The AI CDAIO Office

Your virtual Chief Data & AI Office -- 22 specialized Claude agents that mirror a real CDAIO organization, with a simulated Board of Directors and Industry Advisory Panel providing shareholder-value oversight.

Not prompt templates. Not chatbot wrappers. A complete CDAIO team that produces boardroom-ready deliverables indistinguishable from top-tier consulting output.

Why This Exists

The average CDO tenure is 2.4 years. You have 14 primary responsibilities (Gartner 2024), a board that wants ROI proof, and a CEO who just asked for "an AI strategy by Friday."

You don't need another AI tool. You need a team.

The AI CDAIO Office gives you one -- a virtual office staffed by Claude agents who have the same roles, expertise, and collaboration patterns as a real CDAIO organization.

Background reading: How AI Will Reshape the Enterprise Data Function by 2028 -- the research and thesis behind this project.

The Team

graph TD
    Board["Board of Directors<br/><i>Shareholder Gate</i>"]
    CDAIO["CDAIO<br/><i>The Orchestrator</i>"]
    CoS["Chief of Staff<br/>& Operations"]
    HGov["Head of<br/>Data Governance"]
    HAI["Head of AI<br/>& Analytics"]
    Exec["Exec Comms"]
    PM["Program Mgr"]
    Steward["Data Steward"]
    Custodian["Data Custodian"]
    Compliance["Compliance Officer"]
    Eng["Data Engineer"]
    AIML["AI/ML Lead"]
    UC["Use Case Lead"]
    Analyst["Data Analyst"]
    SR["Stakeholder Relations"]
    QR["Quality Reviewer"]
    FS["Financial Services<br/>Advisor"]
    HC["Healthcare<br/>Advisor"]
    RT["Retail & Tech<br/>Advisor"]
    IND["Industrial<br/>Advisor"]
    PS["Public Sector<br/>Advisor"]

    Board -->|reviews| CDAIO
    CDAIO --> CoS
    CDAIO --> HGov
    CDAIO --> HAI
    CoS --> Exec
    CoS --> PM
    CoS --> SR
    HGov --> Steward
    HGov --> Custodian
    HGov --> Compliance
    HAI --> Eng
    HAI --> AIML
    HAI --> UC
    HAI --> Analyst
    CDAIO -.->|QA gate| QR
    CDAIO -.->|domain review| FS
    CDAIO -.->|domain review| HC
    CDAIO -.->|domain review| RT
    CDAIO -.->|domain review| IND
    CDAIO -.->|domain review| PS

Installation

Claude Code (CLI)

# Add the marketplace and install
claude plugin marketplace add abensrhir/ai-cdo-office
claude plugin install ai-cdo-office

All 20 /cdaio: slash commands, 22 agents, and 10 playbooks are now available in every Claude Code session.

Team installation (shared config)

Install with project scope so teammates get the plugin automatically:

claude plugin install ai-cdo-office@abensrhir-ai-cdo-office --scope project

This adds the plugin to .claude/settings.json in your repo. When teammates open the project in Claude Code, they'll be prompted to install.

Claude Cowork

The AI CDAIO Office is also available as a Cowork plugin. In the Claude Cowork tab, search for ai-cdo-office in the plugin directory and enable it for your workspace.

Verify installation

claude plugin list    # Should show ai-cdo-office

Then open Claude Code and type /cdaio: -- you'll see all available workflows in the autocomplete.

File generation dependencies

Agents produce boardroom-ready PPTX, DOCX, and XLSX files using Claude Code's native code execution. Install the dependencies:

npm install                    # Installs pptxgenjs (PPTX) and docx (DOCX)
pip install openpyxl pyyaml    # For XLSX generation

No additional configuration needed -- agents read the design system and blueprint files, then write production code directly.

Quick Start

Type /cdaio: in any Claude Code session to see all available workflows. Or just describe what you need -- the CDAIO agent routes your request to the right team member(s).

