Unleashing the Power of AI Globalisation at Docusign

AI Globalisation at Docusign: Empowering Customers with...

Learn how you can transform your agreement management with AI. New capabilities to help organisations succeed and grow—uncovering valuable insights, streamlining processes, and eliminating inefficiencies. Insights on how to simplify your team's workflows, whether you're in Sales, Legal, IT, HR, or Procurement. Docusign Iris, our AI engine, is built to understand agreements in multiple languages, helping customers get insights no matter where they do business—using hybrid AI models, secure infrastructure, and built-in governance to perform reliably across regions and languages.

Key Takeaways:

Navigator multilingual insights – Docusign Navigator delivers AI-powered agreement insights in English, French, and German—with expanded insights added in April and support for Brazilian Portuguese, Spanish, and Japanese planned later this year.

AI-Assisted Review – Docusign AI-Assisted Review will be available in French and German in May, with plans to expand to Brazilian Portuguese and Spanish later this year.

Hybrid AI approach – Combines third-party large language models with proprietary multilingual AI to extract agreement insights at scale.

Global enterprise infrastructure at scale – Docusign Iris operates on a secure, globally distributed infrastructure, ensuring low-latency performance at scale to support agreement management worldwide.

AI governance at the heart – Trust and transparency drive our AI globalisation efforts, with in-product controls giving customers full oversight of data use for AI training.

Docusign serves organisations around the world in over 180 markets, which is why we’ve adopted a global-first mindset for AI development—designing our models from the outset to bring AI-powered agreement insights to customers in multiple languages. In December 2024, we reached a major milestone by launching AI globalisation support, expanding our Intelligent Agreement Management (IAM) platform to provide agreement specific multilingual AI capabilities at scale.

As a flagship example, Navigator is the first step in applying this AI-powered approach. It now enables customers to extract agreement insights in English, French, and German, with support for Brazilian Portuguese, Spanish, and Japanese planned for later this year. In April 2025, we added additional insights in English (Global), French, and German to build on the initial December release. See the Language and Regional Support topic for the complete list of available insights per language and more details on language availability and regional expansion. All of this is made possible with Docusign Iris – the AI engine that powers AI capabilities across the agreement lifecycle, enabling you to create agreements faster, commit with greater confidence, and manage for maximum business impact. AI globalisation is pivotal for enhancing the user experience, ensuring our AI tools are accessible to all and ...

AI Infrastructure and Data Management:

A key differentiator for Docusign is our robust and globally distributed infrastructure, designed to operate at scale for large enterprises. With AI services deployed across North America (NA), Europe (EU), Canada (CA), Australia (AU), and soon in Japan (JP), we provide low-latency performance and data residency in regions where our customers do business. This infrastructure allows us to seamlessly handle high volumes of agreements in multiple languages, a feat that many providers often struggle to achieve.

Key benefits for customers:

Scalable infrastructure – Whether processing 100 or 100,000 agreements, Docusign Iris is built to scale to meet business demands. Multilingual AI models are deployed globally, ensuring resilience and availability, with pipelines that adapt to growing operational needs.

Data residency – Agreements and extracted data can be stored within region-specific data centers, to provide customers with peace of mind and safeguards to customer data.

Consistency across regions – Enterprise customers benefit from uniform, enterprise-grade performance across all regions, ensuring smooth operations.

We adopt a hybrid AI model approach to enable accurate and scalable AI across multiple languages. This involves leveraging third-party large language models like Azure Open AI alongside in-house models specifically designed for contract processing and agreement intelligence.

Data Acquisition and Quality:

Sourcing high-quality, diverse datasets is fundamental to agreement AI globalisation. Real-world contracts and agreements vary significantly across languages and countries, making it challenging to build a globally representative dataset for each use case. Our approach focuses on ensuring data quality and relevance through the following methods:

Vendor partnerships – Partnering with trusted data vendors to source multilingual contracts, expanding the breadth and depth of our datasets.

Anonymised, aggregated customer data – By utilising anonymised and aggregated customer data from customers who have consented, Docusign enhances the accuracy and performance of AI-assisted offerings. This includes anonymisation techniques such as redaction, masking, and replacement with non-sensitive alternatives, and aggregation methods that combine and merge data into a unified set to desensitise source data. It is important to note that customers retain full control over their data and can disable data sharing for AI training at any time. Learn more about this in Docusign’s AI-Powered Future, With Customer Control at Its Core.

High quality data – We methodically select a representative set of agreements reflecting diverse legal structures and languages. Duplicates, templates, and irrelevant content are removed to ensure high-quality inputs for model training. By leveraging this diversified data acquisition pipeline, we improve AI development across languages, reduce biases, and enhance model performance.

Dataset Annotation and Quality Control:

Labelling multilingual datasets at scale is complex, requiring precision to ensure AI models perform consistently across languages. The quality of AI models is only as good as the data on which they are trained. To address this, we are enhancing automation and reinforcing quality control through the following initiatives:

Annotation automation – We are investing in smart labelling tools that automate data annotation by combining machine learning with human oversight, accelerating the process while maintaining accuracy.

Parallel annotation – Datasets are annotated simultaneously across multiple languages to ensure consistency in labelling, reducing variability and aligning data quality across regions.

Quality control and validation loops – Our locale-specific subject matter experts validate substantial amounts of annotated data, applying their expertise iteratively to refine labels until the desired precision is consistently achieved. By combining automation with human expertise, we build high-quality datasets that power reliable multilingual AI models.

Model Refinement and Deployment:

By leveraging a combination of AI techniques, we refine our models to deliver consistent, high-quality agreement insights at scale.

Key components of our approach:

Snippet Selection - We employ AI-driven snippet selection using Docusign AI models to extract the most relevant sections from lengthy contracts, reducing noise and improving accuracy. This reduction in document size accelerates contract analysis—giving customers faster, more reliable insights.

Prompt Engineering - We use prompt engineering with large language models to enhance multilingual data extraction. Prompts are designed to account for language-specific legal nuances and are iteratively refined based on feedback and observed performance.

Post-Processing and Normalisation - Extracted data undergoes post-processing to improve accuracy and consistency across languages. This includes error correction, filtering of irrelevant results, and normalisation to maintain standardized outputs.

Fine-tuning Models - Our strategy includes fine-tuning AI models to improve accuracy and scalability over time. Through iterative model improvements, we enhance our ability to process agreements efficiently across multiple languages.

AI Governance and Security:

AI governance is at the heart of our globalisation strategy. From the earliest stages of model development, we prioritize building trust by embedding robust governance practices into our AI systems. Key areas of focus include:

Data protection & security - When you use Docusign Iris, your data is encrypted both in transit and at rest preventing unauthorized users from accessing customer data. If you consent to sharing data for AI/ML training, it is aggregated and anonymized before use. When anonymized data is used to customize AI models, those models are only available for Docusign to use.

Transparency - Your anonymized data is only used for AI/ML training with your consent, and you have the flexibility to manage your consent. Make informed decisions confidently with clear documentation available at your fingertips.

Accuracy & fairness - Docusign Iris is built to minimize errors and harmful outputs, vital in high-stakes scenarios like renewals or negotiations. We use diverse datasets and thorough checks to correct skewed outputs and ensure quality before model deployment. You also have the final say to approve outputs, so they always meet your standards.

Extensibility & adaptability - Our compliant storage and compute platform for data labelling and...