[DRAFT] AI at CMS

content/outbound/ai-at-cms.md

@@ -0,0 +1,124 @@
+---
+title: Artificial Intelligence at CMS
+description: An overview of artificial intelligence at CMS
+permalink: /outbound/ai/
+layout: layouts/page
+section: outbound
+tags: ospo
+eleventyNavigation:
+  parent: ospo-outbound
+  key: ospo-outbound-ai
+  order: 14
+  title: AI at CMS
+sidenav: true
+sticky_sidenav: true
+---
+
+At CMS, Artificial Intelligence (AI) has the power to reshape the way we use data to make decisions. Given CMS' vast data resources, we have an unprecedented opportunity to drive innovation, boost productivity, and enhance service delivery through AI.
+
+Learn more about how AI is leveraged on the HHS and CMS levels:
+
+- [AI at HHS](https://ai.hhs.gov)
+- [ai.cms.gov](https://ai.cms.gov)
+
+## AI Policy and Documents
+
+### HHS AI Strategy
+
+As defined in the [HHS AI Strategy Document](https://ai.hhs.gov/#strategy), HHS is leading a whole-of-department transformation to make AI a practical layer of value across public health, health care delivery, biomedical research, human services, and agency operations — uniting Divisions through a “OneHHS” approach that shares data, tools, and infrastructure while protecting Americans’ privacy and earning public trust. There are 5 pillars to the AI Strategy:
+
+1. Governance & Risk Management for Public Trust
+2. Infrastructure & Platforms for User Needs
+3. Workforce Development & Burden Reduction
+4. Research & Reproducibility through Gold-Standard Science
+5. Care & Public Health Delivery Modernization
+
+### HHS M-25-21 Compliance Plan
+
+The White House Office of Management and Budget (OMB) released two memoranda delineating implementation of Executive Order 14179 "Removing Barriers to American Leadership in Artificial Intelligence" for federal agencies:
+
+1. Memorandum M-25-21 on Accelerating Federal Use of AI through Innovation, Governance, and Public Trust
+2. Memorandum M-25-22 on Driving Efficient Acquisition of Artificial Intelligence in Government.
+
+These memos focus U.S. AI policy towards a “forward-leaning, pro-innovation, and pro-competition mindset” and direct agencies to leverage AI to be “more agile, cost-effective, and efficient.”
+
+To address these directives, HHS created a [Compliance Plan](https://www.hhs.gov/sites/default/files/2025-hhs-ai-compliance-plan.pdf) mirroring the pillars of the memos: driving AI innovation, improving AI governance and fostering public trust in federal use of AI.
+
+### CMS AI Playbook
+
+The [CMS AI Playbook](https://ai.cms.gov/CMS-AI-Playbook.pdf) documents the agency’s evolving approach to AI adoption and maturity, offering CMS-specific guidance, tools, and frameworks for leadership, project teams, and IT and security professionals. Version 4 reflects CMS’ transition from early-stage AI exploration to building foundational maturity, and is informed by federal directives and feedback from CMS staff across components.
+
+The Playbook includes six chapters: an introduction to its purpose and audiences; an AI primer to establish shared understanding; a summary of current CMS AI use cases, maturity, and guiding principles; governance recommendations for AI oversight; a step-by-step guide to conducting AI projects; and a forward-looking chapter on emerging technologies and future readiness.
+
+Serving as both a reference and a guide, the Playbook helps CMS teams understand their roles in AI development and adoption. It supports informed decision-making and cross-functional collaboration as AI becomes more integrated into agency operations.