Playbooks (multi-agent orchestration):

• /cdaio:strategy -- "Build me an AI strategy"
• /cdaio:governance -- "We need a data governance framework"
• /cdaio:board-prep -- "I have a board meeting in two weeks"
• /cdaio:first-90-days -- "I just got promoted to CDAIO"
• /cdaio:quarterly -- "Set up our quarterly reporting"

Programs (multi-step with specialized agents):

• /cdaio:ai-governance -- Responsible AI & model governance program
• /cdaio:org-design -- Data organization redesign
• /cdaio:data-quality -- Data quality improvement program
• /cdaio:cost-optimization -- Data/AI cost optimization sprint
• /cdaio:vendor-eval -- Vendor & technology evaluation

Deliverables (single-agent focused):

• /cdaio:assess -- AI/Data maturity assessment
• /cdaio:policy -- Data governance policy
• /cdaio:deck -- Board deck or executive presentation
• /cdaio:architecture -- Target-state architecture blueprint
• /cdaio:use-cases -- AI use case identification and scoring
• /cdaio:benchmark -- Industry benchmarking scorecard
• /cdaio:compliance -- Regulatory compliance review
• /cdaio:raci -- RACI matrix for any initiative

Utilities:

• /cdaio:review -- Quality review any deliverable
• /cdaio:init -- Set up project folder (context/ + deliverables/)

All commands accept inline context: /cdaio:strategy healthcare company, $2B revenue, 5000 employees

Project Setup

Set up your project with one command:

/cdaio:init

This creates the project structure:

your-project/
├── context/
│   ├── client-context.md       ← fill in your organization details
│   └── (supporting docs: PPTX, PDF, DOCX, XLSX)
├── deliverables/
│   └── (agents create subfolders per deliverable)
└── .gitignore

Or set up manually:

mkdir -p context deliverables
cp "$(claude plugin list --json | jq -r '.[] | select(.name=="ai-cdo-office") | .installPath')/context/client-context.template.md" context/client-context.md

All agents save output to deliverables/<command>-<date>/ automatically. Playbook runs group all outputs in a single dated folder.

Playbooks

Multi-agent workflows for common CDAIO scenarios:

How It Works

Each agent is a single markdown file with:

• Personality: Real expertise, opinions, and voice -- not a job description
• Workflow: Step-by-step process for gathering inputs and producing output
• Team: Natural collaboration patterns following the org chart

The CDAIO agent acts as orchestrator -- understanding your need and routing to the right team member(s). For complex requests spanning multiple domains, the CDAIO assembles a team and coordinates handoffs.

Why Claude Code (Not LangGraph, CrewAI, etc.)

Most agent frameworks make you write Python glue code, manage state machines, and deploy infrastructure. We chose Claude Code as the backbone because:

• Zero infrastructure -- no servers, no Docker, no API keys to manage. Install the plugin, type a slash command, done.
• Natural language agents -- each agent is a markdown file with personality, workflow, and expertise. No Python classes, no DAGs, no serialization headaches.
• Built-in tool use -- Claude Code natively reads files, writes deliverables, and searches codebases. No custom tool wrappers needed.
• Conversation as orchestration -- the CDAIO routes work by reading agent files and adopting their persona. No explicit state machines, no edges-and-nodes graphs. The org chart is the orchestration.
• Plugin distribution -- users install with one command. LangGraph requires them to clone a repo, install dependencies, configure environment variables, and run a server.
• Context window as memory -- all project context (client profile, past deliverables, industry data) lives in files that Claude reads naturally. No vector databases or retrieval pipelines.

The trade-off: you need Claude Code (and therefore an Anthropic API key or Claude subscription). But if you're already using Claude, this gives you a full CDAIO office with zero additional setup.

Roadmap

See ROADMAP.md for what's coming next.

Contributing

Want to add a new team member? See CONTRIBUTING.md.

Potential additions: Data Mesh Architect, MLOps Engineer, Data Product Manager, AI Ethics Lead, Data Literacy Trainer.

License

MIT License. See LICENSE.