Donovan
Donovan is an AI-powered decision platform specifically designed for defense operators. The tool aims to assist operators in understanding,...
Last verified:
What is Donovan?
Donovan is Scale AI's AI-powered decision-making platform specifically engineered for the public sector, including defense, intelligence, and federal civilian agencies. It enables warfighters, operators, and analysts to understand, plan, and act in minutes instead of weeks by ingesting and synthesizing vast amounts of structured and unstructured data. The platform goes beyond simple summarization and chat, bringing advanced capabilities around reasoning, tool use, and agentic workflows that connect data sources, identify critical information, run code, and validate sources.
Key features include the Agent Factory for building customized AI agents with a no-code interface, model comparison and evaluation tools to test multiple LLMs, an Agent Arsenal with pre-built mission-optimized agents, support for 27 non-English languages, document upload capabilities (up to 100 documents, 126 MB max), and seven analytic functions including Closed Question & Answer, Translation, Analytic Reasoning, Tool Use, Conversational Chat, Creative Writing, and Tool Selection. Donovan integrates with SGP for data source access and supports commercial models from OpenAI (ChatGPT 3.5 Turbo), Anthropic (Claude), and Cohere (Command), plus Scale's fine-tuned models like Defense Llama.
Donovan is designed for U.S. government personnel, military service members, defense operators, intelligence analysts, and federal civilian agencies. Leading federal agencies including the U.S. Army and U.S. Air Force leverage Donovan to process mission-critical data, distill it, and produce actionable intelligence in seconds. The platform is deployable on classified and air-gapped networks, making it suitable for secure government environments where classified materials cannot transfer to open-source AI systems.
The platform emphasizes traceability with transparent detailed reasoning, is model-agnostic allowing users to select preferred cloud and model providers, offers cloud-agnostic Kubernetes containerized deployment, and aligns with DoD AI ethics principles and AI readiness goals. It augments human intelligence rather than replacing it, force-multiplying staff bandwidth for mission success.
Donovan pricing
Pricing model: Free
Donovan offers a free trial exclusively for U.S. Government personnel. The standard trial is 7 days per employee for verified .gov and .mil users, while the API trial offers 14 days. Some sources mention a 30-day free trial option. The trial is primarily for FAR Part 10 Market Research purposes and is not intended for US Government production data or official governmental purposes. During the trial, users can upload up to 100 documents (126 MB max) and access seven LLM models. Full access to Donovan for production use is available under Scale's Main Services Agreement (MSA) in line with FAR 12.212 / Defense FAR Supplement 227.7202, accessible to US Government Customers and End Users. No specific pricing for paid plans is publicly disclosed; users must book a demo for enterprise pricing.
Donovan pros
- Deployable on classified and air-gapped networks for secure government environments
- No-code Agent Factory interface for building customized AI agents without developers
- Supports 27 non-English languages for document reading and summarization
- Seven analytic functions including translation, reasoning, and tool use
- Upload up to 100 documents with 126 MB maximum file size
- Access to seven LLM models including ChatGPT, Claude, and Command
- Model comparison tools with factual accuracy and hallucination rate metrics
- Pre-built Agent Arsenal with mission-optimized AI agents
- Cites all responses showing which documents insights come from
- Agentic workflows that run code, correct errors, and validate sources
- Cloud-agnostic Kubernetes containerized platform deployment
- Model-agnostic selection of preferred cloud and model providers
- Transparent detailed reasoning for traceability
- Aligns with DoD AI ethics principles and readiness goals
- Processes 100,000+ pages of live data for actionable insights
- Natural language queries require no training or coding experience
- DISA IL4 and FedRAMP High Authorized for unclassified/CUI GovCloud
Donovan cons
- Trial limited to U.S. Government personnel only
- Seven-day trial period (or 14-day/API trial) not for production data
- Primarily for FAR Part 10 Market Research, not official governmental purposes
- Only supports text files (.docx, .txt, text-heavy PDFs) and PowerPoint text
- Not intended for US Government production data use
- Requires verified government employee credentials for access
- Limited to controlled U.S. government environments for custom defense LLMs
- May require integration with existing systems for full operational use
- Public sector focus limits commercial or private sector applicability
Frequently asked questions about Donovan
What is Donovan and what does it do?
