borghei/Claude-Skills / c-level-advisor/chief-ai-officer-advisor
chief-ai-officer-advisor
UnverifiedAI leadership advisor on AI strategy, governance, risk, investment, and org design. Use when defining an AI strategy, building an AI governance program, scoring AI maturity, or drafting an AI risk register.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.7 KB |
| references/ai-org-and-talent.md | markdown | 8.9 KB |
| references/ai-risk-and-governance.md | markdown | 12.5 KB |
| references/ai-strategy-framework.md | markdown | 8.8 KB |
| scripts/ai_investment_planner.py | script | 13.7 KB |
| scripts/ai_maturity_assessor.py | script | 16.8 KB |
| scripts/ai_risk_register_generator.py | script | 16.0 KB |
references/ai-org-and-talent.md
View on GitHub# AI Org & Talent Reference
Patterns for organizing the AI function, hiring the right roles in the
right order, and developing AI literacy across the company.
## 1. The AI org — three core groups
A mature AI organization usually contains three groups, regardless of
reporting structure:
### A) AI Platform
Owns infra, tools, and shared services that every applied team uses.
- Foundation-model gateway (multi-provider, caching, rate limit, cost attribution)
- Eval harness and benchmark management
- Model registry and approval workflow
- Vector store, retrieval pipeline, document processing
- Observability (latency, cost, quality, drift)
- Annotation and data-labeling tooling
- Fine-tuning and training infrastructure (if needed)
### B) Applied AI / Embedded ML
Owns delivery — usually embedded in product or business teams.
- Use-case definition with product/BU
- Model selection (within approved set)
- Prompt design, retrieval design, agent design
- Eval design for the use case
- Application UX (often paired with product engineering)
- Adoption and value measurement
### C) AI Governance + Risk
Owns the trust posture.
- Policy and standards
- Model approval gates and risk review
- Incident response process
- Regulatory tracking (EU AI Act, sector-specific)
- Third-party AI vendor reviews
- AI literacy program
These are roles, not necessarily separate teams. In a 50-person company,
all three may live inside one team of 5–10. In a 5000-person company,
they're separate functions reporting to the CAIO.
## 2. Hiring sequence (by org size)
### Stage 1 — < 100 engineers
Hire **one senior applied ML engineer** with breadth (eval, prompt, retrieval,
deployment). They double as the platform lead and the governance lead.
Common mistake: hiring an ML researcher when you need someone who ships.
### Stage 2 — 100–500 engineers
Add:
- A **second applied ML engineer** to start the embedded model
- A **platform engineer** (infra background; doesn't need ML depth) for the gateway, observability, cost
- A **part-time AI policy lead** (often inside legal/compliance, dotted-line to AI)
You don't yet need a CAIO — a head of AI or principal engineer can carry the weight.
### Stage 3 — 500–2000 engineers
Add:
- A **head of AI** (or CAIO at this stage) with full P&L for the function
- **Embedded ML leads** per major product area (1 per 50–80 engineers in that area is a healthy ratio)
- A **dedicated AI governance lead** (full time)
- A **data quality / data engineering** counterpart
- **Annotation operations** lead
### Stage 4 — 2000+ engineers
Add:
- **Platform manager** (separate from head of AI)
- **Red team lead**
- **Vendor management** specialist
- **AI security** engineer (paired with CISO org)
- **AI product manager(s)** at the platform and at major product areas
## 3. Role definitions
### Chief AI Officer
- Owns AI strategy, portfolio, and risk
- Peer to CTO, CIO, CISO; reports to CEO (preferred) or COO/CTO
- Operates through the AI Council + Model Risk Board
- Outside view (regulators, partners, customers) is part of the role
### Head of AI Platform
- Owns the gateway, eval harness, registry, infra
- Sets standards every applied team consumes
- Manages foundation-model vendor relationships
- Manages platform cost and capacity
### Applied ML / GenAI Engineer
- Builds production AI features end-to-end
- Designs evals; iterates prompts/retrieval; ships
- Pairs with product, design, and backend engineers
- Owns post-launch monitoring with platform
### ML / Data Scientist
- Predictive modeling, classical ML, decision support
- Stronger statistics + experimentation than GenAI engineers
- Often partners with analytics for impact measurement
### MLOps / Platform Engineer
- Infra, CI/CD, observability, cost
- Backend mindset; doesn't need to train models
- Owns reliability of training and serving
### AI Governance / Risk Lead
- Owns risk register, AIIAs, approval workflow
- Liaises with GC, privacy, CISO, regulators
- Builds the literacy program
### AI Red Team Lead
- Adversarial evaluation, jailbreak testing
- Threat modeling for new systems
- Maintains the adversarial test set in the eval harness
### AI Product Manager
- Use-case prioritization, requirements, eval criteria
- Adoption strategy
- Roadmap and stakeholder alignment
## 4. Compensation realities (2026)
Rough comp ranges in US-tier markets, for calibration only:
| Role | Base | Total comp (approx.) |
|------|------|----------------------|
| Sr applied ML engineer | $200k–$280k | $350k–$500k |
| Staff applied ML engineer | $260k–$360k | $500k–$800k |
| Head of AI (post-Series C) | $300k–$420k | $600k–$1.1M |
| AI governance lead | $200k–$280k | $300k–$450k |
| MLOps engineer | $180k–$260k | $300k–$500k |
| CAIO (mid-stage) | $400k+ | $800k–$2M+ equity-heavy |
European markets generally run 50–70% of US numbers. Frontier-lab competition
inflates the top end.
