DocuSign Achieves 90% Cost Reduction in AI Contract Processing
By adopting smaller, task-specific AI models, DocuSign has cut its processing costs dramatically while enhancing performance, setting a new standard for efficiency in contract management.
Key Facts
- Docusign cut AI processing costs by 90% while increasing throughput eightfold, enhancing efficiency.
- Per-token prices dropped 98% since 2022, but enterprise AI bills rose 320%, indicating budget strain.
- Small models outperformed larger ones in production by 7% accuracy, revealing cost-effective strategies.
- 81%-95% of large U.S. enterprises use AI, but only 5%-10% see full ROI, highlighting investment risks.
- Docusign's affordable AI agents enable automation across sectors, indicating a strategic shift in service delivery.
Summary
Summary
DocuSign faced high costs and delays in processing over 1 million contracts daily using large general-purpose AI models. By switching to smaller, task-specific models, the company reduced AI processing costs by 90% and increased throughput by up to eight times while maintaining accuracy within 2 percentage points of the larger models.
Background
DocuSign operates in the digital signature and contract management industry, processing a vast volume of agreements daily. Prior to the deployment of smaller AI models, the company relied on large general-purpose AI models for contract processing, which became costly and inefficient as the volume of contracts increased.
Challenge
The primary challenge was the high cost and processing delays associated with using large AI models for every contract. As the volume of agreements surpassed 1 million per day, the infrastructure budget was heavily impacted by the per-call costs of these models.
Solution
DocuSign implemented a new AI engine, Iris, which reads contracts and extracts over 50 critical facts. Instead of sending entire contracts to large models, DocuSign developed smaller models trained on Microsoft Foundry for routine data extraction tasks. A filtering system was also introduced to identify and send only relevant passages to the AI, improving efficiency and accuracy.
Results
The switch to smaller models resulted in a 90% reduction in AI processing costs and an eightfold increase in throughput. The per-document cost fell 50-fold compared to the previous system, and the smaller models outperformed the larger models in production by 7 percentage points in accuracy.
Key Insights
Businesses can achieve significant cost savings and efficiency improvements by using specialized AI models for routine tasks rather than relying on general-purpose models for all applications. Targeted data extraction can enhance accuracy and reduce unnecessary processing overhead.
Customer Testimonial
“...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.” — Ramachandra Kota, Senior Director of Applied Science, DocuSign.
Entities Mentioned
Companies
Products
Technologies
People
Key Concepts
Definitions
- AI models
- Algorithms designed to perform tasks that typically require human intelligence, such as understanding and processing contracts.
- data extraction
- The process of retrieving specific information from documents, such as contracts, for analysis or action.
- frontier models
- Advanced AI models used for complex reasoning tasks, such as interpreting intricate clauses in contracts.
- Iris
- Docusign's AI engine that reads agreements and extracts critical business information.
- Agent Studio
- A product from Docusign that automates contract-related tasks across various business functions.
Use Cases
- →Automating contract work across sales, HR, procurement, and legal.
- →Pulling obligations from master service agreements.
- →Connecting legal terms to new work orders in Workday and HubSpot.
Frequently Asked Questions
How does Docusign improve contract processing efficiency?
Docusign enhances efficiency by using smaller, specialized AI models for routine tasks, significantly reducing processing costs and increasing throughput.
What is the role of the Iris AI engine?
The Iris AI engine reads contracts and extracts essential facts, enabling businesses to act on critical information quickly and accurately.
What are frontier models used for?
Frontier models are reserved for complex reasoning tasks that require deeper analysis, such as comparing terms across multiple documents.
How has Docusign's approach to AI models changed?
Docusign shifted from using large general-purpose AI models to smaller, task-specific models, which has led to significant cost savings and improved accuracy.
What impact does AI have on enterprise budgets?
Despite rising AI costs, many enterprises plan to increase their AI budgets, indicating a strong belief in the long-term value of AI investments.