Docusign Wants to Be the AWS of Contracts

When Docusign connects a contract agent to an AI model, it is not just routing a document. It is placing a machine in a role that carries legal weight. On May 21 at its Momentum conference, Docusign unveiled an agent studio for the agreement lifecycle: Iris AI, a no-code builder, pre-built agents, and a tool called Agent Studio, all wired together through the Model Context Protocol (MCP), a shared connectivity standard that lets AI systems talk to each other. The pitch is that Docusign's 1.8 million customers and their billion-plus signed agreements can become the operating substrate for AI agents that not only draft contracts but bind their organizations to them.

Sixty percent of in-house legal teams are evaluating agentic AI for contract workflows in 2026, according to Bind Legal's annual CLM trends report. Docusign is betting it can be the layer those teams build on. Through MCP, Docusign connects to Anthropic's Claude, Google's Gemini, and OpenAI's ChatGPT simultaneously, meaning the system can route contract work to whichever model a customer already uses or prefers.

The open question is where the line runs between automated routing and genuine autonomous execution. A year before Agent Studio, Docusign launched a first generation of AI contract agents that analyzed agreements, highlighted risks, and surfaced issues that usually required human intervention, according to Constellation Research. Those agents did not replace judgment. They surfaced information for a human to act on. The current Agent Studio announcement describes multi-step workflows that route, draft, and send for signature with minimal human intervention. Whether the 2026 version crosses the threshold that the 2025 version did not depends on whether those autonomous decision points are real or aspirational. Docusign has not published case studies showing a live enterprise workflow running end-to-end without a human in the loop.

CTO Sagnik Nandy told The Letter Two that the problem his team built Intelligent Agreement Management to solve is that agreements are slow and inefficient. Iris, the AI assistant in the new stack, is trained on Docusign's proprietary agreement data rather than a generic large language model alone. For contract work, that distinction matters: a model that has processed millions of NDAs, MSAs, and statements of work has a reference point for what a given clause typically means in context, not just what it says on the page. Docusign has surpassed 10,000 IAM customers in Q1 FY2026.

The agents, according to Bind Legal, are designed to handle multi-step contract workflows with minimal human intervention for routine agreements: receive the request, identify the right template, draft the contract, route it for review, answer questions about it, and send it for signature. Docusign is also partnering with legal AI platforms Harvey, Legora, and CoCounsel by Thomson Reuters, layering their capabilities into the same routing infrastructure. If the architecture works as described, a legal ops team could build a contract workflow where Harvey drafts specific clauses, Docusign's Iris extracts relevant data from prior agreements, and the final output routes through Docusign's signature infrastructure without leaving the Docusign control plane. Docusign reported revenue of $763.7 million in Q1 FY2026, up 8 percent year over year, and is targeting fiscal 2026 revenue of $3.13 billion to $3.14 billion.

Pricing for Agent Studio was not disclosed. The tools are in early access in the U.S. today with a full U.S. rollout starting in July 2026, which means the access model today is invite-only, not open distribution. Mid-market teams without a direct Docusign relationship likely cannot build today.

The liability question is not answered. If an agent routes a contract to the wrong template, drafts a clause that exceeds its authorization parameters, or sends something for signature when it should not have, existing contract law does not clearly establish who bears responsibility. Docusign has not published guidance on what happens when an agent makes a binding error. Whether that ambiguity is a product gap or a legal industry problem is an open question.

The Deloitte figure Docusign cites — that organizations using AI-driven workflows with an end-to-end agreement platform see nearly 30 percent higher ROI — comes from a report commissioned by Docusign, so treat it accordingly. The 60 percent evaluation figure from Bind Legal covers the universe of teams surveyed, not the subset already running pilots.

What separates this from a feature announcement is the organizational-authorization problem underneath it. Docusign is arguing that the agreement data it holds, the history of what every customer has signed and on what terms, is the substrate on which AI agents for contracts should operate. Whether that data moat is wide enough to make Docusign the operating system for agentic contract work is the test. Watch three signals: whether independent legal tech platforms build MCP connectors to Docusign on their own, whether the 10,000 IAM customer count grows or stalls, and whether any customer goes on record describing a workflow that ran without a human approving the final signature. Those three numbers answer the question better than any press release.