## 5. Build vs buy — the talent angle
You **can't out-hire** the frontier labs for foundation-model research talent.
What you can hire for:
- **Applied** engineers who understand your domain, data, and product
- **Eval-driven** practitioners who write tests before code
- **Platform** engineers who can run reliable AI infra at your scale
- **Governance** practitioners with regulated-industry experience
Stop trying to hire researchers unless you have a genuine research thesis
and a 5-year capital runway for it.
## 6. AI literacy across the company
A practical curriculum, tiered by role:
### Tier 1 — All employees (60–90 min)
- What AI tools we sanction and why
- What you can and can't put into a prompt (acceptable-use)
- How to escalate a suspected incident
- Where to find the AI policy
### Tier 2 — Engineers and product (4–8 hours)
- Foundation models 101 (capabilities, limits, evals)
- The shared eval harness and how to use it
- Prompt + retrieval patterns from our cookbook
- How to ship a low-risk AI feature
- When to escalate to the Model Risk Board
### Tier 3 — ML practitioners (ongoing)
- Internal eng meeting weekly
- Externally: paper reading group, conferences, open-source contribution
- Annual offsite for the AI org
### Tier 4 — Execs and managers (2 hours, refreshed quarterly)
- Strategy themes and our position
- Risk taxonomy and what to escalate
- The board narrative; how to talk about AI externally
## 7. AI org failure modes
### "Innovation theater"
Big AI team, lots of demos, no production systems.
**Fix:** publish a public production-system count and AI-attributable
business metric. If neither moves, restructure the team around delivery.
### "Center of excellence" without delivery
A central AI team writes whitepapers, runs lunch-and-learns, never ships.
**Fix:** dissolve into embedded squads + a small platform group. The
"COE" pattern almost always fails.
### "Shadow AI" everywhere
Every BU has its own LLM tool, its own vendor, its own (or no) governance.
**Fix:** offer a sanctioned, fast-path option that's better than what
they'd build themselves. Lead with carrots; use sticks selectively.
### Single point of expertise
One staff engineer knows the entire system. They leave.
**Fix:** insist on internal docs, runbooks, eval transparency, and
shared on-call from day one.
### Wrong manager profile at the top
The head of AI is a researcher who hasn't shipped production software.
**Fix:** the head of AI must be a builder. If your top candidate has
no production deployment experience, hire a strong COO/EM as the #2.
## 8. AI / human work — what the team does NOT do
Be explicit. Common boundaries:
- **AI is not a replacement** for engineering judgment in approval flows
- **AI is not a hiring decision-maker** (even if it scores candidates)
- **AI is not a sole compliance source** (it informs; humans decide)
- **AI is not a replacement** for security review on its own outputs (prompt-injection world)
Publishing these explicitly head off both fear and over-reliance.
## 9. Career ladders — what to expect
A typical Sr → Staff → Principal ladder for applied ML/GenAI roles emphasizes:
- **Sr:** ships features end-to-end; runs evals; owns one production system
- **Staff:** sets technical direction for a domain (e.g., search, agents); mentors
- **Principal:** organization-wide impact; drives platform standards; external presence
Avoid creating a separate "AI ladder" — it ghettoizes the function and
limits mobility. Use the engineering ladder with AI-specific examples.
## 10. Outsourcing the right pieces
Things that outsource well:
- **Annotation and labeling** (with your QA layer)
- **Penetration testing and red-teaming** (engaged quarterly)
- **Third-party audits** (SOC 2, ISO 42001)
- **Domain-specific fine-tuning** for adjacent capabilities (e.g., a vendor's medical-coding model)
Things that don't outsource well:
- **AI strategy** (too company-specific)
- **Use-case prioritization** (requires deep product context)
- **Governance program ownership** (you're accountable; agents help, not own)
- **Eval design** for your differentiating use cases (encodes your IP)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |