Vector 1
Agentic AIOrganizations & Work

Announcing AI CEO — An Open-Source, Policy-Aware Executive System

Brian J. Curry
Brian J. Curry
September 13, 2025 · 2 min read

A modular, multi-agent “AI CEO” starter designed for real strategy, real guardrails, and real results.

TL;DR: I’m open-sourcing AI CEO, a production-minded starter that goes beyond copilots. It ships with a clean architecture, multi-agent orchestration, memory interfaces, guardrails/policy checks, a FastAPI service, CI, and infra stubs — so you can stand up a transparent, auditable AI executive system fast.

Repo: https://github.com/Bodhi8/AI-CEO

Why this project

The AI discussion keeps optimizing “individual contributors,” while executive decision-making largely stays manual. AI CEO is a public, remixable baseline that brings policy-checked automation to strategy, finance, and operations — with human-in-the-loop escalation where it matters. It’s a practical codebase for testing how far an “AI executive” can go — openly, transparently, and with guardrails.

What’s included (v0.1)

  • Orchestrator that plans tasks, routes to agents/tools, and logs outcomes
  • Agents: Strategy, Ops, Finance (easy to extend)
  • Guardrails / Policy Engine: pre/post action checks + audit hooks
  • Memory interface: pluggable vector-store abstraction
  • Tool registry: web-search & spreadsheet stubs (drop in your adapters)
  • FastAPI service + a simple CLI example
  • Docs: Architecture, Deployment, Roadmap, Contributing, Security, Governance
  • DX: Tests, Ruff/Black/Mypy, GitHub Actions CI, Docker + Terraform skeleton

Architecture at a glance

A FastAPI endpoint hands objectives to an Orchestrator, which retrieves context from Memory, proposes a plan via the Strategy Agent, executes guarded Tool calls, records decisions, and returns a structured report. You extend it by adding agents, tools, policies, and a real vector store.

Quickstart

python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -e .
uvicorn ai_ceo.server.app:app --reload
# POST {"goal":"Evaluate AI-driven growth strategy for Q4"} to http://localhost:8000/objectives

Roadmap (short)

  • 0.2: Human-approval endpoints, persistent vector DB adapter, email/calendar/file tools, Docker images
  • 0.3: Role-based policies, multi-objective planning, metrics/observability

How you can help

  • Add tool adapters (BI, sheets, ERP/CRM, finance data, risk engines)
  • Wire a real vector store (FAISS/Chroma/Weaviate/Pinecone)
  • Expand policies (SOX-style approvals, spend thresholds, sensitive-action gates)
  • Improve tests, docs, and examples; propose real-world evaluation harnesses

Support the project

If this work helps you, consider supporting it:

  • GitHub: Star the repo and, if you’d like to back ongoing development, use the Sponsor button on my profile/repo (https://github.com/Bodhi8).
  • Medium: You can leave a tip on this post or subscribe to get updates on major releases, deep-dives, and example integrations.

Why now? Leadership decisions shape everything downstream. If we’re serious about AI-driven efficiency, we need transparent, auditable executive automation — out in the open — so teams, boards, and researchers can test it, critique it, and improve it together.

Repo: https://github.com/Bodhi8/AI-CEO
 License: MIT

Related research