Docusign scales AI on Azure, processes 1M+ agreements daily with eightfold throughput

Enterprises execute billions of agreements every year. Buried inside are renewal deadlines, pricing terms, obligations, and risk indicators that shape business decisions, but surfacing that intelligence has historically required manual review.

As organizations look to turn agreement data into business intelligence, the challenge is understanding agreements at scale across volumes of complex, unstructured data without introducing cost and latency constraints. Serving more than 1.9 million customers across 180 countries, Docusign saw an opportunity to transform agreement intelligence into a strategic asset for its customers.

A foundation for innovation at scale

To help customers unlock intelligence hidden within their agreements, Docusign first built its Intelligent Agreement Management (IAM) on Microsoft Azure, scaling agreement intelligence across millions of documents on a platform designed for innovation, efficiency, and growth.

“Having a modern cloud-first stack was going to be a huge advantage for our ability to innovate, run the system sustainably and efficiently, and scale. Azure took care of the building blocks, helping us create the product we are uniquely positioned to build,” says Sagnik Nandy, CTO at Docusign.

The Azure foundation made it possible to scale infrastructure, but scaling AI itself required a different approach. Early efforts relied on large, general-purpose models to process agreements. While effective, those models introduced growing pressure across cost, latency, and throughput when applied to millions of documents, each with variable formats, scanned inputs, tables, and structures.

Prabhakar Marnadi, Senior Principal Engineer at Docusign, and his team began to rethink the problem. Instead of asking how to scale general-purpose models, they asked whether those models were necessary for every task. For highly specific agreement-processing jobs, the answer was often no.

A specialized model could match the performance of a larger model, but at a fraction of the cost and with significantly higher throughput. That insight became a proof of concept and, eventually, a production system.

Rethinking AI economics at scale

“At a million documents a day, the difference between sending a full 100-page contract and sending the 4,000 tokens that actually matter is the difference between a viable business and an economics problem,” says Ramachandra Kota, Senior Director of Applied Science, Docusign.

That realization led to a shift in both model choice and architecture. Docusign moved from relying on general-purpose models to a system of task-specific models optimized for agreement intelligence workflows. Using Microsoft Foundry and Azure OpenAI in Foundry Models, the team implemented a production AI pipeline that’s enterprise-ready.

Foundry helps Docusign operationalize AI workflows within its existing architecture. “Microsoft Foundry gave us the ability to move from large general-purpose models to small, specialized fine-tuned models and combine that with better context engineering. That one investment helped move cost and throughput at the same time, which was a great outcome,” explains Kota. Rather than sending entire agreements to a model, Docusign developed a discipline around context engineering, ensuring models only process the portions of a document relevant to a given task. 

Building a production-ready AI lifecycle

At the center of this system is a model lifecycle designed for continuous improvement. Larger foundation models act as “teacher” models, generating high-quality training data across a wide range of agreement types. That data is validated and used to fine-tune smaller models, which are evaluated before deployment and continuously monitored in production.

Azure Kubernetes Service (AKS) provides the scalable microservices foundation required to process agreement workloads at enterprise volume. Azure Cosmos DB and Azure Database for PostgreSQL support the storage and management of agreement intelligence data, helping power search, automation, and downstream workflows. Together with Foundry and Azure OpenAI, these services bring AI, applications, and data into a unified architecture rather than separate systems.

“By combining Microsoft Foundry, Azure OpenAI in Foundry Models, Azure Kubernetes Service, and Azure database services, we built a platform that lets us continuously improve model performance while scaling agreement intelligence efficiently and reliably across our customer base,” says Marnadi.

The first production application of this architecture was Agreement Manager, Docusign’s agreement intelligence capability, which surfaces more than 50 data points from agreements with high levels of extraction accuracy. Through model fine-tuning and optimization, Docusign increased document-processing throughput by up to eight times while maintaining accuracy within two percentage points of larger models and reducing AI processing costs by 90%. This efficiency enabled the company to scale AI-driven capabilities more efficiently as volume grows.

Instead of introducing new infrastructure or siloed AI systems, Docusign extended its existing Azure environment, reducing complexity and supporting continuous optimization.

Scaling AI with enterprise trust

Agreement data is among the most sensitive information that organizations handle, often subject to regulatory and residency requirements. Docusign processes data in-region using Azure services and isolates workloads at the tenant level on AKS, helping ensure customer data remains highly secure and separated.

Foundry provides governance and traceability across the AI lifecycle, so Docusign can track model interactions while maintaining enterprise requirements for compliance, visibility, and control. “That traceability is what lets us move fast on model rollouts while still satisfying enterprise and regulatory review,” says Kota.

With this foundation in place, Docusign is moving beyond document understanding toward broader automation. The same system that extracts intelligence from agreements can now be extended to act, triggering workflows, surfacing risk, and accelerating decision-making across the business.

From agreement intelligence to business action

Today, IAM processes over 1 million agreement documents each day, transforming agreement content into structured data that organizations can use to make decisions, automate workflows, and drive business outcomes. For users, the impact is immediate.

Tasks that once required manual review, tracking renewal dates, analyzing contract terms, and identifying obligations, can now be surfaced in near real time or handled automatically. This reduces operational overhead, shortens decision cycles, and allows teams to focus on higher-value work.

SEI, a management consulting firm, uses Docusign CLM as a centralized platform for managing client, vendor, partner, and subcontractor agreements across its organization. As part of its adoption of IAM, SEI is utilizing AI-powered capabilities to help identify and organize key contractual information, improving visibility into agreement data and supporting more efficient contract lifecycle management.

“Through several Azure services, we synthesize our Master Service Agreements [MSAs] and contracts and attach them to our core systems, like Workday and HubSpot, to streamline new work orders by taking a holistic look at all of the legal obligations with our client, vendors, and partners,” says Bill Gallagher, CEO, SEI. “Agreements contain important operational and business information that can be difficult to access when buried within individual documents. IAM helps us surface key contract data more efficiently, enabling our teams to spend less time searching for information and more time focused on delivering value to the business.”

Docusign’s platform demonstrates how that shift can be achieved. By combining fine-tuning, context optimization, and a unified cloud-native architecture, the company created a repeatable model for scaling AI across enterprise workloads. This allows AI to move beyond isolated use cases and become a core capability that powers automation, accelerates decisions, and enables new business outcomes at enterprise levels.

As Docusign continues to expand its platform, the question is no longer what intelligence exists within agreements. It is what intelligence can do now.

Discover more about Docusign on Facebook, LinkedIn, X, and YouTube.