, and contribute ideas grounded in their day-to-day work.
Build on sound processes and trusted data
Iyer emphasized that AI should not be used to automate a broken process. Before introducing automation, organizations should ask what can be eliminated, simplified, or made less burdensome. The quality of the underlying data matters just as much: acquisition, cleanliness, ownership, governance, and clear understanding of how data is used all shape whether AI can deliver dependable results.
In other words, AI transformation is also process and data transformation. Those foundations can make it easier to move from a promising demonstration to a solution that works reliably in real business operations.
Make trust part of the design
For organizations serving regulated industries, the question is not only whether an AI system can perform a task. Customers also want to understand how it handles data, why it produced an outcome, and whether its actions can be traced and audited. DocuSign described bringing legal, security, and product and technology teams together to shape governance, with more scrutiny as initiatives move toward production and broad release.
That progressive approach aims to balance two needs: let teams learn early, then apply stronger checks as the potential impact and scale increase. Human oversight, explainability, privacy, and auditability are not afterthoughts; they are part of what makes AI appropriate for consequential work.
Measure what matters at each stage
Not every AI initiative can demonstrate financial impact immediately. DocuSign described evaluating opportunities by effort, cost, and potential impact, while matching measurement to maturity. For a use case tied to a visible business process, teams may be able to track progress through stages of a funnel. For an early-stage tool, initial signals such as repeat use and feature engagement can help show whether people are finding value at all.
This creates a more useful path from experimentation to investment: start by checking whether a solution is being used, then connect usage to workflow outcomes and business impact where the evidence allows.
Explore more customer stories
成果を生み出すWork AI