Donovan is Scale AI's AI-powered decision-making platform specifically designed for the public sector including defense, intelligence, and federal civilian agencies. It helps operators understand, plan, and act in minutes instead of weeks by ingesting and synthesizing vast amounts of structured and unstructured data. Donovan goes beyond summarization and chat with capabilities around reasoning, tool use, and agentic workflows that connect data sources, identify critical information, run and correct code, and validate sources. It generates summary reports, courses of action (COAs), and actionable intelligence to force-multiply staff bandwidth.
Who can use Donovan and is it available to the public?
Donovan is exclusively available to U.S. Government personnel, including military service members, defense operators, intelligence analysts, and federal civilian agency employees. Verified .gov and .mil users can sign up for a 7-day trial. The platform is not available to the general public or commercial users. Leading federal agencies including the U.S. Army and U.S. Air Force already leverage Donovan's capabilities for mission-critical data processing.
Can Donovan be deployed on classified networks?
Yes, Donovan is one of the first LLMs deployed on classified networks and is field-ready on secure government networks. It is available on both unclassified/CUI GovCloud (DISA IL4 and FedRAMP High Authorized) and classified networks. The platform is cloud-agnostic with Kubernetes containerized deployment, allowing seamless embedding with existing applications and tools in secure, air-gapped environments where classified materials cannot transfer to open-source AI systems.
What document types and file sizes can I upload to Donovan?
Donovan supports uploading text files including .docx, .txt, and text-heavy PDFs. It also supports the text-components of Microsoft PowerPoint (.pptx) files. During the trial, users can upload up to 100 documents with a maximum file size of 126 MB total. Donovan processes these documents so LLMs can access them, enabling users to chat with their data and get cited responses.
What large language models are available in Donovan?
Donovan provides access to seven LLM models in the trial, including commercial models from OpenAI (ChatGPT 3.5 Turbo), Anthropic (Claude), and Cohere (Command). Beyond these commercial options, Scale fine-tunes LLMs for unique public sector use cases including Scale GPT4AIIJ, Scale GPT-J English & Chinese, Scale MPT Chat, and Scale Vicuna. Defense Llama, specialized from Meta's Llama 3 for national security missions, is also available exclusively in controlled U.S. government environments through Donovan.
What analytic functions does Donovan support?
Donovan's models are aligned to seven analytic functions: Closed Question & Answer for precise queries, Translation for preserving semantic nuance and contextual meaning, Analytic Reasoning for data analysis, Tool Use for executing external functions, Conversational Chat for interactive dialogue, Creative Writing for report generation, and Tool Selection for choosing appropriate capabilities. These functions enable comprehensive data analysis, report generation, real-time translation for intelligence gathering, and daily briefing generation.
How does the Donovan trial work and what's included?
The Donovan trial is available to U.S. Government personnel for a 7-day period (some sources mention 14-day for API trial, 30-day for general trial) per employee, primarily for FAR Part 10 Market Research purposes. During the trial, users can upload up to 100 documents (126 MB max), access seven LLM models including ChatGPT, Claude, and Command, use all seven analytic functions, search pre-loaded open-source national security and government datasets, and experience Donovan's natural language query capabilities. The trial is not intended for US Government production data or official governmental purposes.
What is the Agent Factory and how does it work?
The Agent Factory is Donovan's no-code interface for building and customizing AI agents. Users can access data sources, create knowledge bases, leverage state-of-the-art models, assign agent instructions, and connect agents to existing systems and tools without relying on external developers. The integration with SGP enables users to build mission-tailored AI agents from concept to combat in record time. Users can customize agents like military intelligence analysts with specific system prompts and knowledge bases containing GDELT Event Data, DoD Policy Documents, and Theatre Operations databases.
Does Donovan support non-English languages?
Yes, Donovan can read and summarize documents written in 27 different non-English languages. This capability enables secure evaluation of classified documents across multiple languages, data sources, and data types, as well as real-time language translation for intelligence gathering. The translation function preserves semantic nuances and contextual meaning to enhance intelligence output quality.
How does Donovan ensure traceability and accountability?
Donovan provides transparent detailed reasoning that helps users understand the steps and resources AI agents used to arrive at an answer. The platform cites all of its responses, so users always know which documents their answers and insights come from. Donovan aligns with DoD AI ethics principles and is engineered for accountability, speed, and scale. The platform includes monitoring & insights, guardrails & redteaming capabilities, and full audit trails for incident response playbooks and automated tasks.