+
+### CMS Technical Architecture Reference
+
+The [TRA](https://www.cms.gov/tra/Application_Development/AD_0010_Application_Introduction.htm) contains guidelines and recommended practices on AI-assisted development in the Software Development Lifecycle:
+
+- AI Code Generation
+- AI Principles for Developers
+  - AI Code Generation Accountability
+  - Tool Selection and Data Privacy Considerations
+- AI Development Methodologies
+  - “Vibe Coding” Policy Restrictions
+  - Suitable Approaches to AI-Assisted Development
+  - AI in the SDLC — Shifting Left: Integrating Quality and Security
+- Examples of AI Benefits in the Software Development Lifecycle
+
+## AI Initiatives
+
+### CMS AI Explorers Program
+
+The [AI Explorers Program](https://confluenceent.cms.gov/spaces/APP/pages/169593017/CMS+AI+Explorers+Program) supports AI organizational maturity at CMS by conducting technical research and development, supporting AI pilots within CMS components, conducting policy research, and engaging the CMS AI community. The team has created numerous [technical application resources](https://confluenceent.cms.gov/spaces/APP/pages/768063421/Technical+Application+Resources) for AI usage and development in model context protocol, agentic engineering, policy analysis, risk + maturity frameworks,prompt engineering, and many more.
+
+## AI Tools
+
+### Want an AI workspace?
+
+The CMS AI Workspace is a JupyterHub environment that provides users with a powerful, user-friendly platform for data analysis, artificial intelligence, and business intelligence projects. Users can create and edit Jupyter notebooks, writing code in a variety of programming languages with essential libraries, LLM APIs, and dependencies pre-installed (e.g., pandas, pip, PyTorch).
+
+To get started on a new project, explore the AI Workspace Cookbooks for ready-to-use JupyterLab notebooks and tutorials that help integrate AI and machine learning workflows.
+
+Visit:
+
+- [https://aiworkspace.cms.gov](https://aiworkspace.cms.gov)
+- [AI Workspace User Guide](https://confluenceent.cms.gov/spaces/SEAS/pages/1006053810/AI+Workspace+User+Guide)
+
+### Want a personal assistant?
+
+#### CMS Chat
+
+`CMS Chat` is CMS' LLM developed by OIT with connections to CMS-specific knowledge sources to answer questions and assist with various tasks. It has the ability to process all PII, PHI, CUI as well as integrate with CMS Jira, Confluence, and CMS OneDrive/SharePoint.
+
+Visit: [cms.chat.gov](http://cms.chat.gov) and login using your EUA credentials
+
+#### Tools provided by HHS
+
+`HHS ChatGPT` is an AI chatbot provided by HHS best used to assist with the following tasks: governance and policy discussions, OA/FISMA analysis, dashboard and metrics design, and complex problem-solving. It is strong at structured thinking, reasoning through tradeoffs, and providing context across systems.
+
+_Be mindful of inputting sensitive data/PII/PHI in this platform as it is not protected – CMS Chat is best for this._
+
+Visit: [go.hhs.gov/chatgpt](https://go.hhs.gov/chatgpt) and login using your CMS email address
+
+`HHS Claude` is another chatbot provided by HHS. It has more Claude platform functionality than CMS Chat (e.g. projects, teams). It is most suitable for generating clear writing and working with large amounts of text whether it be drafting, summarizing, editing, and reviewing documents.
+
+_Be mindful of inputting sensitive data/PII/PHI in this platform as it is not protected – CMS Chat is best for this._
+
+`HHS Gemini` is the third AI chatbot provided by HHS. It is best for real-time research with direct Google search integration for up-to-date information, analyzing massive documents or multimedia files, and deep integration with Google's ecosystem.
+
+_Be mindful of inputting sensitive data/PII/PHI in this platform as it is not protected – CMS Chat is best for this._
+
+Visit: [go.hhs.gov/gemini](https://go.hhs.gov/gemini) and login using your CMS email address
+
+### Want a personal assistant on Google Workspace?
+
+`Gemini` functionality is accessible through the CMS Google Pilot Workspace such as prompt-drive image generation in Google Drive and text editing in Google Docs.
+
+### Want a personal assistant on Outlook?
+
+`Microsoft Copilot` is best used for day-to-day productivity within Microsoft tools. It works well for in-workflow assistance rather than deep analysis (e.g. searching/summarizing your email, calendars, documents, etc).
+
+### Want a personal assistant on Slack?
+
+`SlackAI` is already embedded in our CMS enterprise Slack, with limited functionality for summarizing specific channels/threads within set time ranges.